EVXL — Electric Vehicle & Tesla News

Volkswagen Chased A Flying Car In China For Five Years And Left Its Project Lead Facing A Criminal Case

Haye Kesteloo — Sat, 01 Aug 2026 20:45:00 GMT

Volkswagen spent five years and three full-scale prototypes trying to build a flying car in China, and killed the program in June 2024. Reuters reconstructed the whole thing this week from hundreds of pages of internal documents and court records.

The wreckage did not stop at a written-off research line. A Shanghai startup accused Volkswagen of misappropriating its trade secrets, police opened a criminal investigation in December 2023, and the person they named as the suspect was not Volkswagen. It was Zhou Jin, the Chinese manager Volkswagen had promoted in its own videos as the face of its innovation program.

I wrote in October 2024 about Ralf Brandstätter warning Volkswagen’s supervisory board that Chinese rivals were leaping ahead. This file is what that warning looked like from the inside, one product line over.

Volkswagen Built A Four-Seat Flying Car For China’s Wealthy Buyers

Volkswagen launched the project in 2019 with a team of fewer than ten people in Beijing, chasing an electric vertical takeoff and landing aircraft branded V.MO. Internal marketing materials named the target customer as China’s “established life enjoyers,” wealthy individuals buying status and exclusivity.

The cabin had to seat four passengers with luggage and carry entertainment screens. The pitch was intercity travel, Beijing to Tianjin, at roughly half a helicopter’s noise. It was a moonshot that had already defeated two of the group’s own brands: Porsche partnered with Boeing, Audi partnered with Airbus, and neither venture produced a commercial aircraft.

Zhou, then the 39-year-old director of Volkswagen’s Beijing innovation center, led the team. Volkswagen concluded it needed a Chinese partner because the local market was running ahead of European research and development, a judgment Zhou delivered on camera in a company promotional video.

The Partner Search Ran Straight Into China’s State Defense Contractor

Finding a capable Chinese aerospace partner without political exposure turned out to be the first unsolvable problem, and one internal presentation warned explicitly that government ties could trigger scrutiny outside China. Volkswagen worked the list and kept hitting walls.

An August 2020 assessment concluded that Guangzhou-based EHang had a gap to close against Volkswagen’s automotive engineering standards. Military drone maker Sichuan Tengden withdrew on its own, citing capacity constraints that a Volkswagen note connected to the China-US arms race. Volkswagen’s lawyers separately flagged that Tengden was founded by a veteran of Aviation Industry Corporation of China, the state defense contractor that builds the J-20 stealth fighter.

Volkswagen then hired an AVIC subsidiary, AVIC General Huanan Aircraft Industry Company, as its flying-car consultant in 2021. AVIC faced US export restrictions at the time, though it was not a fully blocked entity under US sanctions. That decision is where the legal trouble starts.

A $590,000 Contract Turned Into A $30 Million Lawsuit

That same year Volkswagen picked Pantuo Aviation, a two-year-old Shanghai startup run by Zhang Qiong, for a feasibility study on a Volkswagen-badged aircraft. Pantuo’s four-rotating-wing concept, known as Pantala, impressed Volkswagen managers. Zhang was less impressed in the other direction, telling Reuters the automaker treated a novel aircraft program like a parts-assembly job.

“They had never done this before,” she said.

A September 2021 Volkswagen status document declared the concept carried “uncontrollable risk,” with aerodynamic performance leaving projected range far short of target. A later internal assessment called the Pantuo collaboration a mistake and blamed Volkswagen’s own lack of in-house aviation experience. Volkswagen offered to terminate the roughly $590,000 contract for about $414,000. Pantuo refused and went to arbitration, arguing Volkswagen had improperly shared its concept work with the AVIC unit.

The arbitrator sided with Volkswagen in April 2023 and ordered Pantuo to pay damages and fees above $120,000. Volkswagen’s legal team called the ruling an outstanding result in an internal email. Pantuo escalated in September 2025 with a $30 million suit in a Guangdong court against Volkswagen and AVIC GA. China’s Supreme People’s Court dismissed the claim against Volkswagen in June on arbitration grounds while allowing the case against the AVIC unit to proceed, which leaves Pantuo free to return to the arbitrator.

Zhou Jin Is The Only Individual Carrying Criminal Exposure

Pantuo filed its criminal complaint in 2021 against Volkswagen. When Shanghai police opened the investigation in December 2023, case documents named Zhou as the suspect, and prosecutors took the matter up in 2024. She has been under travel restrictions since.

Zhang has confirmed she filed against the company, not the manager. Zhou told Reuters she cannot understand being left alone to face “a criminal case I was innocently dragged into,” and said she has repeatedly asked Volkswagen to resolve the underlying commercial dispute in full. She called it shocking and heartbreaking that Volkswagen treats the case against her as an individual legal matter, a position she says senior executives restated to her as recently as July.

Volkswagen declined to comment publicly on the flying-car program or the case against its employee, saying only that the claims against the company are “entirely without substance.”

China Is Now The Plaintiff In Technology Disputes

For decades the anxiety ran one way, with foreign firms worried that Chinese partners would copy their technology. The Volkswagen case is the inversion, and the numbers say it is not an outlier.

Chinese courts accepted 11,066 intellectual-property cases involving foreign parties in 2025, up 34 percent from the prior year, according to a Supreme People’s Court report. Foreign-related matters at China’s Intellectual Property Court grew at an average 19 percent a year between 2019 and 2025, reaching 2,546 cases. In February, the national agency overseeing prosecutions said it was stepping up enforcement on technology-leakage claims.

Mark Cohen, former senior intellectual-property attaché at the US embassy in Beijing, framed the shift in seven words: “The shoe is on the other foot.” As China owns more of the technology worth defending, he said, it becomes the plaintiff more often.

Brandstätter Killed It, And The Partner Bought Volocopter

Volkswagen’s top management committee reviewed the program in March 2024 and green-lit it, but left the real decision to China chief Ralf Brandstätter. Legal, compliance, finance and strategy had all raised objections, and the strategy paper warned that momentum in China was already held by established local players, naming Xpeng and Geely.

Brandstätter ended it that June, concluding Volkswagen needed to focus on a core business whose passenger-car sales were sliding. He told the team it was disbanded and broke the news to Chinese partner Wanfeng Auto Holding Group by WeChat. A Wanfeng unit disclosed in a November 2024 regulatory filing that the joint venture agreement had ended.

Wanfeng did not slow down. In March 2025 its Austrian subsidiary Diamond Aircraft acquired the assets of insolvent German eVTOL developer Volocopter for about €10 million. Emerson Xu, chief executive of consultancy NexAvian and the former China head of that same Volocopter, told Reuters that firms lacking local speed “don’t stand a chance.”

The core business tells the same story. Volkswagen’s China deliveries peaked at 4.2 million in 2019 and fell to 2.7 million last year, dropping it to third behind BYD and Geely. Volkswagen has since announced plans to cut vehicle models and up to 100,000 jobs. To be fair to Wolfsburg, the Chinese market has punished its own champions too, as BYD’s own profit collapse and the 2026 subsidy cliff showed. But VW is losing to Geely, the same Geely that already builds cars in South Carolina, and to a BYD that still holds the global BEV crown even in a shrinking quarter.

EVXL’s Take

Volkswagen sold “in China, for China” as risk management. The flying car shows the opposite. Localization concentrated Volkswagen’s exposure rather than spreading it, and the file reads like a company that thought hiring Chinese engineers and Chinese consultants was the same thing as acquiring Chinese speed. It is not. Zhang put her finger on it precisely: Volkswagen approached a novel aircraft the way it approaches a car, as parts that exist and get bolted together. That is a manufacturing mindset applied to a research problem, and it is the same instinct that produced the ID. software mess.

The detail that should bother every Western executive with a China operation is not the failed prototype. It is Zhou Jin. Volkswagen put her in its promotional videos as the face of Chinese innovation inside a German company, and when the legal exposure arrived, the company’s position was that her criminal case is her own matter. Whatever the legal merits, that is a recruiting problem in a market where Volkswagen needs local engineering talent to choose it over Xpeng or NIO. Every senior Chinese hire at a Western automaker read this story this week.

I have argued for two years that legacy automakers are failing on strategy rather than on the EV transition itself, and this is the purest example yet. Nobody sanctioned Volkswagen out of the flying-car market. It had money, access, a Premier photographed in its mock-up and a five-year head start on most of its Chinese competitors. It lost anyway, to companies that moved faster and to a legal system that now defends domestic technology with criminal referrals.

Watch what Volkswagen does with Xpeng from here. That partnership carries the EV turnaround, and Xpeng’s Aridge unit is simultaneously one of the flying-car competitors Volkswagen’s own strategy paper named as the reason to quit. Wolfsburg is now dependent on a partner that beat it in an adjacent market, and it continues to lean on Chinese firms elsewhere, including Horizon Robotics for autonomous driving. Whether that dependence deepens or quietly narrows over the next year is the real signal about how much leverage Volkswagen has left in China.

Sources: Reuters, Supreme People’s Court of China, Wanfeng Auto Holding Group regulatory filings.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

BYD Sales Rise 21.8% in July as Exports Hit a Record, but the 5 Million Target No Longer Adds Up

Haye Kesteloo — Sat, 01 Aug 2026 17:05:00 GMT

BYD sold 419,211 vehicles in July, its strongest month of 2026 and a third consecutive month of year-over-year growth. The number underneath that headline matters more: 179,841 of those cars went to buyers outside China, the highest monthly export volume the company has ever posted.

Reuters calculated the totals from BYD’s official statements released Saturday. The year-over-year gain came in at 21.8%, a sharp step up from June’s 5.46% and May’s 0.26%.

I have followed these monthly filings through eight straight months of decline, a February collapse of 41.1%, and now three months of recovery. The recovery is real. It is also happening almost entirely outside China, and BYD’s own full-year guidance stopped working somewhere around the end of June.

Exports Now Account for 42.9% of Everything BYD Sells

BYD shipped 179,841 passenger vehicles and pickups outside China in July, up 124.3% from a year earlier, according to Reuters calculations drawn from the company’s official statements. Those cars made up 42.9% of total sales, a fraction below June’s 43.46% share, because domestic volume finally moved in the right direction.

One caveat belongs on that number. These are wholesale shipments, not registrations. A record export figure tells you what left the port, not what found a driveway, and Chinese manufacturers have been scrutinised before over channel inventory and cars exported as nominally used vehicles.

The monthly export sequence tells a more complicated story than the year-over-year percentage suggests. April brought 135,098 units abroad, May 160,644, June 175,349, and July 179,841. The sequential gains have shrunk in each of the last three months: 25,546, then 14,705, then 4,492. The annual comparison looks explosive because July 2025 was a weak base. The month-to-month momentum is flattening.

Cumulative overseas sales through seven months now stand at 972,097 units, based on BYD’s reported first-half figure of 792,256 plus July. That already puts the company most of the way to the 1.5 million export goal it raised from 1.3 million earlier this year.

By brand, the core BYD marque accounted for 350,178 passenger units in July. Fang Cheng Bao contributed 41,213, Denza 19,196, and Yangwang 485. Those four add to 411,072, with the remaining 8,139 vehicles coming from the commercial side of the business.

BYD’s Home Market Is Still Shrinking, Just More Slowly

Subtracting overseas shipments from the global total puts BYD’s July domestic sales near 239,370 vehicles, down roughly 9% year over year. That is still a contraction, but it is the shallowest one the company has recorded at home since the slide began, and a long way from June’s 22.02% drop.

The damage traces back to January 1, when China halved its new energy vehicle purchase tax exemption from a maximum waiver of 30,000 yuan ($4,200) to 15,000 yuan ($2,100). Domestic sales fell 39.57% in the first half to 1,016,255 units. In March I wrote that BYD’s home market was deteriorating faster than the headline numbers suggested, when February domestic volume collapsed 65% and, for the first time on record, came in below overseas shipments.

The price war has been expensive. BYD’s first-quarter net profit dropped 55% to 4.08 billion yuan ($561 million), its weakest in more than three years, a figure we covered alongside the company’s second-quarter crown over Tesla.

The 5 Million-Vehicle Guidance No Longer Survives Contact With Arithmetic

BYD guided investors to 5.0 million to 5.5 million new energy vehicles for 2026. Through seven months it has sold 2,227,722. Clearing even the bottom of that range requires 554,456 vehicles a month for the remaining five months, which is 32% above July’s best-of-year total.

Put that in context. BYD’s best quarter on record is the fourth quarter of 2024, when it sold 1,524,270 vehicles, an average of about 508,090 a month. The pace now required to reach the low end of guidance is higher than the best three-month stretch the company has ever managed, and it would have to hold for five months while the lineup finishes its retrofit to the second-generation Blade Battery.

The obvious defence is seasonality, since BYD always sells more in the back half of the year. It does not rescue the number. In 2025 the company’s second half ran 14.5% ahead of its first. Apply exactly that uplift to this year’s first-half total of 1,808,511 and BYD lands near 3.88 million for 2026, still more than a million vehicles short of the bottom of its own range.

The export target is the opposite problem. With 972,097 units already shipped, BYD needs only 105,581 a month to reach 1.5 million. It is running at 179,841. If July’s rate held through December, overseas sales would finish near 1.87 million, well past the goal.

One of those two numbers is going to get revised. It will not be the export one.

BYD’s Japan Launch Shows What the Export Push Actually Looks Like

BYD launched the Racco, an all-electric kei car built specifically for Japan, on July 28. It starts at 2,145,000 yen (about $13,100) and falls below 2 million yen after a 150,000-yen government subsidy, undercutting the Nissan Sakura that currently leads Japanese EV sales at roughly 2.44 million yen.

Kei cars account for close to 40% of new vehicle sales in Japan, and Suzuki, Daihatsu and Honda together hold about 80% of that segment. BYD is attacking it with a car it does not sell at home. The entry 200 trim carries a 22.4 kWh lithium iron phosphate pack rated at 210 km on the WLTC cycle. The 300 Plus and 300 Premium step up to 35.84 kWh and 320 km, which BYD says makes the Racco the first kei EV in Japan past the 300 km mark.

Here is the detail worth pausing on. DC fast charging on the Racco is capped at 50 kW. BYD spends enormous marketing energy at home on flash charging and megawatt-class claims, and the car it built to crack Japan tops out at a rate that would have looked ordinary in 2019. Export buyers are getting a different product than the domestic press releases imply, and that gap is the thing owners outside China should watch as BYD scales.

Geely’s July Numbers Show This Is an Industry Pattern, Not a BYD Story

Geely sold 250,161 vehicles in July, up 5.23% year over year, with record exports of 106,663 units and domestic sales down 29.12% to 143,498. The shape of that result matches BYD’s almost exactly: overseas volume carrying the company while the home market contracts.

Geely’s new energy exports reached 62,604 units in July, up 616% year over year and 59% of its total exports, which shows the overseas mix tilting away from combustion models. Cumulative Geely exports for the year hit 580,891, up 164.78%.

The comparison contains a reversal worth noting. In February, Geely was taking share from BYD inside China while BYD’s domestic volume fell 65%. In July, BYD’s domestic decline of roughly 9% is a third of Geely’s 29.12%. BYD is losing at home more slowly than the rival that was beating it six months ago.

Both are building the same escape hatch. BYD runs plants in Thailand, Uzbekistan, Brazil and Indonesia, with Szeged in Hungary moving toward series production, a build-out we tracked when BYD confirmed it would double its European dealer network to 2,000 locations. Geely agreed in July to take over part of Ford’s Spanish factory through a joint venture. Tariff walls are not stopping Chinese volume. They are relocating where it gets assembled.

EVXL’s Take

Every outlet will run the 21.8% number today. The honest framing is different: BYD has stopped being a Chinese automaker with a healthy export business and started becoming an exporter with a Chinese division attached. Nearly 43% of its volume now sells outside its home market, and the only reason the global figure grew at all is that foreign buyers made up for Chinese ones who stayed home.

The guidance is dead and somebody should say so plainly. Five million vehicles required 554,456 units a month from August through December. July, the best month of the year, delivered 419,211. That is not a stretch goal, it is a number the company will quietly stop mentioning. At BYD’s shareholder meeting in Shenzhen last month, Chairman Wang Chuanfu told investors he wants BYD to be the world’s largest automaker within five years. I would take that ambition more seriously if the company could hit a target it set seven months ago.

The export engine has its own warning light, though, and I want to be even-handed about it. Those sequential gains shrank from 25,546 to 14,705 to 4,492 in three months. If that curve keeps flattening, exports top out before they can carry a company this size, and then BYD has no growth story left in either hemisphere.

An accountability check on my own record. In that March piece I predicted BYD’s domestic sales would not recover to year-ago levels until the third quarter at the earliest. July is the first month of Q3, and domestic volume is still down about 9%. The call is holding, though I want to flag that it was a low bar to clear and the recovery is arriving faster than the pace I implied. If August lands within a few points of flat, I was too pessimistic and I will say so.

What bothers me is the Racco. BYD is winning abroad on price, partly by shipping hardware that would not headline a product launch in Shenzhen. A 50 kW charging cap on the car built to break the Japanese incumbents in their own backyard tells you where the cost cuts landed. Export buyers deserve the same scrutiny of range and charging claims we apply to every automaker, because the flash-charging marketing is not travelling with the cars.

The last piece is one I keep writing and will keep writing. Americans still cannot buy any of this. A 100% tariff wall keeps the cheapest competent EVs on earth out of the country, even as 40% of U.S. consumers say they would welcome Chinese brands. I am against subsidy distortion from Beijing and I am against protectionism from Washington, and right now Washington is the one keeping a $13,100 electric car out of American driveways. Japanese buyers get to decide for themselves this month. Americans do not.

Watch August 3. Denza opens pre-sales on the Z9S with a claimed range of up to 920 km. If BYD leads that launch with the range figure rather than the charging figure, you will know which number the company thinks it can actually defend.

Sources: Reuters, “BYD’s sales rise for third month, buoyed by exports,” by Che Pan and Ryan Woo, August 1, 2026. Also CnEVPost on Geely’s July results, CnEVPost on the BYD Racco launch in Japan, and BYD’s monthly sales statements.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Renault Returns To Profit With One Hard Rule: Every New EV Must Earn Like A Hybrid

Haye Kesteloo — Sat, 01 Aug 2026 01:01:00 GMT

Renault swung to a €721 million net profit in the first half of 2026 on revenue of €30.25 billion, up 9.5%, while selling 0.4% fewer vehicles than a year ago. Fewer cars, more money per car. That inversion, not any volume record, is the story of the French group’s turnaround.

Electric is doing the pulling. EV sales across Renault Group’s European operations jumped 47.6% and now make up 18.8% of its sales there, with electrified models at 52%. The growth came while BYD and Chery expand in the same small-car segments Renault depends on, and one year after a €9.3 billion Nissan-related writedown dragged the group to an €11.1 billion first-half loss. Strip out that one-off and underlying group-share profit still climbed €244 million.

The comeback rests on one internal rule, and it runs opposite to how most legacy automakers approached electrification.

Renault’s One Rule Makes Every New EV Earn Like A Hybrid

Renault now approves a new model only if its electric version is projected to earn the same margin as the group’s full hybrids, a gate CEO Francois Provost confirmed on Wednesday’s earnings call and one that leaves no room for loss-making compliance cars built to satisfy CO2 targets. For every new model decision, Provost told analysts, “the profitability of electric should be the same as full hybrid.”

The gate sits inside futuREady, the strategy Provost launched in March 2026. The cars carrying the growth are the Renault 5 E-Tech electric, the Renault 4 E-Tech electric, and the new sub-€20,000 Twingo E-Tech electric. Renault-brand BEV sales in Europe grew 63.2% in a market up 34.2%, and the R5 is the continent’s best-selling B-segment EV. The Douai plant built its 100,000th example in December 2025, 15 months after launch, and Renault expects the model to pass 200,000 cumulative units before the year is out.

The yellow Renault 5 Turbo 3E caps the range whose €25,000 base car carries the volume. Photo courtesy of Renault / Autocar

The clearest outside validation came from Detroit. Ford concluded last December that its own EV platforms were too expensive for Europe and hired Renault to build two small electric models on French underpinnings, with CEO Jim Farley calling it more competitive than the platform Ford had engineered in the U.S. When a rival pays you to use your platform, the spreadsheet argument is over.

Chinese Price Pressure Meets A 20.7% Electric Market

Battery-electric cars took 20.7% of EU registrations in the first half of 2026, up from 15.6% a year earlier according to ACEA, and Chinese brands such as BYD and Chery are scaling in the same small-car segments where Renault earns its living. Buyers registered 1,220,890 battery-electric cars in the EU over six months, up 40.5%, with France growing 62.9% and Germany 48%. The market Renault bet on has arrived. So has the competition chasing it.

Provost expects price pressure in Europe to run through the rest of the year, Reuters reports, and the strain is visible in Renault’s own accounts: price and mix effects cost the group €425 million in the half. The response was to protect value rather than volume. Retail channels made up 60% of passenger-car sales in Renault’s five biggest European markets, 17.7 points above the market average. Dacia volumes fell 8.7% in Europe after early-year logistics problems, a gap Renault chose not to buy back with discounts, in line with the pricing discipline Provost stressed on the call.

Compare that with Stellantis, which in February took a €22.2 billion charge to unwind its EV plans while leaning on budget models to hold the value end of the market. Two European groups faced the same Chinese pressure. One cut product. The other cut cost.

Cost Cuts Of €400 Per Car Fund The Pricing Defense

The rule only works because the cost base keeps shrinking: Renault hit its Renaulution target of cutting variable costs by €400 per vehicle in 2025 and now aims to remove roughly another €400 per car every year on average over the medium term under futuREady. The first half delivered on schedule: the cost program removed €184 million, with purchasing gains and lower warranty spending more than offsetting raw-material inflation. New projects now launch with supplier entry tickets as much as 40% below the previous generation.

None of this makes Renault bulletproof, and the honest reading of these numbers says so. Automotive operating margin came in at 3.0%, down from a year earlier, while Mobilize Financial Services, the group’s financing arm, contributed €753 million, roughly half of total operating profit. Net income attributable to the group, at €705 million, landed under the €770 million analysts expected. Renault itself warns the raw-material and inflation hit will roughly double to about €400 million in the second half. The machine works. It is not coasting.

EVXL’s Take

In January, when Provost folded the Ampere EV unit back into the mothership, I asked whether any European automaker had a viable path against BYD, and six months later Renault has delivered the closest thing to an answer I have seen. Detroit answered the same question with a checkbook: we have tracked roughly $53 billion in EV write-downs across the Big Three since October, capital spent un-committing from electrification. Renault spent the same stretch stripping €400 a car out of its build costs and shipping a small EV people actually line up for.

One approach treats Chinese competition as a reason to retreat behind tariffs and lobbying. The other treats it as a spec sheet to beat. EVXL has argued from the start that unfair Chinese subsidies deserve pushback, but a tariff wall that lets legacy automakers keep bloated cost structures protects executives, not buyers. A mid-size French automaker just grew EV sales 47.6% against BYD and Chery on the same tariff terms every rival already enjoys. Anyone still begging Brussels for more protection should have to explain that number first.

I am not declaring victory for Renault. A 3% automotive margin and a financing arm carrying half the profit add up to pricing power, not immunity. The real test of the hybrid-parity rule is the cheapest car in the range: the sub-€20,000 Twingo scales through a second half in which Renault has already flagged a doubled raw-material bill. Watch the full-year results next February. If automotive margin holds with Twingo volume in the mix, the rule survived its hardest case. If Mobilize is still doing half the lifting, the cheering gets premature.

The EV transition does not need rescuing from China. It needs more automakers willing to do what Renault just did: the slow, unglamorous work of building electric cars people want at a cost that earns money.

Sources: Renault Group, Reuters, ACEA, Motor1

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Zoox Wins NHTSA Approval for Paid Robotaxi Rides

Haye Kesteloo — Thu, 30 Jul 2026 18:38:31 GMT

The National Highway Traffic Safety Administration announced federal approval Thursday for Zoox to charge passengers for rides in its purpose-built electric robotaxis, giving the Amazon-owned company a legal route to commercial service without a steering wheel, pedals, or other controls meant for a human driver.

The temporary exemption covers up to 2,500 vehicles annually for two years and is scheduled to take effect Friday, July 31, when the grant notice is published in the Federal Register. Zoox says it will start collecting fares in Las Vegas in August after completing remaining state requirements. State and local approvals remain necessary wherever the vehicles operate.

This is the first NHTSA commercial exemption for a purpose-built robotaxi. It turns Zoox’s free passenger program into something closer to a real transportation business, while keeping the fleet under federal limits and oversight.

NHTSA Gives Zoox Two Years Under Adaptive Oversight

NHTSA’s temporary exemption allows Zoox to deploy as many as 2,500 vehicles during each of two 12-month periods, while preserving federal authority to revise oversight as commercial data arrives, making this a capped test of a new vehicle architecture rather than blanket approval for unlimited production.

That annual 2,500-vehicle limit comes from the temporary-exemption authority in Part 555 of federal law. Zoox requested relief from portions of eight Federal Motor Vehicle Safety Standards, covering windshield defrosting and wiping, lighting, rear visibility, brakes, interior-impact protection, glazing, and occupant crash protection.

Calling the robotaxi “non-compliant” without context would be misleading. Several rules assume a human sits in a forward-facing driver’s position and needs a windshield, mirrors, pedals, and controls. Zoox instead built a bidirectional carriage with two inward-facing rows of seats and no driver’s position at all.

NHTSA said the exemption sits inside an adaptable enforcement framework rather than replacing safety review. Administrator Jonathan Morrison described the agency’s approach as removing unnecessary barriers while retaining enforcement oversight and developing measurable automated-vehicle performance requirements.

Paid Las Vegas Rides Are Scheduled For August

Zoox says Las Vegas will become its first paid market in August, after the company completes Nevada’s remaining commercial requirements, while San Francisco remains subject to a separate California process because the federal exemption clears specified equipment rules but does not authorize paid operations in every city.

The company launched free public rides in Las Vegas last year and an Explorers testing program in San Francisco. It now claims more than half a million completed riders and another half million people on its waitlist. Those figures come from Zoox and have not been independently audited.

The distinction between federal and local authority matters. NHTSA can exempt the vehicle from specified federal equipment standards, but Nevada and California still control commercial operations on their roads. A federal yes removes the vehicle-design barrier. It does not erase operating permits, service-area limits, or local enforcement.

A Purpose-Built Robotaxi Finally Gets A Commercial Path

Zoox’s approval gives the United States its clearest regulatory path yet for a passenger vehicle designed around autonomy rather than converted from a conventional car, because NHTSA has now identified how a purpose-built vehicle can enter paid service under a temporary exemption tied to continuing federal review.

EVXL previously covered NHTSA’s National AV Safety Forum, where regulators focused on remote assistance, measurable safety performance, emergency response, and the move from limited demonstrations to commercial scale. This exemption is the first concrete commercial decision to emerge from that broader policy push.

NHTSA is also working with SAE Industry Technologies Consortia on a three-year, $5 million program to develop automated-vehicle performance standards. The agency says those standards could eventually replace some vehicle-by-vehicle exemptions with national rules written for vehicles that drive themselves.

That is the right long-term destination. Temporary waivers can get a small fleet on the road, but an industry cannot scale around one-off permission slips forever.

EVXL’s Take

Zoox has reached the kind of robotaxi milestone that matters: federal permission to collect fares in a purpose-built vehicle, earned through a documented exemption process that names the rules its design cannot literally satisfy and limits the deployment while regulators gather evidence from commercial operation.

The company did not rename a supervised driver-assistance system, stage a short demonstration, or ask investors to accept another future promise. It submitted the vehicle to a federal process and accepted a capped deployment under continuing oversight.

The exemption also exposes the weakness in Washington’s old rulebook. Requiring a steering wheel in a vehicle with no driver’s seat does not protect anyone by itself. Requiring Zoox to demonstrate equivalent safety for the exempted standards, report how the fleet behaves, and operate under continuing federal oversight does. Regulation should police outcomes and evidence, not preserve hardware whose only job was to serve a human driver.

Zoox still has to prove that paid service works outside a controlled launch. Federal approval is the starting gun, not the finish line. But once the order takes effect and local requirements are met, the company will hold something many robotaxi announcements lack: permission to collect money from passengers in a vehicle built without human controls.

Sources: National Highway Traffic Safety Administration, Zoox, and the NHTSA grant notice.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's Cybertruck Tooling Is Held By A Supplier That Just Lost A $30 Million Appeal

Haye Kesteloo — Tue, 28 Jul 2026 18:00:00 GMT

The supplier refusing to hand back Tesla’s Cybertruck tooling walked into that standoff with a courtroom record already attached: a six-year court fight that ended in a $30 million judgment, a subsidiary that went through Chapter 11 because of it, and a federal appeal it lost five months before telling Tesla the Troy, Texas plant would close. EVXL reviewed the court records behind Angstrom Automotive Group’s recent legal history, and they recast the tooling fight EVXL first reported on July 25, 2026.

The Sixth Circuit affirmed the Eaton Corporation verdict against Angstrom on February 13, 2026. Angstrom notified Tesla on July 13, 2026, exactly five months later, that it would shut the Troy facility. Between those two dates sits the question every Cybertruck buyer waiting on a delivery now has an interest in: whether the company holding the dies can afford to give them back.

Angstrom Fought Eaton For Six Years And Lost $30 Million

Angstrom Automotive Group spent six years fighting power management company Eaton Corporation over broken clutch levers, lost a $30 million jury verdict in June 2024 in the U.S. District Court for the Northern District of Ohio, and then lost the appeal this February. The case began long before Tesla entered the picture. Eaton filed suit on April 24, 2020, over levers that Angstrom subsidiary Wrena, LLC supplied for Eaton’s electric clutch actuation systems, after field failures Eaton traced to manufacturing defects.

The trial opened June 3, 2024, before Judge Bridget Meehan Brennan, and the jury’s verdict became public on June 17. Jurors found both Angstrom and Wrena liable for breach of contract and for breaches of express and implied warranty, and rejected the companies’ counterclaims against Eaton.

One detail from the appellate record deserves attention in light of the Tesla dispute. Angstrom built the levers strictly to Eaton’s specifications and carried no design responsibility. That is the same build-to-print arrangement Tesla describes in its own complaint, where Tesla owns the designs and the tooling and the supplier runs the presses.

The Verdict Pushed Subsidiary Wrena Into Chapter 11

Wrena, LLC, the Angstrom subsidiary that physically made the clutch levers, filed for Chapter 11 protection in the U.S. Bankruptcy Court for the Eastern District of Michigan on September 23, 2024, listing liabilities between $10 million and $50 million against assets of no more than $10 million. The company’s announcement cited an isolated warranty claim that produced a large judgment, alongside rising costs and market pressures, and its chief restructuring officer called a court-supervised sale “the best path forward.”

The sale process the bankruptcy court approved on October 28, 2024 opened with a $5.65 million cash bid. That bid came from a related entity, an arrangement bankruptcy law permits and judgment creditors rarely enjoy, since it can keep the operating assets close to the family while the liability stays behind with the estate.

The Appeal Died Five Months Before The Troy Shutdown Notice

The U.S. Court of Appeals for the Sixth Circuit affirmed the full $30 million judgment against Angstrom on February 13, 2026, closing off the company’s last realistic path to erasing the Eaton liability just as its newly acquired Texas die-casting plant was supplying Tesla’s Cybertruck line. The panel’s opinion runs through Angstrom’s pre-suit notice defense and expert challenges before disposing of the appeal in two words: “We affirm.” A footnote records that Wrena, already in bankruptcy, had voluntarily dismissed its own appeal along the way.

Five months after that ruling came the July sequence detailed in the complaint Tesla filed on July 23: the July 13 shutdown notice, a demand for $250,000 per week to keep the plant running, 700 finished parts that never shipped, and Tesla representatives turned away at the gate with law enforcement standing by on July 21. EVXL covered the emergency filing and its five-to-six-month tooling rebuild estimate when it landed.

The Anderton Purchase Closed Ten Months Before The Plant Closure

Angstrom acquired Anderton Castings, the aluminum die-casting company that ran the Troy, Texas plant, in a deal announced September 9, 2025 by sell-side advisor Angle Advisors, which described a supplier of structural castings for body-in-white applications and battery-electric propulsion components including inverter housings. The advisor’s announcement also lists a second Anderton facility in Monterrey, Mexico.

A clarification on timing is in order here. Bloomberg Law’s report described the Tesla relationship beginning when Angstrom acquired a company already supplying Tesla in early 2025, and EVXL’s July 25 article carried that framing. The deal advisor’s own announcement dates the Anderton transaction to September 9, 2025. The complaint should settle which acquisition and which date the court record relies on, and EVXL will update its reporting when it does.

Either way, the arc is short. Ten months separate the acquisition announcement from the shutdown notice. In between, the truck the plant feeds went through its own whiplash: the $59,990 introductory Cybertruck that lasted ten days and pushed delivery estimates into 2027, arriving after a 2025 in which Cybertruck sales fell 48.1% to 20,237 units. The plant Angstrom bought was built around a volume story that had already broken, and the order surge that might have fixed it arrived months before the padlock.

EVXL’s Take

On July 25, I wrote that a supplier willing to burn a relationship this publicly usually believes it has nothing left to lose. The court records now put numbers on that instinct. Angstrom carried an affirmed $30 million judgment into July, watched a subsidiary’s assets move through a bankruptcy sale, and spent 2025 buying plants anyway, including the one it is now closing ten months after the purchase.

My position: this is not a rogue supplier story, it is a supplier diligence story. Every fact above sat in public dockets before the tooling standoff began. Tesla single-sources castings its own filing says cannot be replaced for five to six months, through a parent whose last major customer dispute ended in a $30 million judgment and a subsidiary bankruptcy. A company built on just-in-time manufacturing ran a critical line through a counterparty with a hole that size in its balance sheet. The people paying for that choice are the buyers holding $59,990 order confirmations and 2027 delivery estimates.

Watch two filings. First, the ruling on Tesla’s emergency motion, which should come quickly given the posture. Second, Angstrom’s response. If that response includes a counterclaim for money Angstrom says it is owed, the purchase orders and volume forecasts behind the Cybertruck program go into a public record, under oath. I said on July 25 that those numbers are the story. Nothing in Angstrom’s paper trail changes my mind.

Sources: Bloomberg Law, Crain’s Cleveland Business, U.S. Court of Appeals for the Sixth Circuit, PR Newswire, Angle Advisors.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Boring Company Director Puts Real Numbers On The Vegas Loop, Including The One That Undercuts His Own Pitch

Haye Kesteloo — Tue, 28 Jul 2026 03:20:00 GMT

The Boring Company director Mike Baier spent Monday evening on X handing out the kind of specifics the tunneling firm rarely volunteers: the Vegas Loop moves about 40,000 passengers on a busy convention day, the airport connector tunnels open in a few months, and ridership should triple or quadruple when they do.

Baier, the company’s Director of Production, Electrical Engineering, and Software Engineering, made the claims in a thread posted at 6:48 p.m. Eastern on July 27 that drew 35,000 views and 1,600 likes within hours. He kept answering questions well into the night, covering tunnel costs, boring speeds, Full Self-Driving coverage, and fares of $4.50 per ride or $13 for a day pass.

The peak-day figure, he argues, beats most American light rail systems. He’s right. He’s also comparing his best day to everyone else’s average one, and that gap is the real story.

The Airport Connector Timeline Keeps Sliding To The Right

The 2.25-mile Airport Connector twin tunnels, which Boring Company president Steve Davis targeted for a first-quarter 2026 opening back in December, remain under construction in late July, and Baier’s new guidance of a few months pushes the opening toward the end of the year at best. Once open, he expects ridership to grow 3x to 4x, which would put the Loop, by his math, in the top 10 of US public transportation systems and eventually the top 3 or 4.

The Las Vegas Review-Journal reported in December that Davis promised airport rides in the first quarter, with the system handling 17,000 to 20,000 people per hour once the line opened. The company has been running above-ground airport trips through partial tunnel routes since late 2025 while it finishes digging. As of a July 18 Review-Journal report, the company was still boring toward a site just north of Harry Reid International Airport, with a direct airport station under discussion but not yet agreed.

Baier called the connector twins two of the longest transportation tunnels built in the US in the past 25 years. The full approved build-out remains 68 miles of tunnel and 104 stations across Clark County and the City of Las Vegas, against roughly a handful of miles open today through 11 stations.

Baier Concedes The Machines Run At A Quarter Of Their Paper Speed

Tunneling costs the company between $5 million and $10 million per mile depending on ground conditions, Baier wrote, and its Prufrock machines currently manage about a mile per month against a design spec of roughly a mile per week, an admission of a 4x gap between marketing and reality that executives in the Musk orbit almost never make in public. The fix, in his telling, is logistics and reliability work, not a redesign.

That candor matters because the mile-a-week number is load-bearing across the company’s expansion pitch. When EVXL covered the start of Nashville Music City Loop tunneling on February 25, the Prufrock-MB1 machine’s target of better than a mile per week was the basis for the company’s aggressive Q1 2027 opening estimate on a 13-mile route. At the mile-a-month pace Baier now describes as actual, that math changes by years, not weeks.

Baier also argued the machines are reusable, so every unit that leaves the line adds annual capacity, and he sees a path to hundreds of miles per year sooner than skeptics expect. One reply asked for a machine per week. His answer: on it.

FSD Covers Only Some Routes While Cybercabs Wait Next Door

A rider who took six Loop trips this week reported that none used Tesla Full Self-Driving, and Baier’s explanation was that FSD simply doesn’t support every route yet, nine months after the company began supervised FSD trials in the tunnels in October 2025. Route-by-route rollout in a closed, mapped, single-operator tunnel system is a slower crawl than the environment would suggest.

The obvious next question came up too: when do Cybercab robotaxis enter the tunnels? That one had no visible answer from Baier in the thread, but the pieces are moving around him. Tesla’s July 22 Q2 earnings named Las Vegas as one of the next two robotaxi metros, as EVXL noted in today’s Charlotte staging report, and the company has 245 finished Cybercabs parked at Giga Texas looking for somewhere to earn a fare. A geofenced tunnel network in a robotaxi launch city is about the friendliest deployment surface that inventory could ask for.

EVXL’s Take

Baier’s thread is the most useful public accounting of the Vegas Loop in months, and I want to give him credit for the mile-a-month admission before I take the rest apart. A director at a Musk company voluntarily stating that real-world output runs at a quarter of the design spec is rare honesty, and it’s worth more to me than any ridership boast.

Now the boast. The 40,000-rider figure, which Baier wrote is “more than the majority of light rail systems in the US,” checks out on its face: Houston’s METRORail, the tenth-busiest light rail system in the country, averages about 35,100 weekday riders, so a 40,000-passenger day genuinely beats the average weekday of most American light rail. But that’s the trick. Convention-day peaks are the Loop’s Super Bowl, and Baier is scoring them against everyone else’s Tuesday. The Boring Company’s own project page counts more than 4 million total passengers through 11 stations since the system opened in June 2021. Spread over those five years, that averages out to roughly 2,200 rides per day, and even weighting the count toward the busier recent years, the typical day sits in the low thousands. On its typical day, the Loop wouldn’t crack the top 25.

None of that makes the Loop a failure. It makes it a convention shuttle with genuine peak capacity and a thin base load, which is exactly what a system anchored to a convention center should look like, and the airport connector is the one addition that could actually change the base-load math. That’s why the slipping timeline is the number to watch, not the ridership. Davis said Q1 2026. It’s late July and Baier says a few months. Given that track record, I’ll take the under: paying passengers ride the full underground airport route closer to Q1 2027 than to Halloween. When those tunnels open, the ridership claims stop being marketing math and start being measurable. I’ll be checking.

Source: Mike Baier on X, Las Vegas Review-Journal, The Boring Company

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Cybercabs Surface in Charlotte, Giving the Giga Texas Pile Its First Visible Destination

Haye Kesteloo — Mon, 27 Jul 2026 22:00:00 GMT

Fourteen Tesla Cybercab robotaxis and 14 Model Y vehicles, all wearing Texas plates, turned up in a Charlotte-area parking lot this week, photographed by longtime Tesla investor Michael Konen. Tesla is also advertising for an AI Safety Operator in Charlotte to “oversee daily operations of ride-hailing services” and collect data by driving a defined zone six to eight hours a day.

The sighting lands nine days after EVXL counted a record 245 finished Cybercabs staged at Gigafactory Texas with no commercial fleet to absorb them. Test prototypes have left the country before, including the steering-wheel Cybercab now winter testing in New Zealand, but Charlotte is the first documented multi-car batch of that gold inventory staged outside Texas for what looks like service preparation.

Tesla did not respond to Axios Charlotte’s request for comment. The company’s latest earnings report named Las Vegas and Phoenix as the next metros in line. Charlotte was not on the slide.

Charlotte Sets Up a Direct Race With Waymo

Charlotte is shaping up as a head-to-head market between Tesla and Waymo, because Alphabet’s robotaxi unit has been testing and mapping the city for months while Tesla now has vehicles on the ground, a job posting live, and no announced launch date for either service.

The AI Safety Operator listing on Tesla’s careers site describes daily oversight of ride-hailing operations and hours of data-collection driving in a set area. Similar operator postings are live for more than 30 other locations, including Nashville and Atlanta, so Charlotte is one front in a much wider staging effort.

Konen, who shared the photos on X, told Axios Charlotte the city isn’t usually first to new technology, adding, “But I love that technology is coming here.”

The Cybercab Pile Starts to Move

The Charlotte batch is the first sign that the record inventory of finished Cybercabs at Gigafactory Texas has somewhere to go, after EVXL’s July 18 count of 245 staged units documented a production line running far ahead of any deployed commercial service.

Tesla currently runs driverless robotaxis in six metros: Austin, Dallas, Houston, Miami, Orlando, and Tampa. San Francisco cars still carry safety drivers because, as EVXL reported in March, California classifies Tesla’s service as a supervised limo operation, not an autonomous one. In its Q2 report on July 22, Tesla said preparations were underway in Las Vegas and Phoenix for further rollouts.

The safety-operator role matters for reading the timeline. A human driving a mapped zone for hours a day is the data-collection phase Tesla ran in every prior market before opening service. Charlotte is at the start of that playbook, not the end.

EVXL’s Take

I’ve spent months asking where the Giga Texas Cybercab pile would drain to, and Charlotte just gave me the first answer. Fourteen cars in a Carolina parking lot won’t dent a 245-unit backlog, but it proves the inventory has started shipping.

Read the two signals together, though. The earnings deck named Las Vegas and Phoenix. The parking lots named Charlotte. When Tesla’s slides and Tesla’s logistics disagree, I trust the logistics, because this company communicates by deployment, not press release. Thirty-plus operator postings say the real map is far bigger than anything investors were shown on July 22.

The caveat is the job description itself. A safety operator driving six to eight hours a day is mapping, not service, and every market Tesla has launched so far opened on Model Ys while the Cybercab stayed in validation. Waymo has months of Charlotte mapping already banked. Whether Tesla’s arrive-late, launch-fast pattern from Florida repeats here, or whether Waymo converts its head start into the city’s first driverless fare, is the race worth watching. The October earnings call is the next forced disclosure. If Charlotte makes that slide, the lots called it first.

Source: Axios Charlotte

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Sues Cybertruck Supplier Over Tooling Locked Inside A Plant It Is Closing

Haye Kesteloo — Sat, 25 Jul 2026 15:20:00 GMT

Tesla asked a federal judge on Thursday, July 23, 2026, for an emergency order forcing supplier Angstrom Automotive Group LLC to open the doors of its Troy, Texas plant so Tesla can haul out the multi-ton tooling that stamps Cybertruck parts. Angstrom told Tesla on July 13 that it planned to shut the facility down. Ten days later, according to the complaint, Tesla still could not get its equipment out.

The supplier is also sitting on 700 finished Tesla components that were scheduled to ship on July 17 and never left the building. Tesla told the court it will be unable to build “several thousand Cybertrucks that are currently in or planned for production,” most of which already have buyers attached. Replacing the tooling is not a workaround: Tesla says the equipment is extremely complex and would take five to six months to rebuild.

The case is Tesla v. Angstrom Automotive Group LLC, No. 6:26-cv-0477, filed in the U.S. District Court for the Western District of Texas. Tesla is represented by Norton Rose Fulbright US LLP. Angstrom declined to comment when contacted on July 24.

The Standoff Escalated Over Nine Days In July

Angstrom notified Tesla on July 13 that it intended to close the Troy plant, then refused to work out a retrieval plan for the tooling, and by July 21 Tesla representatives had shown up at the gate accompanied by law enforcement and been turned away, according to the account in Bloomberg Law’s report on the filing.

Between those two dates, the July 17 shipment of 700 parts failed to move. Tesla then received a compensation demand covering modifications the supplier had made to the tooling, plus an offer to keep running the plant if Tesla paid $250,000 on top of whatever it already owed on open purchase orders.

A note on that figure, because it matters and the record is not clean. Bloomberg Law’s summary line describes the payment as $250,000 per day. Its account of the July 20 demand describes $250,000 per week. Those are wildly different numbers, roughly $91 million against $13 million annualized, and EVXL is not going to pick the scarier one to make a better headline. The complaint itself will settle it.

The relationship is barely a year old. Bloomberg Law reports the partnership began when Angstrom acquired a company already supplying Tesla in early 2025. Angstrom is a tier-one supplier founded in 1999 and headquartered in Southfield, Michigan, and it spent 2025 buying up rivals, including Canadian supplier KSR International in a deal Automotive News reported created a business with more than $1 billion in annual revenue and 4,500 employees.

Finished Cybertrucks fill staging lots at Gigafactory Texas. Tesla says tooling held at a supplier's plant north of Austin feeds parts to this line. File photo credit: Joe Tegtmeyer.

Tesla Wants The Tools Back And Nothing Else Decided

The filing asks for one narrow thing, retrieval of the tooling, and explicitly declines to litigate any other dispute between the two companies, which strips the money fight out of the emergency posture and leaves the judge a single question about physical access to property.

That framing is a tell. Tesla runs just-in-time manufacturing, so a supplier holding a single set of dies can idle a line in days rather than months. The company is not asking a court to decide who owes whom. It is asking a court to get the presses onto trucks before the parts bins at Gigafactory Texas run dry.

The five-to-six-month rebuild estimate is the number that gives this dispute its weight. It means Tesla has no dual-sourced fallback for these parts and no ability to buy its way out on a useful timescale. Whoever physically holds that tooling holds the Cybertruck line, and Angstrom appears to have understood that before Tesla did.

The Cybertruck Has No Cushion Left To Absorb A Stoppage

The truck enters this fight already diminished, with U.S. sales down 48.1% to 20,237 units in 2025 and a Texas factory that has visibly reorganized itself around a different vehicle, which means a production interruption now lands on a program with no volume buffer and little internal priority.

EVXL laid out the collapse in the 2025 full-year sales analysis published January 13: 38,965 units in 2024, then a near-halving, with Q4 alone down 68.1% year over year. Meanwhile the drone flyover of Giga Texas on July 18 counted 245 finished Cybercabs staged across the lots, more than double the tally from five weeks earlier. Austin’s attention has moved.

The balance sheet offers no slack either. Tesla’s Q2 results, reported on July 22, showed operating margin compressed to 1.4% and free cash flow negative for the first time in more than two years. An idled line is an expense this income statement is in no shape to absorb.

EVXL’s Take

Read the calendar, not the adjectives. A supplier announces it is closing a plant, then declines to hand back a customer’s tooling, then names a price to keep the lights on. That is not a shakedown by a healthy business. That is a company trying to get paid for a facility it built around volumes that never arrived, and Tesla’s $250,000 demand and Angstrom’s plant closure are two symptoms of the same disease. Musk projected 250,000 to 500,000 Cybertrucks a year. Austin installed capacity for 125,000. America bought 20,237 in 2025. Somebody had to eat that gap, and tier-one suppliers are usually the ones holding the plate.

Which does not make Angstrom right. If Tesla’s ownership documents are as clean as the complaint implies, self-help has no place here, and courts exist precisely so that commercial disputes get resolved on paper instead of at a locked gate with police present. Tesla should get its equipment. It should also recognize that a supplier willing to burn the relationship this publicly usually believes it has nothing left to lose.

Here is the part Tesla will not enjoy. Its own filing claims several thousand Cybertrucks in or planned for production with buyers already attached, which is a considerably healthier order book than 20,237 units a year suggests. Both statements cannot be casually true. One of them is going to get tested in a public docket, under oath, by a supplier with every incentive to produce the purchase orders and volume forecasts Tesla actually sent it. That document, whenever Angstrom files it, will tell EV buyers more about real Cybertruck demand than any quarterly delivery release has.

Watch for Angstrom’s response and the ruling on the emergency motion, both of which should land within days rather than weeks given the posture. The tooling will almost certainly come home. The numbers attached to it are the story.

Source: Bloomberg Law, reporting by Ryan Autullo.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's Founders Bankroll The Anti-Cybertruck, A $41,520 Mini Truck Running Their Original Playbook

Haye Kesteloo — Fri, 24 Jul 2026 21:05:00 GMT

Tesla’s Founders Bankroll The Anti-Cybertruck, A $41,520 Mini Truck Running Their Original Playbook

Tesla founders Marc Tarpenning and Martin Eberhard, the two engineers who started the company and recruited Elon Musk as its first big investor, are now backing TELO, a California startup building a mini electric pickup its own team frames as the opposite of the Cybertruck. Forbes reported the investment on July 23, twenty years after the original Tesla Roadster introduced the company to the world.

TELO has raised about $27 million to date, including a $20 million round in September 2025 that drew Tarpenning’s Spero Ventures and Eberhard as investors. Its truck, the MT1, starts at $41,520, measures about the length of a two-door MINI Cooper, and is targeted to reach first customers by the end of this year or early 2027.

EVXL has tracked the American small-truck counterattack since Slate’s April 2025 reveal, and the pattern keeps repeating: the most interesting U.S. EV startups right now are building smaller, not bigger. This one comes with a twist, because the people funding it wrote Tesla’s original playbook.

Tarpenning And Eberhard Bet On The Truck Tesla Won’t Build

Marc Tarpenning and Martin Eberhard, the two engineers who founded Tesla in 2003 and convinced Elon Musk to write the first check, are now investors in a San Carlos startup whose product rejects nearly everything the Cybertruck stands for: small footprint, soft styling, low volume, and quiet ambition. Tarpenning sits on TELO’s board through Spero Ventures, where he has invested in cleantech since leaving Tesla in 2008. Eberhard, forced out of Tesla in late 2007, joined the September round and also advises Swiss minicar maker Microlino, according to Forbes.

The symbolism runs deep. TELO operates out of the same Silicon Valley office park where Tesla was headquartered when the Roadster launched in 2006. Cofounder Forrest North is a former Tesla engineer. And CEO Jason Marks openly cites the original Tesla ramp, 2,400 Roadsters before roughly 10,000 Model S sedans, as the template: TELO plans just 500 MT1s at first, then a 5,000-per-year run rate, and Marks claims the company can turn a profit even at that volume.

Neither founder hides what they think of Tesla’s pickup. “I personally think Tesla’s lost its way a bit, especially with the Cybertruck,” Tarpenning told Forbes, adding that no professional designer would have drawn it without Musk’s fingerprints on the file.

The MT1 Starts At $41,520 And Ships Without A Factory

TELO’s MT1 is a pickup roughly the length of a two-door MINI Cooper, priced from $41,520 with 260 miles of range, and the company plans to build its first bodies through Michigan contract manufacturer Schwab Industries rather than a factory of its own. A $45,500 long-range version stretches to 350 miles, and adding a second motor pushes the price to $50,000. TELO installs its own battery pack and components in San Carlos, which keeps its capital needs a fraction of what Rivian burned reaching production.

The engineering bet is packaging. The MT1’s flat battery pack with integrated motors is designed to stiffen the frame, and Marks says the team started from the minimum frontal crash structure needed for a five-star rating, then packed a working truck behind it. That claim faces its first real test when crash testing begins later this year. Computer modeling is not a crash cell, and a truck this short has no margin for error.

Pricing is the other open question. The MT1 undercuts the roughly $55,000 average price of a new EV, but $41,520 is not cheap for a vehicle this small, and the federal purchase credit that would have softened it is gone. Tarpenning concedes the price will need to come down for mass-market reach. The startup claims roughly 12,000 reservations, per TechCrunch.

The Founders’ Verdict On Today’s Tesla Lands Harder Than The Money

Both founders told Forbes they remain proud of what Tesla achieved over two decades, and both watched the company hand its global EV lead to Chinese rivals while Musk redirected the business toward robotaxis, humanoid robots, and AI instead of vehicles people can actually buy. Their money is now a statement about that drift.

The contrast with Tesla’s own truck is brutal. EVXL documented 10,000 unsold Cybertrucks worth roughly $800 million sitting on U.S. lots in May 2025, against a product once projected at 250,000 units a year. Meanwhile Musk’s attention sits elsewhere: EVXL reported in June that Tesla asked Nevada for a 5,000-vehicle robotaxi permit while running roughly 20 driverless cars nationwide.

TELO is not alone in the segment it is entering. Slate, backed by Jeff Bezos with more than $700 million raised, is chasing the stripped-down end of the same small-pickup market. Eberhard sees room for both, arguing TELO’s value is that it competes in a lane nobody else occupies rather than fighting Tesla or Chinese makers head-on.

EVXL’s Take

The two people who invented Tesla’s playbook just endorsed a competitor that runs it better than Tesla currently does, and that says more about where the company under Musk has drifted than any single delivery chart or earnings call ever could. Start small. Prove the product. Delight the buyer. Scale when the math works. That was the 2006 plan, and in 2026 it describes TELO more than it describes Tesla, which is asking regulators for 5,000 robotaxis it cannot field while its halo truck rusts on inventory lots.

I’m not anointing TELO. Twenty-seven million dollars is lunch money in car manufacturing, the crash test hasn’t happened, and $41,520 without a federal credit is a real ask for a truck the size of a hatchback. Slate’s 100,000 reservations show the demand signal for small and cheap; TELO still has to prove the demand for small and $42,000.

Two checkpoints will settle it, and both are on the calendar. Watch the crash testing later this year: if the MT1 pulls a five-star result at a 152-inch footprint, the industry’s “big trucks are safer” excuse dies in public. Then watch whether the first 500 customer trucks actually ship by early 2027 as promised. Startups that hit their first delivery window earn the right to be taken seriously. The founders’ judgment of the Cybertruck is already on the record. Now TELO has to earn it.

Source: Forbes, TechCrunch

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's Musk Premium Just Got Priced For The First Time, And The Number Is Ugly

Haye Kesteloo — Fri, 24 Jul 2026 20:34:00 GMT

Tesla’s Musk Premium Just Got Priced For The First Time, And The Number Is Ugly

Tesla shares fell 14% after Wednesday’s earnings release, and Barron’s argues the drop was never really about the $398 million operating profit that missed Wall Street’s $1.7 billion bar. The magazine’s diagnosis cuts deeper: investors are starting to doubt Elon Musk’s ability to sell the future, and that ability is the asset the entire stock price rests on.

The numbers behind the argument are stark. Tesla missed profit expectations by $1.3 billion in a quarter where it delivered a record 480,126 vehicles, roughly 80,000 more than analysts projected. Selling more cars than anyone expected and earning less than half of last year’s profit on them is the pattern I flagged in EVXL’s Q2 earnings breakdown yesterday: the record was bought with discounts, and the bill landed on the operating line.

Barron’s puts a face on the market’s mood through Future Fund co-founder Gary Black: “Investors are losing patience in hype followed by a lack of follow-through.”

The Selloff Ran Seven Times Deeper Than The Analyst Cuts

Tesla shares fell 14% even though the average analyst price target slipped just $8 to $392, a decline Barron’s calculates at about $30 billion in market value, meaning investors sold roughly seven times harder than the professionals who cover the company adjusted their models. No analyst upgraded or downgraded the stock in the immediate aftermath.

That gap between the professional reaction and the market reaction is the story. Analysts trimmed around the edges while shareholders repriced the man. Canaccord’s George Gianarikas kept his Buy rating while cutting his target by $40 to $410, per Barron’s reporting syndicated by moomoo, and framed betting against Musk as historically unwise. He may be right. The sellers on Thursday clearly wanted more than history.

Musk gave them the familiar menu instead. The earnings call featured robotaxi growth talk, robots as the biggest product ever, and a brand-new business line called the Megapod, a modular AI data center built on Tesla hardware. A subdued Musk pitching three future businesses while the current one posts a 1.4% operating margin is exactly the tension Barron’s captures.

Wall Street Keeps Its Buy Ratings While The Multiple Does The Talking

Tesla trades at roughly 175 times expected 2026 earnings while the rest of the Magnificent Seven average about 24 times, per Barron’s, a multiple that leaves the valuation resting almost entirely on faith in future businesses that produce no material revenue on today’s income statement. Value the company like Toyota Motor and Barron’s puts the shares near $20.

None of this is news to EVXL readers. When I documented Tesla’s $44 billion free cash flow collapse in April, the forward multiple stood at 178x and the disconnect was already the story. What changed this week is that the doubt moved from the spreadsheets into the tape.

Musk’s track record supplies both sides of the argument. Tesla has sold nearly 10 million EVs, and Full Self-Driving (Supervised) subscriptions hit 1.48 million this quarter, up 56% in a year. He has also promised 20 million cars annually by 2030, a 2023 target Barron’s calls well out of reach, and robotaxis without safety monitors on timelines that keep sliding.

EVXL’s Take

Barron’s says Musk needs to show results to restore the premium. I’d flip that. The premium was never a reward for results. It was a loan against promises, extended at 175 times earnings, and this week the market made its first margin call.

I’ve covered the pattern too long to treat this crash as noise. The 20 million cars by 2030. Unsupervised robotaxis in three weeks, promised last December, while EVXL’s fleet tracking in June still counted roughly 20 driverless cars across three cities. Now the Megapod arrives on cue, a shiny new future business unveiled in the exact quarter the current one earned 1.4 cents per revenue dollar. That timing isn’t coincidence. It’s the playbook, and as I wrote yesterday about the pay package built on these same milestones, the company is being reshaped around a scorecard of promises rather than a car business that funds them.

Here’s the part the doom headlines will miss: the FSD subscription curve is real, growing, and audited. If Musk wants his magic back, that table is where it lives, not in another keynote product. Watch the Q3 report in October. If the market sells a delivery beat again, the premium isn’t compressing. It’s gone.

Sources: Barron’s, Tesla Q2 2026 Update (SEC filing).

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Cybercab Hits Public Roads in New Zealand, Caked in Winter Testing Grime

Haye Kesteloo — Fri, 24 Jul 2026 17:00:00 GMT

A Tesla Cybercab appeared driving on a public road in New Zealand in footage that surfaced Friday, July 24, 2026, what appears to be the first video of the two-seat robotaxi moving under its own power outside the United States. The short clip, shot by Stephen Hillier, shows the gold prototype turning through a rural two-lane intersection, its body coated in road grime, a fitted tan cover over the roof, and a yellow New Zealand number plate on the tail.

The sighting closes the loop on a story EVXL readers already know the first half of. In mid-June, at least two Cybercabs were spotted on a car transporter on the South Island, and Tesla later confirmed the vehicles were headed to the Southern Hemisphere Proving Grounds near Cardrona for cold-weather testing. Five weeks on, a gold prototype is out of the compound and working New Zealand’s public roads.

Cybercab spotted winter testing in New Zealand.
— Sawyer Merritt (@SawyerMerritt) July 24, 2026

The Video Shows a Registered Test Car, Not a Trailer Queen

The short clip shows a Cybercab doing something we had not seen one do outside the US until now: rolling down an ordinary public road, plated, dirty, and clearly deep into a test program rather than posing for one. That dirt is the detail worth pausing on. Prototypes get filthy when they run daily loops between a proving ground and open roads, and this car looks like it has been doing exactly that since the transporter sighting in June.

The New Zealand plate matters too. Photos shared by local Tesla owner groups in June showed the arriving Cybercabs were left-hand-drive cars fitted with steering wheels, even though the production robotaxi was unveiled without a wheel or pedals. New Zealand drives on the left, so a left-hand-drive prototype with manual controls and local registration points to one purpose: a human test driver logging validation miles on real winter roads, not a regional launch.

Tesla builds both configurations. As EVXL reported when the EPA certification filing surfaced in June, the certified car is a front-wheel-drive single-motor design rated at 219 hp, and the company has kept a steering-wheel variant in the mix for markets and test programs that require one.

New Zealand Is Tesla’s Standing Winter Laboratory

The Southern Hemisphere Proving Grounds, sitting between Queenstown and Wanaka near the Cardrona ski field, gives automakers sub-zero test conditions from June through August while facilities in Scandinavia and Canada sit in summer thaw. The facility’s winter season typically runs June through late August, and Tesla is a repeat customer.

EVXL has covered this pattern before. In July 2023, Model 3 Highland prototypes and camouflaged Cybertrucks turned up at the same facility for the same seasonal reason. Three years later the program is running the same playbook with a far stranger vehicle.

For the Cybercab, the winter checklist is longer than usual. A camera-only autonomy stack has to prove it can read frost-glazed roads, low winter sun, and fogged lenses. Battery preconditioning, cabin heating, and traction behavior all need cold-weather data before Tesla can put the pedal-free version to work anywhere with an actual winter. The 245 finished units EVXL counted staged at Giga Texas on July 18 are waiting on exactly this kind of validation work.

EVXL’s Take

This is the most encouraging Cybercab footage I’ve seen all year, and it’s encouraging precisely because it’s boring. A dirty prototype with a legal plate grinding out miles on a rural road is what a real engineering program looks like. It’s the unglamorous opposite of the choreographed 2024 unveiling, and it’s worth far more.

But let’s be clear about what this is not. A left-hand-drive car with a steering wheel and a human aboard, in a country that drives on the left, is not a New Zealand robotaxi launch. It’s Tesla borrowing winter. The company that promised full autonomy “next year” every year for a decade is now doing the slow, dusty validation work it should have front-loaded years ago, and I’d rather see this than another demo night.

The SHPG season runs through the end of August. Expect more sightings like this one on Central Otago roads over the next five weeks, and judge the program by how many show up. A test car you keep seeing is a test program that’s actually running.

Sources: CarExpert, MotoringNZ. Video by Stephen Hillier.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's Pay To Musk Now Exceeds Everything It Ever Earned

Haye Kesteloo — Thu, 23 Jul 2026 14:14:00 GMT

Tesla’s Pay To Musk Now Exceeds Everything It Ever Earned

Tesla closed its second quarter with $398 million in operating income. One of the milestones in Elon Musk’s new compensation package requires $400 billion in annual adjusted profit. That is not a typo, and the gap between those two numbers is the honest frame for the question every Tesla owner and shareholder should be asking after Wednesday’s earnings: what exactly is this company paying for?

The pay stack is now the largest in corporate history by a wide margin. In December, the Delaware Supreme Court reinstated Musk’s 2018 award of 303 million split-adjusted stock options, valued at roughly $139 billion at the time of the ruling, according to CNN. A month earlier, shareholders had approved a second package worth up to $878 billion if Musk hits its targets over the next decade, a vote EVXL covered as it happened.

For scale: Reuters calculated last fall that a $26 billion award alone would exceed half of everything Tesla has earned since it became profitable in 2019. The 2018 package was worth more than five times that figure at the December valuation. The pay already outweighs the profits, cumulatively, from the company’s entire existence.

The Two Packages Stack On Top Of Each Other

The December court ruling and the November shareholder vote were separate events with separate awards, and Musk keeps both, which means the reinstated 2018 options and the new decade-long package now run concurrently rather than one replacing the other. The court called rescinding the 2018 deal an extreme remedy and slashed the plaintiff attorneys’ $345 million fee, per Gibson Dunn’s analysis of the decision.

The new package grants up to 423.7 million additional shares across tranches tied to milestones: an $8.5 trillion market capitalization, $400 billion in annual adjusted profit, 20 million cumulative vehicle deliveries, 10 million active FSD subscriptions, one million Optimus robots, and one million robotaxis in operation. Glass Lewis estimated the entire S&P 500 paid all of its CEOs about $9 billion combined last year. Even conservative valuations put Musk’s new package near $8.8 billion per year on its own.

Musk has framed the award as a control question, saying he won’t build what he calls a “robot army” without stronger voting influence. Full payout would lift his stake to roughly 27% of the company.

Q2’s Results Sit A Very Long Way From The Targets

Wednesday’s earnings, covered in detail in EVXL’s Q2 report, put hard current numbers against every milestone in the new package for the first time since its approval, and the distance between where Tesla stands and where the payouts trigger is enormous on every line. Operating income of $398 million annualizes to about $1.6 billion against a $400 billion target. Cumulative deliveries stand at 9.7 million against a 20 million bar.

FSD subscriptions reached 1.48 million against the 10 million milestone. That number is growing fast, up 56% in a year, but part of the growth is manufactured: Tesla killed the one-time FSD purchase option in February and stripped free Autopilot from new cars, funneling buyers toward the $99 monthly subscription, as EVXL reported in January. When a CEO’s personal payout depends on a subscription counter, watch what happens to the free features.

The robotaxi and robot milestones are further out still. Tesla runs 175 registered robotaxis in Texas against a one-million-vehicle target. Optimus production lines are under construction, with initial units earmarked for training data collection, not customers. Tesla’s market cap sits around $1.4 trillion against the $8.5 trillion ceiling tranche.

The SpaceX Entanglement Ran Straight Through Q2’s Profit Line

Tesla disclosed in its first-quarter 10-Q that it invested $2.0 billion of shareholder cash in SpaceX common stock, and that stake came straight back around in Wednesday’s numbers, where it helped fill the wide gap between operating income and the net income line. Operating income was $398 million. Net income was $1.11 billion. The difference came from below the operating line, from interest income and other income that included a mark-to-market gain on the SpaceX holding.

That gain was struck at the June 30 quarter close. SpaceX, which merged with xAI and held a record market debut in June, has since lost more than 40% of its value from its peak close, per CNBC. If SPCX stays below its June 30 mark, the same accounting that flattered this quarter’s net income will subtract from the next one. Layer on Terafab, the chip plant Tesla plans to build with SpaceX and Intel, and the merger speculation both companies have fed, and a pattern emerges: Tesla capital increasingly flows toward, and Tesla profit increasingly depends on, the CEO’s other company.

EVXL’s Take

The strongest defense of the package is real and I’ll state it plainly: Musk gets nothing unless shareholders get an $8.5 trillion company, and the 2018 award only paid because he delivered a twelvefold value increase nobody thought possible. Pay for performance, at absurd scale, is still pay for performance.

Here’s why that defense doesn’t settle it. The question was never whether Musk gets rich when shareholders do. The question is what the incentives steer him to do along the way, and the steering is already visible. The 10 million FSD subscription milestone arrived in November. By February, the purchase option was dead and free Autopilot was gone from new cars. The robot milestones arrived, and Model S and Model X production ended to clear Fremont floor space for Optimus lines. The company is being reshaped around a scorecard, and the scorecard belongs to one man.

Meanwhile the car business, the thing generating the $28.2 billion in quarterly revenue that funds all of it, just posted a 1.4% operating margin, and the net income line needed a paper gain on the CEO’s other company to hold nearly flat. Tesla shareholders have now committed potential compensation worth many multiples of every dollar of profit in the company’s history, to be paid in dilution, for targets that today’s numbers approach at a crawl.

Is he worth it? My answer: nobody is, at this structure, because the structure itself is the problem. A package that pays out on subscription counts and robot counts invites exactly the owner-hostile monetization we’ve documented since January. Watch two things from here. First, whether Q3 brings a SpaceX markdown that shows how much of Tesla’s reported profit was ever Tesla’s. Second, whether the proxy disclosures ahead of the next annual meeting show any tranche progress at all. The package runs ten years. The patience of people buying the cars may not.

Sources: Tesla Q1 2026 Form 10-Q (SEC), CNN, Gibson Dunn, CNBC, Reuters.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Q2 Earnings Reveal What The Delivery Record Cost

Haye Kesteloo — Thu, 23 Jul 2026 12:40:59 GMT

Tesla Q2 Earnings Reveal What The Delivery Record Cost

Tesla reported $28.2 billion in second-quarter revenue on Wednesday, up 26% from a year ago, alongside its first negative free cash flow in more than two years: a $1.1 billion outflow driven by $5.8 billion in capital spending on AI infrastructure and robotics. Shares fell nearly 3% in after-market trading, according to The Wall Street Journal.

The record 480,126 deliveries that powered the revenue growth were already public. What Wednesday’s report added is the bill. Operating margin collapsed to 1.4%, down from 4.1% a year ago and 4.2% last quarter. GAAP net income slipped 5% to $1.11 billion, and non-GAAP earnings of $0.33 per share missed the $0.55 consensus Tesla itself had compiled from analysts.

Yesterday, hours before this report, I wrote that automotive gross margin was the number to watch, not the delivery record everyone already knew, in EVXL’s look at Tesla extending 3.99% APR to the Model Y Performance. Total gross margin came in at 16.8%, the lowest in five quarters.

Revenue Grew 26% While Profit Went The Other Way

Every revenue line in Tesla’s Q2 update grew by double digits, yet income from operations fell 57% to $398 million because operating expenses jumped 47% to $4.35 billion and the gross margin that funds them compressed by more than four percentage points from the first quarter. That is the whole quarter in one sentence.

Automotive revenue rose 23% to $20.5 billion on the delivery record. Energy generation and storage added $3.1 billion, up 13%, on 13.5 GWh deployed. Services and other revenue grew fastest of all, up 50% to $4.6 billion.

The margin side tells the opposite story. Gross margin of 16.8% fell from 21.1% in Q1. A caveat on causes: Tesla’s summary table doesn’t break out automotive gross margin, so segment mix played some part, and the 10-Q will show how much. What is documented is the incentive stack EVXL has tracked all quarter, from 0% APR on the base Model Y to the 3.99% rate Tesla extended to the Performance trim the day before earnings. Cox Automotive estimates Tesla’s US deliveries fell 20.1% year over year even as global volume set a record, and the Journal cites Motor Intelligence data showing a similar 20% US decline. The growth came from Europe and China. The discounts hit margin everywhere.

One genuine bright spot: operating cash flow rose 85% to $4.7 billion. The core business generates cash. The investment program simply consumes more of it than the business produces.

Capital Spending Nearly Tripled To Fund The AI Pivot

Tesla spent $5.8 billion on capital expenditures in the quarter, a 142% increase over a year ago, and that single line converted $4.7 billion of operating cash flow into a $1.1 billion free cash flow deficit, the company’s first since early 2024. Management had warned on the Q1 call that free cash flow would stay negative for the rest of 2026.

Among the largest commitments is Terafab, the chip-manufacturing plant Tesla plans to build and run with SpaceX and Intel in Texas, per CNBC’s earnings coverage. The Journal reports the project has fueled investor speculation that Musk could try to merge SpaceX and Tesla into a single AI-focused company.

The spending also covers the Cybercab ramp. Tesla says volume production of the pedal-free two-seater is underway, which matches what drone flyovers have documented on the ground: EVXL counted 245 finished Cybercabs staged at Giga Texas on July 18, none of them yet carrying a paying passenger. On Optimus, Tesla told investors it is installing first-generation production lines and will start building robots soon, though the initial units are earmarked for training data collection rather than customers. The capacity table lists Cybercab at more than 125,000 units of installed annual capacity, with Optimus lines in California and Texas still under construction.

FSD Subscriptions Are The Growth Engine Tesla Actually Landed

Active subscriptions to Full Self-Driving (Supervised) reached 1.48 million in the quarter, up 56% from a year ago and roughly 200,000 ahead of Q1, making the software Tesla sells to US drivers for $99 a month the clearest driver of the 50% jump in services revenue.

The Robotaxi side moved too. Per the Journal, Tesla expanded the service this month to limited areas of Miami, Orlando, and Tampa, on top of its Texas operations in Austin, Dallas, and Houston. In June, the National Highway Traffic Safety Administration proposed eliminating the requirement that autonomous vehicles carry manual brake pedals, a change that would directly benefit the controls-free Cybercab.

EVXL’s Take

The story most outlets will run tonight is the AI spending headline, and it is the wrong lens. Tesla told everyone in April that free cash flow would go negative. The $1.1 billion burn actually beat the $3.25 billion outflow Wall Street expected. The real news is quieter and worse: the car company inside the AI company earned 1.4 cents of operating profit per revenue dollar in its best delivery quarter ever.

That is what buying a record looks like on an income statement. I wrote on July 1 that a consensus beat built on incentives would be Tesla grading its own homework, and the grade is now in. Deliveries up 25%, operating income down 57%. The bull will point out that net income held within 5% of last year on 26% more revenue, and that’s a fair card to play. It just doesn’t answer the question this quarter raised. Volume was never the question. Pricing power was.

The FSD number is the honest counterweight, and I’ll give Tesla full credit for it. Adding 200,000 subscribers in one quarter is the recurring-revenue business Musk has promised for years finally showing up in a table instead of a keynote. If that curve holds, it changes the margin math more than any robot does.

Here is the test I’ll hold Tesla to. The incentive rates that bought this record, 0% to 3.99% APR across the Model Y range, run through the full third quarter. If automotive gross margin stabilizes in Q3 while those rates stay live, Tesla has the pricing power its valuation assumes. If margin keeps sliding, this quarter was not an investment phase. It was a margin war Tesla started against itself, funded by a cash pile that shrank $1.2 billion this quarter to $43.5 billion. Watch the Q3 report in October, not the robot demos between now and then.

Sources: Tesla Q2 2026 Update, CNBC, The Wall Street Journal.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Extends 3.99% APR To The Model Y Performance, The Last Trim It Was Holding At Full Price

Haye Kesteloo — Tue, 21 Jul 2026 19:51:11 GMT

Tesla is now offering 3.99% APR financing on the new Model Y Performance for loan terms up to 72 months, according to a marketing email the company sent to customers on Tuesday afternoon, July 21, 2026. The one that landed in my inbox at 3:13 PM pitches the trim’s “instant torque” alongside a nudge to subscribe to Full Self-Driving (Supervised).

The rate applies to well-qualified buyers only, and Tesla’s fine print notes the offer is subject to change or end at any time. The company did not attach an end date in the email itself.

The promotion matters because of which trim it covers. The Performance has been the lone Model Y variant excluded from Tesla’s rate war on itself, the holdout Tesla kept at standard financing while it bought down every cheaper version of the car. That wall just came down.

The Performance Was The Only Model Y Locked Out Of Tesla’s Rate Cuts

Until this week, the Model Y Performance was the single trim in the lineup without a subsidized interest rate, while the rest of the range enjoyed rates between 0% and 0.99% for the same 72-month terms. Third-party trackers of Tesla’s current offers listed no Performance financing incentive as recently as July 8.

When Tesla rolled out 0% financing on the Model Y Rear-Wheel Drive and 0.99% on the All-Wheel Drive and Premium trims in February, the Performance was explicitly left out. Earlier this year, buyers of the top trim faced Tesla’s standard rate of 5.34%, per Cars.com data from the spring.

At 3.99%, the Performance still pays a premium over its cheaper siblings. A buyer financing the 460-horsepower flagship pays roughly four percentage points more in interest than someone financing the base Rear-Wheel Drive car at 0%. Tesla is discounting its halo trim, but not all the way.

I’ve tracked this pattern since Tesla first cut Model Y financing to 1.99% in May 2025, when the freshly redesigned Juniper was barely a month old and already needed help. The tool has stayed the same. Only the trims have changed.

2026 Model Y Performance. Photo: Tesla

Cheap Money Is Doing The Job The Tax Credit Used To Do

Tesla’s financing promotions have grown steadily more aggressive since the $7,500 federal EV tax credit expired on September 30, 2025, and the company has been absorbing the cost of that lost subsidy through rate buy-downs, lease credits, and eliminated down payments ever since.

EVXL documented that shift in November, when Tesla dropped the down payment on Model Y leases entirely and layered a $6,500 internal lease credit on top. Every one of those dollars now comes out of Tesla’s margin instead of the Treasury’s.

The US numbers explain the urgency. Tesla posted its best second quarter ever with 480,126 global deliveries, up 25% year over year, but Cox Automotive estimates its US deliveries fell 20.1% to 114,629 in the same quarter. The growth came from overseas. At home, where this 3.99% offer applies, demand is still digesting the loss of the credit.

Sub-1% financing on a car loan is not free for the lender. Tesla eats the spread between its promotional rates and market rates on every contract, which is one reason analysts expect revenue near $25.7 billion for the quarter, down about 3% from a year ago despite the delivery record.

EVXL’s Take

Read the timing. This email went out on July 21. Tesla reports Q2 earnings on July 22. The company spent the quarter buying a delivery record with incentives, and the day before it has to explain what that record cost, it extends subsidized financing to the one trim it had protected.

The Performance exists to be the trim people pay full freight for. It’s the 3.3-second halo car, the version that carries the fattest margin in the lineup. When the halo needs a rate buy-down, the message isn’t “treat yourself.” The message is that US demand at the top of the Model Y range is soft enough that Tesla would rather eat interest than watch Performance inventory sit.

There’s a second tell in the email itself: Tesla invites buyers to “let your Model Y drive for you under your supervision.” The company is bundling its highest-margin software pitch into its financing pitch, because recurring FSD revenue is the only thing that makes 3.99% money on a performance SUV pencil out.

I flagged before the delivery report that a consensus beat built on incentives would be grading its own homework, and the beat arrived exactly that way. Tomorrow’s earnings call will show the bill. Watch automotive gross margin, not the delivery number everyone already knows. If margin holds while rates like this one spread across the lineup, Tesla has pricing power its rivals should fear. If it doesn’t, this email is what a margin war looks like when it reaches the last trim standing.

For buyers, none of this is bad news. A 3.99% rate on the quickest Model Y beats the 5.34% it replaced, and Tesla’s history says these offers vanish without warning. Just know why the deal exists. It’s not generosity. It’s arithmetic.

Source: Tesla.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's Giga Berlin Ran At 54% Capacity In 2025 And Still Grew Profit 36%, Annual Report Shows

Haye Kesteloo — Mon, 20 Jul 2026 10:08:07 GMT

Tesla’s German factory operated at 54% of capacity in 2025, built 202,000 vehicles against 211,000 the year before, and closed the year with a net profit of EUR 77.1 million, up 36% from 2024. All three numbers come from the same document: the annual report of Tesla Manufacturing Brandenburg SE, the legal entity behind Giga Berlin, filed with Germany’s Bundesanzeiger and signed in Grünheide on May 7, 2026.

The filing covers the fiscal year from January 1 to December 31, 2025, and it reads very differently from the celebratory excerpts circulating on social media this weekend. Fans are sharing the 700,000-vehicle milestone and the battery investment figures. The report itself documents a plant that spent 2025 running at just over half its potential, a fact that gives real weight to the expansion sprint now underway. EVXL has tracked that tension since February, when expansion at the site was still being dangled as a bargaining chip.

Contract Manufacturing Math Turns Falling Output Into Higher Profit

Giga Berlin earned more money in 2025 while building fewer cars and billing less revenue, because the entity is a contract manufacturer whose income is structurally insulated from Tesla’s actual sales performance in Europe, with revenue falling from EUR 7,675.0 million to EUR 7,123.3 million as production costs dropped.

Tesla Manufacturing Brandenburg SE builds the Model Y exclusively for Tesla Motors Netherlands B.V. under a contract manufacturing agreement, which means its revenue directly tracks its production costs plus a margin. Net profit rose from EUR 56.8 million in 2024 to EUR 77.1 million in 2025. The German entity gets paid to build cars, not to sell them, so a brutal European sales year barely dents its bottom line. The filing pins the lower output on the Model Y changeover at the start of 2025 and the ramp of new variants, including the Performance version and a Canadian export build.

One line deserves more attention than it will get: warranty provisions jumped from EUR 563.7 million to EUR 802.2 million, a 42% increase the company attributes to the growing cumulative number of vehicles it has produced. The site employed 11,083 people including temporary workers as of December 31, 2025, up from 11,003 a year earlier, and supplied more than 30 markets.

Battery Cell Production Slips To 2027 As The Bill Nears EUR 1 Billion

The report confirms in writing what years of announcements kept vague: Giga Berlin’s battery cell factory is not expected to reach full production until 2027, five years after the plant opened, even as the cumulative investment in cell manufacturing approaches EUR 1 billion.

During fiscal 2025, Tesla decided to invest further in cell production capacity to reach up to 8 GWh per year, with additional spending in the three-digit millions bringing the total cell investment to nearly EUR 1 billion. Current work at the cell facility focuses on component manufacturing and preparation, with full production expected in 2027. That target has since grown: plant chief Andre Thierig said in late June that new investments will enable “18 GWh of 4680 manufacturing capacity starting from 2027”.

Overall capital spending tells a quieter story. Investments in property, plant, and equipment fell to EUR 320.6 million in 2025 from EUR 500.4 million the year before. The report frames bundling the full value chain, from battery cell to finished vehicle, as unique in Europe, while conceding that economical cell production on the continent faces major cost disadvantages against the USA and China.

The 2026 Forecast Leans On A European Recovery That Showed Up

For 2026, the filing forecasts a sharply higher production volume and correspondingly higher capacity utilization, and unlike many corporate forecasts, this one already has three quarters of supporting evidence: record output, record deliveries, and two announced production increases stacked on top of each other.

Giga Berlin crossed 700,000 cumulative Model Y units by March 31, 2026, per the report, and set a quarterly record with 61,000 vehicles in the first quarter. In April, Tesla announced a roughly 20% output increase to about 6,200 vehicles per week from July. On June 25, Thierig announced a second step to 7,500 vehicles per week starting in October, each phase adding around 1,000 jobs. Tesla then delivered a record 480,126 vehicles globally in the second quarter, a recovery EVXL previewed the day before those numbers landed, driven largely by the European demand rebound.

The report also lists regulatory tailwinds among its opportunities. It cites the April 2026 approval of FSD Supervised in the Netherlands, which EVXL covered the day the RDW issued it, and the EU Commission’s December 2025 proposal to soften the 2035 CO2 fleet target from a 100% to a 90% reduction. Tesla’s own read on that softening is telling: the company argues legacy rivals will now burn capital running parallel drivetrain programs while Tesla concentrates on a single electric platform.

EVXL’s Take

The number in this report that matters most is 54%. Not the 700,000 milestone, not the billion-euro cell investment. A factory built for 375,000 vehicles a year produced 202,000 in 2025, and its own management logged utilization slipping from 56% to 54%. Every celebratory repost of this filing skips that line.

Remember what happened in February. Elon Musk told Giga Berlin’s workers that Tesla wouldn’t expand the plant if IG Metall gained influence, and I wrote then that the threat was hollow because a factory running this far below capacity had no expansion needs regardless of who won the works council vote. This filing proves the half-empty part in audited print. IG Metall then lost the March election, its share falling to 31.1%, and Tesla announced its first production ramp weeks later, so Musk’s supporters will read that sequence as a promise kept. I read it differently: a plant at 54% expands for one reason, returning demand, and demand came back on fuel prices and a record quarter, not on a council vote. An honesty check on my own record: in that February piece I predicted IG Metall would convert its votes into a seat majority. It went the other way. The union call missed. The capacity math did not.

On the cell factory, hold Tesla to its own paper trail. Battery production at Grünheide has been promised since the site opened in 2022. The audited target is now full production in 2027, and Thierig’s 18 GWh ambition carries the same start date. That is a concrete, falsifiable deadline from a company with a documented habit of missing self-set dates, and I will be checking it. Two things to watch: whether the 7,500 vehicles per week ramp actually holds through Q4 against the 54% baseline this report establishes, and whether cells come off the line in volume in 2027. If both land, Giga Berlin becomes the counterexample to every Tesla timeline joke I have ever written. The filing gives it a fair shot. The history says verify first.

Source: Tesla Manufacturing Brandenburg SE annual report 2025 (Bundesanzeiger filing), Not a Tesla App.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Stages 245 Cybercabs at Giga Texas, Hitting EVXL's 200-Unit Call Ten Weeks Early

Haye Kesteloo — Sun, 19 Jul 2026 20:50:26 GMT

A drone flyover of Gigafactory Texas on Saturday, July 18, 2026, captured 245 Tesla Cybercab units staged across the Austin factory’s lots, a new record and more than double the 102 cars observers counted just five weeks earlier. The gold two-seaters sat in tight rows across multiple staging areas, some in the familiar outbound lot, others spread across a second parking field on the campus.

EVXL has tracked this staging count since six Cybercabs sat in simultaneous crash testing back in February, and the curve keeps steepening. Roughly 60 units in early April. More than 70 by mid-May. A record 102 on June 14. Now 245, a tally not yet independently confirmed but counted from aerial footage shot by the same drone channel that documented the June record.

The line is sprinting. The question that matters has not changed: where do these cars go?

The Staged Fleet More Than Doubled in Five Weeks

The July 18 count of 245 units marks a 140% jump from the 102 cars EVXL reported at the same factory on June 14, which was itself a record at the time, and it confirms the production ramp has moved well past its early stumbles. The first production Cybercab rolled off the Giga Texas line on February 17, 2026, and volume manufacturing followed in April. If the observer counts hold, Tesla built and staged roughly 143 additional finished units in 34 days, a pace of more than four cars per day added to the lots alone. Actual output runs higher, since some units have already shipped to test sites and showrooms in other cities.

The reflective gold finish that makes these counts possible from the air is not paint. Tesla applies a mirror-like coating that throws back the Texas sun, a detail drone observers have flagged since the first glossy units appeared in late April. It also makes miscounting hard. These cars are not subtle.

What makes a lot this full legally possible is Tesla’s decision to self-certify the Cybercab under FMVSS, the federal safety standards, rather than seek a NHTSA exemption. As EVXL detailed when the EPA certification filing surfaced in June, that route sidesteps the 2,500-unit annual cap that exemptions impose on operators like Waymo. The factory can build without a ceiling. And it clearly is.

A second section of the July 18 flyover shows additional staged Cybercabs lined up across a separate Gigafactory Texas parking area. Credit: Airwave Dynamics via YouTube.

Deployment Still Trails the Production Line

Not one of these 245 Cybercabs has carried a paying passenger, because the vehicle remains in engineering validation while Tesla’s commercial robotaxi service in Austin continues to run on Model Y vehicles fitted with conventional controls, steering wheel and pedals included. Cybercabs began public-road testing in Austin around the end of June with a safety monitor aboard, and the Texas Department of Transportation has confirmed the controlless design of the roughly 34 units testing downtown. On July 11, Tesla announced that employee rides would start soon at Giga Texas, not on the public network.

The Model Y fleet is scaling in parallel. EVXL reported that Tesla registered 58 robotaxis in Texas in a single day on July 15, lifting its state total to 175 vehicles, still behind Waymo’s 642. Every one of those additions was a Model Y. The purpose-built robotaxi, the one stacking up by the hundreds outside the factory that builds it, is not yet on that tracker earning its keep.

So the picture on July 18 is this: 245 finished Cybercabs in the lots, 34 or so validating downtown with minders, employee shuttle duty pending on factory grounds, and a commercial network that still runs entirely on retrofitted crossovers.

EVXL’s Take

On June 14, I wrote that before the end of Q3 2026 we should expect a drone flyover showing more than 200 staged Cybercabs at Giga Texas with no commercial fleet deployment to match it. That flyover arrived July 18, ten weeks ahead of the deadline, and both halves of the call landed. The lot filled before the network did.

I take no victory lap on the second half, because it is the half that should worry Tesla. A 245-car pile of finished robotaxis is capital parked on concrete. These cars cannot be sold to consumers, cannot be driven by a human, and cannot earn a fare until unsupervised FSD clears Tesla’s own validation bar in a vehicle with no fallback controls at all. Elon Musk has spent a decade promising autonomy timelines the software could not keep. Now the factory is keeping timelines the software cannot match, which is a more expensive version of the same problem.

The honest read: this is a genuine manufacturing achievement, likely the fastest ramp of a purpose-built autonomous vehicle anyone has managed, and it proves nothing about the service until a Cybercab picks up a stranger and takes their money. Whether the employee rides announced on July 11 convert into public Austin service before the staged count crosses 300 is the open question worth watching. At the current build rate, Tesla has about two weeks to beat its own parking lot.

Source: Giga Texas drone flyover, July 18, 2026 (Airwave Dynamics)

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Model Y With FSD Is One Dutch Father's Answer to the Blackout Crash That Took His License

Haye Kesteloo — Sun, 19 Jul 2026 00:18:47 GMT

Wilco de Kreij spent the past year in the passenger seat. The Dutch entrepreneur blacked out behind the wheel a year ago, crashed straight into another car with his three children in the back, and lost his license to an epilepsy diagnosis. This week, days after getting that license back, he picked up a Tesla Model Y with FSD Supervised and explained why in a post on X that struck a nerve well beyond the Netherlands.

“Being allowed to drive and feeling safe are two different things,” de Kreij wrote. His youngest child was two months old at the time of the crash. Everyone walked away, but the fear of a second blackout did not. The Model Y is his answer: a licensed, medically cleared driver adding a system that keeps watching the road if he can’t.

A Blackout, a Crash, and a Year in the Passenger Seat

De Kreij’s crash happened in a Volkswagen Passat, the family car on a vacation trip, when he lost consciousness without warning and drove straight into another vehicle; the diagnosis that followed, a form of epilepsy, took his license for a full year, a detail he shared in replies to his own post.

His wife drove him everywhere for that year, three young kids in tow. When the license came back last week, the legal question was settled. The medical one wasn’t. Epilepsy can be managed, but de Kreij was blunt about the residual risk: he hopes another blackout never comes, and knows nobody can promise that.

One detail the feel-good framing skips: de Kreij already owned a Tesla. His previous Model 3 ran Hardware 3, which the European FSD build does not support. The safety net he wanted required a new Hardware 4 car plus Tesla’s €99 monthly subscription. The safety layer is real. So is the price of admission.

The Purchase Lands Three Months After the Approval EVXL Tracked From Denial to Sign-Off

FSD Supervised became legal for Dutch consumers on April 10, 2026, when the RDW (Rijksdienst voor het Wegverkeer, the Dutch vehicle authority) issued Europe’s first type approval for the software, a decision that came five months after the same regulator publicly corrected Tesla’s premature approval claims.

That November 2025 episode, in which the RDW denied a committed approval date and asked Tesla fans to stop calling its offices, is worth remembering when the purchase stories start rolling in. The regulator took more than 18 months of track and road testing before signing anything. Since April, Lithuania, Estonia, Denmark, and Belgium have recognized the Dutch approval, and the five-country fleet passed 50 million kilometers on FSD Supervised on July 14.

The Dutch numbers frame de Kreij’s decision. The RDW’s June statement counted nearly 40,000 FSD-equipped Teslas in the Netherlands covering roughly 24 million kilometers since approval, in the regulator’s words, “without any relevant incidents.” He is not an early adopter taking a flyer on unproven software. He is joining a monitored national fleet with a published track record.

FSD Supervised Watches the Driver and Stops the Car When the Driver Stops Responding

The approved system pairs its driving software with cabin-facing driver monitoring: eye-tracking cameras check whether the driver is watching the road, escalating visual, audio, and haptic alerts fire when attention drops, and if the driver never responds, the software is designed to bring the car to a controlled stop.

That controlled-stop behavior is the part of the type approval that maps directly onto de Kreij’s fear. A 2025 Passat with an unconscious driver keeps going until physics decides. A Model Y running FSD Supervised with an unresponsive driver is engineered to slow down and stop. For a father who has lived the first scenario, the difference is not a spec-sheet line.

The limits matter just as much. The RDW classifies FSD Supervised as a driver assistance system, not autonomy, and full legal responsibility stays with the human. De Kreij can use it because he holds a valid license and is medically cleared to drive. The system supplements a fit driver. It does not replace an unfit one, and the RDW’s approval documents say so in plain language.

EVXL’s Take

This is the story the April 10 approval was written for, and I say that as the person who spent nine months documenting every slipped date and premature Tesla claim along the way. A legally cleared, medically fit driver using a supervised system as redundancy against a risk he can name: that’s the strongest real-world argument for FSD Supervised I’ve seen since the RDW signed off.

It’s also where I want to pump the brakes, because the replies under his post show how fast this story inflates. One commenter wrote that FSD would preserve freedom of movement for people who may never drive again. Another declared it will give the elderly and disabled their lives back. That is not what the RDW approved. It approved an assistance system that legally requires a fit, attentive human at the wheel, and it went out of its way to say so. The gap between de Kreij’s careful reading and his comment section’s hopeful one is the gap between a safety layer and the next liability headline.

De Kreij bought redundancy. His replies are buying autonomy. Only one of those is for sale in the Netherlands, and Tesla’s product name keeps blurring the two.

Source: Wilco de Kreij, RDW.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Adds 58 Robotaxis To Texas Fleet In One Day

Haye Kesteloo — Wed, 15 Jul 2026 21:46:58 GMT

Tesla added 58 more Model Y vehicles to its Texas robotaxi fleet on Wednesday, according to the Texas Autonomous Vehicle Tracker, pushing the company’s total registered fleet in the state to 175.

The tracker page now shows three numbers worth watching: 175 total vehicles, 58 added today, and 0 days since last added. Its fleet-growth chart also shows Tesla jumping from 117 vehicles on July 14 to 175 on July 15, which makes today’s increase the biggest single-day step visible on the page.

Texas Fleet Growth Suddenly Looks More Aggressive

The tracker says the new vehicles were registered since midnight CST and lists fresh Jul. 15, 2026 VIN entries in its latest registrations table. The newest entries shown are all Model Y vehicles, which fits the vehicle Tesla is already using as the base for its early robotaxi rollout.

That matters because this is not a marginal increase. A 58-vehicle jump in one day expands Tesla’s Texas-registered robotaxi pool by nearly 50% from the prior day’s 117 total. Even if registration does not mean every vehicle is already carrying riders, it does show Tesla still feeding inventory into the program instead of freezing the fleet where it is.

Tesla Is Still Behind Waymo On Raw Texas Fleet Size

The same tracker still shows Waymo well ahead on total Texas registrations, with 642 vehicles, versus 317 for Avride, 175 for Tesla, and 44 for Zoox. So Tesla is gaining scale, but it is not yet the largest autonomous fleet on paper in the state.

EVXL’s Take

This is the kind of data point Tesla fans will run with, and fairly so. A one-day addition of 58 vehicles is real movement, not another vague autonomy promise on a conference call.

But registration growth is not the same thing as robotaxi success. The next question is whether these extra Model Ys quickly translate into real service coverage, ride volume, and economics that hold up under scrutiny. For now, the tracker shows momentum. It does not yet prove Tesla has won Texas.

Source: Texas Autonomous Vehicle Tracker

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla FSD Hits 50 Million Kilometers in Europe, 28 Times the Dataset That Won Its Approval

Haye Kesteloo — Tue, 14 Jul 2026 20:50:54 GMT

Tesla’s European fleet has driven more than 50 million kilometers on FSD Supervised, the company announced on July 14, 2026. The total comes from customers in the Netherlands, Estonia, Belgium, Lithuania, and Denmark, the only five countries on the continent where customers can legally use the driver assistance software. The post from Tesla’s Europe, Middle East and Africa account drew over 410,000 views within hours.

The milestone lands 95 days after the Dutch vehicle authority RDW issued the first European type approval, a saga EVXL has tracked since the regulator publicly corrected Tesla’s premature claims in November 2025. Just two days ago, our coverage of a Dutch driver forced to switch FSD off at the German border put the Netherlands tally alone at roughly 24 million kilometers.

Five Countries Took the Fleet From Zero to 50 Million Kilometers

The 50 million kilometer total accumulated in 95 days across the Netherlands, Estonia, Belgium, Lithuania, and Denmark, the only five European countries where FSD Supervised is legal, which works out to an average of more than 525,000 kilometers of supervised driving added every single day since April 10.

The real pace is steeper than that average. The Netherlands drove alone until Lithuania joined on May 20 and Estonia on May 29, and, as EVXL reported when Denmark and Belgium signed off within 48 hours of each other, the last two clearances arrived on June 9 and 10. Tesla’s own June 9 tally claimed 23.6 million kilometers in the Netherlands through June 5. The fleet has more than doubled its lifetime total in the five weeks since.

The base doing the driving is small. The RDW’s June statement counted nearly 40,000 FSD-equipped Teslas in the Netherlands, covering about 24 million kilometers since approval, in its words, “without any relevant incidents.” Access costs €99 per month, subscription only, on Hardware 4 cars exclusively, which locks out much of the existing European fleet. Fifty million kilometers came from a fraction of Tesla’s cars in five smaller markets.

Customer Kilometers Now Dwarf the Dataset That Won the Approval

The RDW based its April type approval partly on 1.8 million kilometers of FSD Supervised driving collected in Europe over more than 18 months of testing, which means European customers have now produced roughly 28 times that entire pre-approval dataset in a little over three months.

That gap is the logic of Tesla’s whole approach to autonomy. FSD Supervised runs on neural networks trained end to end on fleet video, so every customer kilometer through a Dutch roundabout, a Vilnius intersection, or a Danish coastal road in fog becomes potential training material for the Europe-specific build, which the RDW insists differs from the American software. Tesla’s global fleet has passed 11.9 billion FSD miles per the company’s safety data, but almost none of it was earned on European roads until April.

The usual caveats travel with the number. This is a Level 2 system, so every kilometer measures a driver-plus-software combination, not the software alone. The figure is also Tesla’s own, from a company Reuters caught feeding regulators safety statistics that independent researchers called misleading, and the RDW said in June it could not confirm Tesla’s mileage claims.

EVXL’s Take

The 50 million is marketing. The slope is the story. Tesla posts round numbers because they travel well on X, and I discount self-published stats from a company with Tesla’s record on safety data. But the rate doesn’t need Tesla’s spin: a subscription-only, Hardware 4-only fleet in five smaller markets is doubling its lifetime mileage in five-week windows. You can’t fake that adoption with a press release.

Here’s what the 22 holdout countries should sit with. Germany, France, and Italy aren’t pausing Tesla’s learning curve by waiting. They’re only deciding whose roads feed the training set. The EU build keeps improving on Dutch roundabouts and Baltic winters either way; the holdouts guarantee it learns nothing about an Autobahn merge until years after everyone else. If the RDW’s monthly reports keep coming back clean, deliberation stops looking like caution and starts looking like a decision to stay out of the dataset.

Don’t let the odometer stand in for the safety case, though. Kilometers prove usage, not safety, and Tesla has a documented habit of blurring that line. The number that matters is in the RDW’s monthly in-use reports, not on X. In June I called at least eight national recognitions by December 31. Today’s post makes that bet look conservative.

Source: Tesla Europe, RDW, TeslaNorth.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla FSD V14.3.5 Lets You Watch Any Camera, Including the Cabin View, While Driving

Haye Kesteloo — Mon, 13 Jul 2026 20:20:00 GMT

Tesla pushed FSD V14.3.5 to the fleet this week, and buried in the changelog is a small interface tweak that Tesla watcher Sawyer Merritt flagged as one of the update’s most useful additions: drivers can now open Camera Preview at any time, even while the car is actively driving. Previously, the feature only worked when the vehicle was parked, according to Merritt’s post on X, which included a screenshot of the split-screen view, navigation on one side, a live camera feed on the other, running while the car was in motion.

Cabin Camera Access Is the Real Story

The update covers every camera Tesla vehicles carry, including the interior cabin camera that watches the driver. That detail matters more than a simple UI unlock. Parents driving with kids in the back seat can now pull up the cabin view mid-trip to check on them without pulling over. Tesla’s cabin camera has mostly lived in the background, quietly feeding the driver-monitoring system that watches for attentiveness. Making it accessible on demand, while driving, turns a passive safety sensor into something drivers can actually use.

Merritt’s screenshot shows the feed with a small disclaimer overlay, “This live stream is not recorded,” a reminder that Camera Preview is a live monitoring tool, not a way to pull saved footage. For that, drivers still need TeslaCam and the vehicle’s onboard storage.

What Else Shipped in V14.3.5

Camera Preview wasn’t the only change. Per official notes tracked by Not a Tesla App, V14.3.5 rewrote Tesla’s AI compiler using MLIR, delivering what Tesla says is a 20 percent faster FSD reaction time. The update also upgraded the vision encoder, improving how the car reads rare, low-visibility scenes and expanding traffic sign recognition.

Tesla says the release reduces unnecessary lane biasing and tailgating, increases decisiveness picking parking spots, and improves responses to emergency vehicles, school buses, and right-of-way violators, with added training for small-animal safety.

The same branch brought Actually Smart Summon to the Cybertruck for the first time, plus dashcam clip encryption, expanded parental controls, and better eye-gaze tracking.

EVXL’s Take

None of this is a headline leap for autonomy, and Tesla isn’t pitching it as one. It’s the kind of incremental usability fix that matters more in daily life than a marquee announcement. Letting a parent glance at the back seat without stopping, or letting any driver spot-check a blind-spot camera mid-drive, is the sort of change that gets used constantly once people know it exists.

It’s also a reminder of how Tesla ships software now. Big swings like the MLIR compiler rewrite land in the same point release as something as mundane as unlocking a camera button, and the small stuff is often what drivers notice first. Merritt’s post pulling nearly 30,000 views in under half an hour says as much about Tesla owners’ appetite for these details as it does about the feature itself.

EV Drivers Now Save $43 A Month On Fuel In Nearly Every State, And It Took A War To Prove It

Haye Kesteloo — Mon, 13 Jul 2026 13:40:00 GMT

American EV drivers now pay $43 a month less to fuel their cars than gasoline drivers do, and the advantage holds in nearly every state. Average monthly gas spending has climbed to $149, a 37 percent jump since the U.S. went to war against Iran in February, while EV drivers pay an average of $106 a month to charge, up 16 percent over the same stretch, according to fuel-price tracker OPIS and EV charging data platform Eco-Movement, both Dow Jones companies.

That gap, worth $516 over a year at current rates, lands on a pattern EVXL has tracked since the war’s opening weeks. When we covered the pump shock on April 5, gas had just crossed $4.02 a gallon, up 35 percent in five weeks, while regulated residential electricity rates barely moved. Three months later, the pump has not given that ground back, and the fueling math has flipped in states where gas held the edge as recently as January.

Gas Spending Jumped 37% While Charging Costs Rose 16%

The average American drives 943 miles a month, and powering that driving with gasoline now costs $149 a month against $106 for an EV on charging, a spread produced by a 37 percent gas increase colliding with a 16 percent rise in charging costs since February. The figures come from a state-by-state analysis published July 13 by The Washington Post, reported by Naema Ahmed and Julie Z. Weil. The methodology assumes 31 miles of daily driving, 25 miles per gallon for gas cars, and 3.5 miles per kilowatt hour for EVs.

Higher oil prices lift electricity rates too, since some power generation still burns fossil fuels. The difference is speed. Retail electricity rates are set annually by regulators, so the oil shock reaches home charging bills slowly and partially. Pump prices reprice in days. Wisconsin shows the split in its starkest form: gas surged there while electricity rates barely budged, per the Post’s data.

The human version of that spread showed up at a Wawa off the Pennsylvania Turnpike, where the Post found carpenter Que Walker refueling the gas-powered Dodge Charger he bought in January, weeks before the war began. “It’s an extra $30 to $45 every week just out the window,” Walker said. He told the paper he’d be shopping EVs if he were buying today.

Eight States Lost The Last Gas-Price Advantage In The Country

Before the war began, eight states, including West Virginia, still made monthly gas fueling cheaper than EV charging, and that list has now shrunk to nearly zero as pump prices climbed faster than the regulated utility rates that set most home charging costs across the country. The flip matters because those eight states were the last data points gas defenders could cite. They’re gone from the ledger.

A Bloomberg analysis published in late June ran the same comparison with annual numbers: swapping a gas car for a home-charged EV now saves the average American driver close to $1,500 a year, with Washington state leading at $2,346 and Oregon at $2,057. Electricity prices did rise 8.6 percent over the past 12 months, partly on AI data center demand, but that increase is a rounding error next to the surge at the pump.

Cheaper Fueling Meets A Market That Lost Its $7,500 Crutch

The fuel-cost advantage arrives in a new-EV market still shrinking from the loss of the federal $7,500 purchase credit, which expired September 30, 2025, and which pulled first-quarter EV sales down 28 percent year over year to 212,600 units, according to Cox Automotive data. We documented the first month of that collapse when October 2025 sales cratered 24 percent from September’s pre-deadline rush.

Hybrids caught the demand EVs dropped. Hybrid sales hit a record 756,000 units in Q4 2025, up 57 percent year over year, and hybrid share overtook zero-emission vehicle share in California for the first time in Q1 2026. The Post’s own framing captures the tension: EV sales plummeted after the credit died, yet the fueling economics have never favored electric more clearly than they do right now.

EVXL’s Take

For the first time since I started covering this market, the operating-cost case for EVs is standing entirely on its own. No tax credit. No mandate. Just a $43 monthly gap in nearly every state, produced by market forces Washington didn’t design. That’s the case EVXL has argued subsidies were always papering over, and I’d rather see it proven this way than by another decade of $7,500 checks.

Two caveats before anyone laminates that $516 figure. First, the $106 average is built on state electricity rates, the price home charging pays. Drivers who depend on public fast charging pay up to 60 cents per kilowatt hour, a rate that erases most of the gap, and apartment dwellers don’t get a vote on that. Second, this is a war premium, not a moat. Oil has already retreated from its $119 peak, and the 2022 Ukraine spike taught us that fuel-price converts evaporate when the pump cools off. I flagged that false-peak risk in April and nothing since has changed my read.

So here’s my position: the industry has a closing window to convert fuel-cost shoppers into owners while the gap is this visible, and the used EV market is where that conversion actually happens at post-credit prices. Watch the Q3 used EV transaction data when it lands this fall. If dealers can’t move affordable used electrics with gas at $149 a month, the problem was never the fuel math. It was everything else the industry still hasn’t fixed.

Source: The Washington Post, with additional data from Bloomberg (via Claims Journal) and Cox Automotive.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla FSD Outdrives Humans in the Data While Europe Approves It One Border at a Time

Haye Kesteloo — Sun, 12 Jul 2026 19:09:00 GMT

A Dutch Tesla driver with nearly 10,000 kilometers on FSD Supervised will switch the system off next week at the German border. Not because it stopped working, but because Germany has not approved it. Hans Noordsij posted on X on July 10, 2026, that he had covered a 240-kilometer family trip the night before without a single intervention, then noted his upcoming vacation drive from Venlo to Passau crosses a country where the software he trusts remains illegal. The post drew 385,000 views in a day.

His frustration is the European FSD story in miniature. EVXL has tracked this approval saga since the RDW publicly corrected Tesla’s premature claims in November 2025, and the pattern holds: the safety data keeps stacking up on one side while the approval map fills in one border at a time.

Five Countries Have Approved FSD Supervised, Twenty-Two Have Not

The Netherlands approved FSD Supervised on April 10, 2026, Lithuania followed on May 20, Estonia on May 29, Denmark on June 9, and Belgium on June 10, which leaves Germany, France, Italy, Spain, and 18 other EU member states without a national approval as of this weekend. The pace within that group of five was quick. As EVXL reported when Belgium and Denmark signed off within 48 hours of each other, the gap between national approvals shrank from 40 days to a single day.

Every one of those clearances is provisional. They all rest on the type approval the RDW issued under UN Regulation 171 and an Article 39 exemption, the decision EVXL covered on April 10 after the Dutch regulator spent more than 18 months testing the system. If the European Commission rejects the software at the bloc-wide level, the Dutch approval lapses after six months and takes every national recognition down with it. The Commission’s Technical Committee on Motor Vehicles has discussed the file but has scheduled no vote.

Germany’s KBA is still reviewing. That is why Noordsij’s post tags Germany’s federal transport ministry directly: he can drive hands-free from his front door to the German border and not one meter past it.

The Safety Numbers Keep Pointing the Same Way, Caveats Included

Tesla’s Vehicle Safety Report, as of its February 2026 update, puts FSD Supervised at one major collision every 5.3 million miles in North America against a US average of roughly 660,000 miles, and the RDW counts about 24 million kilometers driven in the Netherlands since April, in its words, “without any relevant incidents.” The February update to Tesla’s report, covered by Not a Tesla App, also confirmed the global fleet had passed 8.2 billion cumulative FSD miles.

The caveats are real and EVXL has reported them. FSD Supervised is a Level 2 system, so incident-free supervised miles measure the driver-plus-software combination, not the software alone. Safety researcher Philip Koopman has detailed how Tesla’s baseline comparisons flatter the system, since FSD owners skew toward safer demographics and newer cars. And the EU build differs from the US software, a distinction the RDW itself insists on. A Reuters investigation into Tesla’s marketing statistics reached the Dutch parliament, where the transport minister defended the RDW on June 16 by pointing to the agency’s own independent testing rather than Tesla’s slides.

Discount the numbers as hard as you like. Cut Tesla’s ratio in half for demographics and road mix, ignore the company’s marketing entirely, and lean only on what an independent European regulator measured across 18 months of its own testing and two months of live monitoring, now reported monthly instead of annually. The direction never flips. The supervised system plus an attentive human is producing fewer wrecks than the human alone.

The Lidar Cost Argument Died in a $1,099 Drone

New Jersey’s Senate Bill S1677 would require robotaxis to pair cameras with two additional sensor types, and the redundancy argument behind it deserves a straight answer now that lidar, once a $100,000 research instrument, ships in production Chinese EVs and in a $1,099 consumer drone. As EVXL reported yesterday, the bill’s sponsor rests his case on physics: cameras struggle in glare and heavy rain, radar cuts through fog, and lidar maps the road in full darkness.

Hesai, the Chinese sensor maker supplying Mercedes-Benz’s Level 3 program, says it has cut lidar pricing from $100,000 down to roughly $200 per unit, according to its co-founder Kai Sun. The company is expanding capacity to 4 million units this year, and solid-state units already sell in the $400 to $500 range, per The Robot Report, while IEEE Spectrum reports US supplier MicroVision is targeting sub-$200 production pricing. The technology is so cheap it flies: DJI’s Air 3S, a $1,099 camera drone our sister site DroneXL reviewed in October 2024, carries a forward-facing lidar unit for obstacle avoidance in the dark.

Elon Musk’s case against lidar was never only about money. He argues roads were designed for eyes and neural networks, and that fusing conflicting sensor inputs adds failure modes rather than removing them. That is a legitimate engineering debate for a supervised Level 2 system, where the human is the redundancy. It gets much weaker for the unsupervised robotaxi future Tesla is selling, where nothing backs up the cameras when a lens is blinded by low sun or caked in slush. GM’s eyes-off Cadillac program, Mercedes-Benz’s Drive Pilot, and Waymo all chose sensor fusion. When the backup sensor costs $200, the burden of proof sits with the company leaving it out.

EVXL’s Take

I’ll hold both positions at once, because the story demands it. FSD is not what Elon Musk promised. We were told robotaxis by 2020, and what shipped six years late is a very good Level 2 system that still needs your eyes on the road. EVXL has held Tesla to that record for years, and the “Full Self-Driving” name still writes checks the software can’t cash.

And yet Europe’s slow-walk has become its own safety problem. The RDW built the template: test independently for 18 months, approve provisionally, monitor monthly, and keep a kill switch. Two months and 24 million kilometers later, its data shows no relevant incidents. If regulators believe their own numbers, every month Germany, France, and Italy spend deliberating is a month of purely human-driven miles that the evidence says would be safer supervised. Noordsij will cross Germany next week the old way, at exactly the moment Dutch roads are demonstrating the alternative. That’s not caution anymore. That’s inertia dressed up as caution. Copy the Dutch homework, keep the monthly reporting, and approve.

On sensors, I’d flip the challenge back at Tesla. The vision-only bet made sense when lidar cost more than the car. At $200 a corner, with the sensor now standard on Chinese EVs and bolted to consumer drones DroneXL has covered since 2024, refusing redundancy for the unsupervised era saves pennies and stakes the entire robotaxi program on a lens never getting blinded. Cameras got FSD this far. Cheap lidar is how it survives its first blizzard without a human in the seat.

Watch two dates: the Commission’s TCMV calendar, still without a scheduled vote, and Germany’s KBA, which faces mounting pressure from its own Tesla drivers. When we covered the Belgium and Denmark approvals on June 10, we called at least eight national recognitions by December 31. Nothing since has moved that bet.

Sources: Hans Noordsij (X), RDW, Tesla, optics.org, The Robot Report, IEEE Spectrum.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla's New Patent Could Keep You Cool in a Heatwave Without Losing the Glass Roof

Haye Kesteloo — Sat, 11 Jul 2026 00:00:00 GMT

Tesla has filed a patent for a glass roof structure designed to stop cabin overheating without giving up the automaker’s signature panoramic glass, according to a filing published by the US Patent and Trademark Office.

A Perforated Inner Pane With an Air Gap Wired to the HVAC System

The patent, titled “Automotive Perforated Glass Structure,” describes a double-layered glass panel: a solid outer pane for weatherproofing, and a separate inner pane facing the cabin that’s perforated with thousands of tiny holes. Between the two layers sits an air gap — small enough to stay hidden inside the roof structure but large enough to carry airflow.

That gap connects directly to the vehicle’s climate control system. Cool or warm air is pumped into the space between the panes, then filters evenly through the perforations and into the cabin, diffusing air across the entire roof surface rather than blasting it from dashboard vents. A honeycomb structure inside the gap keeps the dual-pane assembly rigid while also acting as an acoustic dampener, letting the design cut down on road and wind noise at the same time it manages temperature.

Aimed at Glass Roofs’ Biggest Complaint

Tesla’s panoramic glass roofs have long drawn praise for structural strength and cabin openness, but also a common owner complaint: they turn into greenhouses on hot, sunny days, since tinting and UV coatings only do so much to block solar heat gain. By ventilating the glass itself rather than just treating it, Tesla could neutralize heat closer to where it enters the cabin — potentially clearing the way for even larger glass surfaces in future models without making the overheating problem worse.

It’s unclear whether or when this design would reach production. Patent filings describe engineering possibilities a company is protecting, not confirmed roadmap features.

EVXL’s Take

Tesla has filed a string of cabin-comfort patents this year aimed at the same complaint: glass roofs look great and feel spacious, but they bake occupants in direct sun. This one is clever because it doesn’t try to fix the problem with better glass — it turns the glass itself into the vent. Worth remembering this is a patent, not a production announcement; plenty of automaker patents never make it past a filing cabinet. But if Tesla can pull off diffuse ceiling-level cooling without the dashboard blast, that’s a genuine passenger-comfort upgrade, not just a spec-sheet curiosity.

Source: USPTO filing via Stuff

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Files to Open a Dallas Showroom on Knox Street, Steps from Rivian's Spaces Location

Haye Kesteloo — Fri, 10 Jul 2026 22:30:00 GMT

Tesla has filed to open a new Dallas showroom at 3113 Knox Street, according to a Texas Department of Licensing and Regulation (TDLR) registration, putting the automaker on the same retail strip where rival Rivian opened a Spaces showroom last year.

The Filing Details a 2,513-Square-Foot Tenant Buildout

The TDLR project record, filed under registration TABS2026022334, describes a “2,513 SF tenant improvement for ‘Tesla, Inc.’ auto sales including site improvements” at the Knox Street address. The filing was registered June 8, 2026, cleared design review July 1, and lists an estimated construction start of August 15, 2026, with a targeted completion around December 2027. Property records list the Tabesh Family Trust as the site’s owner and WT Group AEC as the design firm of record. The Dallas Morning News, which first reported the filing, notes such preliminary TDLR details are subject to change before a store actually opens.

The estimated project value is roughly $101,000, a modest buildout cost consistent with a showroom-and-sales space rather than a service center.

Tesla Would Land Directly on Rivian’s Turf

Tesla previously had a presence at nearby NorthPark Center but has since exited that spot, per the Dallas Morning News. The new Knox Street location would place Tesla on the same strip where Rivian unveiled its Spaces retail concept, a showroom-style format built for browsing and test drives rather than traditional dealership sales. Rivian’s Knox location has operated for more than a year.

Tesla already operates other Dallas-area sites, including in Plano and Flower Mound, according to the Dallas Morning News’s reporting. The Knox Street addition comes as Cox Automotive’s Q2 2026 report shows the broader US EV market grew quarter-over-quarter but remained down year-over-year.

Knox Street Is Filling Up With New Retail Names

Tesla wouldn’t be the only new arrival to the neighborhood. Jewelry brand Hart, a South Carolina-based company, has said it will open its first Texas store at 3212 Knox St. — the former Pottery Barn space — following the recent launch of a permanent New York shop. Activewear brand Alo is also filing for a more-than-6,000-square-foot store at 3333 Knox St., with an estimated cost topping $1 million and a targeted completion around November, per city and TDLR filings.

EVXL’s Take

A TDLR filing is a real regulatory record, not a rumor, but it’s also not a grand opening — Tesla’s own projected completion date sits more than a year out, in December 2027, and the filing itself says details remain preliminary. What is notable is the location: choosing to sit directly alongside Rivian’s Spaces showroom, rather than avoiding it, suggests Tesla sees enough retail foot traffic in the Knox Street corridor to compete head-to-head rather than carve out separate turf. Worth revisiting once construction actually breaks ground in August and again as the completion date approaches, since TDLR timelines are frequently revised.

Source: The Dallas Morning News

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

NHTSA Chief Says Agency Will 'Absolutely' Consider Dropping Steering Wheel Rule for Driverless Cars

Haye Kesteloo — Fri, 10 Jul 2026 20:45:00 GMT

The federal government’s top auto safety regulator says he will “absolutely” consider dropping the steering wheel requirement for cars designed to never have a human driver, a potential regulatory opening for Tesla’s pedal-and-wheel-free Cybercab.

National Highway Traffic Safety Administration Administrator Jonathan Morrison made the comments in a CNBC Squawk Box interview on July 9. “If you’re developing a vehicle that is designed never to be driven by a human operator, does it make any sense to require manual control for the vehicle? I think the answer is pretty clear there,” Morrison said.

The Change Would Only Apply to Purpose-Built Driverless Vehicles

Morrison’s comments follow NHTSA’s move last month to update federal safety standards and remove the mandate for manual brake pedals in autonomous vehicles. Any change to steering wheel rules would only apply to vehicles designed to operate exclusively without a human driver, per Bloomberg’s reporting picked up by Claims Journal, leaving existing requirements unchanged for every other type of car. Morrison didn’t specify a timeline for any formal steering wheel rule change.

The distinction matters for how differently automakers have approached autonomy. Cars built for rideshare fleets, including Tesla’s Cybercab, are designed without a steering wheel or pedals from the start. Others, like the vehicles Waymo operates, keep manual controls in place and can be taken over by a remote human driver if needed, according to CNET’s reporting on the interview. Removing brake pedals and steering wheels entirely means no human could take physical control if a driverless car stalls or a dangerous situation requires intervention.

Not Everyone in the Industry Wants the Rules Loosened This Fast

Spencer Penn, CEO of AI procurement company LightSource and a former Tesla engineer and Waymo product manager, told CNET that a steering wheel as a human backstop “only makes sense in the middle of that ladder, where you’re not confident the car can finish the job on its own.” But Penn also cautioned against removing manual controls before replacing them with something else: “Remove the wheel, fine, but tell me what replaces it. Passengers need a clear way to stop and get out. First responders need the car to behave predictably. Regulators need real performance numbers, not company slides.”

Morrison’s agency is simultaneously investigating separate incidents in which autonomous vehicles have stalled in construction zones or been slow to move out of the way of emergency responders, per CNET. Morrison called those incidents rare but said “every single one of these circumstances goes too far.”

EVXL’s Take

This is the clearest signal yet that Washington is willing to rewrite vehicle safety rules around Tesla’s bet rather than the other way around. Tesla has spent years building the Cybercab with no steering wheel and no pedals, effectively betting that federal rules would eventually catch up to the hardware. Morrison’s comments suggest that bet might pay off faster than skeptics expected.

But Penn’s warning is the one worth sitting with: removing a steering wheel is a hardware decision, and replacing the safety function it serves is a much harder problem NHTSA hasn’t solved yet. As EVXL reported in June, Tesla’s own driverless fleet sits at roughly 20 cars nationwide, and the Senate has separately demanded NHTSA review Tesla’s FSD safety math because the company’s self-reported figures aren’t independently auditable. Loosening the hardware rules before there’s a real answer to “how does a passenger stop the car” is a sequencing risk, not a done deal. Watch whether NHTSA actually opens a formal rulemaking process, and watch whether it demands audited safety data as the price of admission. Morrison’s words were an opening position, not a new regulation.

Source: CNET

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Is Hiring 'AI Safety Operators' in 32 Cities Across 20 States, Signaling a Wider Robotaxi Push

Haye Kesteloo — Fri, 10 Jul 2026 19:10:00 GMT

Tesla is actively recruiting for “AI Safety Operator” roles in 32 cities spanning 20 states and Washington, D.C., according to job listings compiled by Tesla-focused X account Ming (@tslaming) on July 10. The listings, filed under Tesla’s Engineering & Information Technology career track, cover a footprint far wider than the markets where Tesla’s Robotaxi service currently operates.

The Listings Span Far Beyond Tesla’s Live Robotaxi Cities

Per Tesla’s own Robotaxi page, driverless rides are currently offered only in Austin, Dallas, and Houston, Texas, and in Miami, Florida. The new “AI Safety Operator” postings reach well past that list: Arizona, Georgia, Illinois, Kansas, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Nevada, New Jersey, New York, North Carolina, Pennsylvania, Tennessee, Utah, and Washington state all show listings, on top of confirmed cities like Reno, San Francisco, San Diego, and Washington, D.C.

Job titles like “AI Safety Operator” replace the “Rental Readiness Specialist” and earlier “Safety Driver” postings Tesla has used to staff up ahead of past Robotaxi launches, per Teslarati’s reporting on similar hiring patterns tied to Tesla’s international expansion.

California listings are worth a caveat. EVXL has previously reported that Tesla’s California “Robotaxi” service is a chauffeured limo-permit operation, not a true driverless service, per the state’s own public utilities regulator. New AI Safety Operator listings in California cities don’t confirm driverless deployment there is imminent.

EVXL’s Take

Job listings are a leading indicator, not a launch date, and Tesla has a well-documented habit of staffing up markets years ahead of an actual driverless rollout. But a 32-city, 20-state hiring footprint is a much bigger signal than a handful of test cities, and it lines up with a pattern EVXL has tracked all year: Tesla’s actual driverless fleet remains small (roughly 20 cars nationwide as of Tesla’s own Nevada filing in June), while its public ambitions keep scaling faster than its hardware or regulatory approvals can follow.

Watch whether any of these new states start seeing actual driverless permits filed, not just job postings. Until then, this reads as Tesla building the operations bench for a national rollout it hasn’t been cleared to run yet.

Source: Ming (@tslaming) on X

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla Gets Written Out of New Jersey's Robotaxi Bill, and Its Own Record Helped Draft It

Haye Kesteloo — Fri, 10 Jul 2026 17:59:00 GMT

New Jersey lawmakers are weighing a bill that would keep Tesla’s robotaxi out of the state before it ever carries a passenger there. Senate Bill S1677 would create a three-year autonomous vehicle pilot program overseen by the state Department of Transportation and the Motor Vehicle Commission, and it requires every participating vehicle to pair its cameras with two additional sensor types, such as lidar or radar. Tesla builds its autonomous stack on cameras alone, so the requirement works as a hardware ban on Full Self-Driving and the Cybercab that depends on it.

The bill was introduced on January 16 and now sits in the Senate Budget and Appropriations Committee after clearing the Transportation Committee this spring. Tesla has responded with open lobbying against it. The timing stings: three weeks back, EVXL documented a record count of production Cybercabs staged at Gigafactory Texas with no commercial fleet to absorb them. Every state that closes its doors makes that gold-finished inventory pile harder to explain.

The Bill Mandates Cameras Plus Two Backup Sensor Types

S1677 sets up a three-year pilot program for fully autonomous vehicles in New Jersey, overseen by the state Department of Transportation and the Motor Vehicle Commission, and it requires every participating vehicle to carry a camera system plus two distinct sensing modalities that keep working if the cameras fail. The operative language demands crash-avoidance systems built around “a camera system and two distinct sensing modalities” capable of detecting and tracking obstacles when the cameras go dark.

The requirements stack up from there. Operators must submit a law enforcement interaction plan and document their redundant safety systems and data recording capabilities. Fifty thousand supervised, crash-free miles in the state come before any human safety driver leaves the seat. The bill also restricts where autonomous vehicles can operate, keeping them out of school zones and active construction areas. For Waymo and Zoox, which already run lidar and radar alongside their cameras, the sensor clause is paperwork. For Tesla, it is a wall.

State Senator Andrew Zwicker, the Democratic sponsor of S1677 and a physicist at the Princeton Plasma Physics Laboratory, drafted the legislation after riding in a Waymo in Phoenix, and he rejects the charge that the sensor requirement was written to keep any single company out of New Jersey. “This is not anti-Tesla. I’m pro-New Jersey safety,” Zwicker told The Verge, which first reported the bill’s collision course with Tesla’s camera-only approach.

Zwicker’s argument rests on redundancy. Cameras read signs and lane markings but struggle in glare and heavy rain. Radar cuts through fog. Lidar builds a three-dimensional map of the road regardless of lighting. His position is that the evidence does not yet show a single sensor type paired with software handling everything a human driver can. New Jersey is the most densely populated state in the country, which raises the cost of being wrong. The bill’s framework mirrors recommendations from SAVE-US, a nonprofit lobbying for tighter autonomous vehicle oversight, and New York legislators are working on a near-identical mandate next door, according to Road & Track.

Tesla Is Lobbying Its Own Customers to Kill the Bill

Tesla’s response has been to activate its owner base through Engage Tesla, the company’s public policy platform, which now hosts a call to action urging New Jersey residents to contact their state legislators and oppose both S1677 and its Assembly companion, A3968, before the measure advances further. The campaign page calls the legislation “anti-competitive favoritism that will cause New Jersey to fall behind” and argues the bill denies mobility to elderly and disabled residents who cannot drive.

The page leans on New Jersey’s 578 traffic deaths last year and the well-worn statistic that human error contributes to more than 94% of serious crashes. Both numbers are real. What the page skips is the reason the sensor debate exists at all: Tesla has never submitted its safety claims to an independent audit that would let a legislator take them at face value.

The Steering Wheel Language Is the Quieter Problem for the Cybercab

Beyond the sensor mandate, the bill favors keeping traditional steering wheel and pedal controls available in autonomous vehicles, a preference that collides directly with the Cybercab, the two-seat robotaxi Tesla is now building at volume in Texas with no steering wheel, no pedals, and no side mirrors. Adding lidar to a Cybercab is a hardware retrofit. Adding a steering wheel is a redesign of the product’s entire premise.

That is why this bill lands harder than a typical state skirmish. As EVXL reported on June 14, the outbound lot at Gigafactory Texas hit its fullest visible count yet, with a reported 102 Cybercabs staged and waiting while deployment lags behind production. A factory that can build cars faster than regulators will accept them needs every large market it can get. New Jersey and New York together would carve the densest metro region in America out of the map.

EVXL’s Take

I’ll say the part that cuts against my own instincts first: I don’t love prescriptive hardware mandates. Writing sensor counts into statute freezes today’s engineering consensus into law, and if camera-only autonomy someday proves itself, New Jersey will be stuck amending a statute instead of updating a performance standard. Regulating outcomes beats regulating parts lists.

But Tesla handed legislators the pen. Performance-based regulation only works when regulators can trust the performance data, and Tesla has spent years making that impossible. When EVXL covered the Senate letter on Tesla’s FSD safety math on June 16, the core complaint was that Tesla picks its own denominators and publishes ratios nobody can audit. A company that won’t open its books to independent verification can’t act shocked when a physicist-legislator decides to regulate the inputs he can see instead of the outcomes he can’t.

The lobbying page’s claim of proven technology also has a scoreboard problem. When EVXL dug into Tesla’s Nevada filing on June 6, independent trackers put the company’s actual driverless fleet at roughly 20 cars nationwide, against Elon Musk’s promise of hundreds of thousands of autonomous Teslas by the end of this year. Waymo runs thousands. Calling a 20-car fleet proven while the market leader fields more than a hundred times as many vehicles is the kind of gap that makes lawmakers reach for lidar mandates.

Watch the Senate Budget and Appropriations Committee. If S1677 clears it and New York’s companion effort keeps moving, Tesla faces a hardware wall around the entire New York metro market, built partly from bricks the company fired itself. The fix was never a lobbying page. It was independent safety data, and Tesla still hasn’t offered any.

Sources: Road & Track, The Verge, New Jersey Legislature, S1677, Engage Tesla.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Tesla FSD v14 Lite Lands In South Korea, But Only US-Built Model 3 And Y Owners Get It

Haye Kesteloo — Fri, 10 Jul 2026 17:41:00 GMT

Tesla began rolling out Full Self-Driving (Supervised) v14 Lite in South Korea on July 10, making the country the first market outside North America to receive the software written for older Hardware 3 vehicles. Tesla Korea said the update goes out sequentially by vehicle and warned owners that timing will vary from car to car.

The catch sits in the fine print. Only US-built Model 3 and Model Y vehicles running the Full Self-Driving Computer 3, with FSD (Supervised) already active, qualify. The Shanghai-built cars that fill much of Korea’s Tesla fleet stay locked out. None of this should surprise EVXL readers. When Tesla excluded HW3 cars from its 1.5 million-vehicle FSD trial last December, the company was already steering those owners toward a separate Lite build rather than the full v14 experience.

Korea Becomes The Second Market For Tesla’s HW3 Software

South Korea is the second market worldwide to receive FSD v14 Lite, following the United States rollout that began June 29 with software update 2026.20.5.1, and the announcement arrived just eleven days after American early-access testers got their first taste of the distilled build. Source: Seoul Economic Daily confirmed the launch on July 10, quoting Tesla Korea’s description of the system as a driver assistance feature covering route navigation, lane changes, steering, and parking under supervision.

The timing matters because HW3 owners have waited a long time for anything at all. This is the first FSD update for the legacy fleet in roughly 14 months, per Not a Tesla App’s tracking. Those cars sat on FSD v12.6 since early 2025 while Hardware 4 vehicles cycled through the entire v14 series. EVXL flagged the widening hardware divide on December 1, when Tesla left every HW3 car out of its holiday FSD trial.

US-Built Cars Qualify While Shanghai-Built Cars Stay Locked Out

Tesla Korea limited the rollout to US-built Model 3 and Model Y vehicles equipped with the Full Self-Driving Computer 3 and an active FSD (Supervised) purchase or subscription, which excludes the Shanghai-built cars that account for a large share of Korea’s Tesla fleet. Seoul Economic Daily reports that recently imported vehicles produced at Gigafactory Shanghai, including the new Model Y, are not eligible for the update.

Tesla Korea also repeated its standard caution in the announcement on X: the system is not fully autonomous, and “the driver must always remain attentive and be prepared to take immediate control.” FSD (Supervised) remains a Level 2 driver assistance system in every market where it operates.

For a Korean owner, eligibility comes down to where the car was built, a detail no buyer chose with software access in mind. Two identical-looking Model Ys parked side by side in Seoul can now live on different software branches for years.

v14 Lite Compresses New Software Into Seven-Year-Old Hardware

Tesla built v14 Lite by distilling the driving behavior of the Hardware 4 version of v14 into a model the older AI3 computer can run, a compression job made harder by the fact that AI3 has roughly 15% of AI4’s effective memory bandwidth, per Elon Musk. Tesla AI chief Ashok Elluswamy described the build as v14’s driving model adapted to AI3’s cameras and compute when the US rollout began on June 29.

The build brings capabilities HW3 owners have never had: starting FSD from Park, automatic reversing out of a spot, and Arrival Options that let the driver pick where the car ends the route. Not a Tesla App notes that some drivers may notice slower decision-making in complex scenes, since the older chip is running a model architecture designed for newer silicon. The v14 line itself has been moving fast on Hardware 4, as EVXL covered on April 7 when v14.3 shipped with a rewritten AI compiler.

Korea’s front-of-line position is consistent. The country was the first market outside North America to receive FSD v14, and it holds an active fleet of roughly 50,000 HW3 vehicles, per Not a Tesla App. Tesla’s AI team said on X that v14 Lite will reach all eligible AI3 owners in Korea after further validation, without committing to a date.

EVXL’s Take

I’ll give Tesla’s engineers their due. Squeezing v14’s driving behavior into a computer with roughly 15% of the memory bandwidth it was designed for is real work, not marketing. Early US feedback suggests the distillation mostly holds. That’s the credit.

Now the debit. HW3 owners paid up to $15,000 for software Musk repeatedly promised would turn their cars into appreciating assets and, eventually, robotaxis. What they’ve received, seven years after that computer started shipping, is a compressed approximation of the current build, delivered 14 months after their previous update, on hardware Tesla has already renamed and moved past. And since Tesla killed the FSD purchase option in February, the company owes new customers nothing beyond what the software does this month. The people still holding the original promise are exactly these HW3 owners.

In Korea, the deal narrows further: your car qualifies only if it left a US factory instead of Shanghai. That’s not a line any owner could have seen when they signed the paperwork. Tesla says all eligible Korean AI3 owners will get v14 Lite after validation. Hold the company to it, and watch whether Europe, where HW3 owners are already organizing a collective action over being left out, gets v14 Lite as goodwill or only once lawyers force the issue. Tesla’s track record on stated FSD timelines says don’t hold your breath.

Sources: Tesla Korea, Seoul Economic Daily, Not a Tesla App.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Photo credit: Tesla

Tesla's Rental Program Is Expanding to Maryland and Kansas, Job Listings Show

Haye Kesteloo — Fri, 10 Jul 2026 16:45:00 GMT

Tesla is hiring for its direct rental program in Maryland and Kansas, according to job listings spotted by Tesla-watcher Sawyer Merritt, extending a $60-a-day rental push that started as a Southern California pilot less than a year ago.

Listings for “Rental Readiness Specialist” positions are live for Owings Mills, Maryland (Req. ID 276050) and Lenexa, Kansas (Req. ID 274235), both full-time roles under Tesla’s Vehicle Service job category. Merritt’s July 9 posting on X also names Washington and Pennsylvania as states where the program is expanding, though no job listings for those states have surfaced publicly yet.

The Job Postings Describe Fleet Work, Not Storefront Retail

Both listings use identical language: the Rental Readiness Specialist “coordinates the receipt of incoming new and used vehicle inventory” and is responsible for “fleet/lot management, movement of vehicles, vehicle readiness, rental invoicing, and customer hand off.” The role handles vehicle charging and cleaning, registration and plating paperwork, readiness inspections, and direct customer communication for pickups and drop-offs.

That’s a staffing pattern EVXL has seen before. Tesla used the same “Rental Readiness Specialist” job title when it expanded the program to Austin, Nashville, Boston, Fort Worth, and Houston in November 2025, after piloting the concept in San Diego and Costa Mesa. Tesla is running these rentals itself rather than through a third-party agency like Hertz, which has publicly scaled back its own EV rental ambitions.

The Pricing And Perks Match Tesla’s Existing Markets

Per Merritt’s reporting and a Tesla promotional page for the San Diego location, the rentals run Model 3 and Model Y at $60 a day and Cybertruck at $75 a day, typically for three-to-seven-day bookings with unlimited mileage. Renters get free Supercharging and access to Full Self-Driving (Supervised), plus a $250 credit toward a Tesla purchase if they buy within seven days of the rental. Those terms are consistent with what Tesla has offered at its existing rental locations since the November expansion, though Tesla’s own promotional material notes prices can vary by location.

Tesla hasn’t announced a launch date for the Maryland and Kansas locations, and the company doesn’t typically comment on unannounced program expansions. The job listings themselves are the only confirmation the program is coming to these markets.

EVXL’s Take

This isn’t a marketing rollout, it’s a fleet-logistics buildout, and that distinction matters. Tesla isn’t announcing rentals with a press release, it’s posting fleet-management jobs and letting Tesla-watchers connect the dots. That’s consistent with a program Tesla is scaling quietly rather than promoting loudly, because the real product here isn’t the rental itself, it’s the seven-day trial of Full Self-Driving and Supercharging that’s designed to convert into a purchase.

Every new state is another data point for how far Tesla intends to take this before Q1 2026, which is the timeline EVXL flagged back in November. Maryland and Kansas are unglamorous, non-coastal markets compared to Austin or Boston, and that’s the tell: if Tesla is comfortable running rental fleets in Owings Mills and Lenexa, this isn’t a coastal novelty program anymore, it’s becoming standard Tesla infrastructure. Watch whether Washington and Pennsylvania listings actually materialize; if they do, the program has effectively gone national in under eight months.

Source: Sawyer Merritt

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.

Rivian Files To Sell 75 Million Shares Into Its R2 Rally, The Dilution We Called In February

Haye Kesteloo — Mon, 06 Jul 2026 22:12:08 GMT

Rivian filed to sell 75 million new Class A shares after Monday’s market close, a deal worth roughly $1.5 billion at the day’s $20.14 closing price. The company says the proceeds will fund general corporate purposes, including the equity contribution requirements tied to its Department of Energy (DOE) loan for the Georgia factory.

The timing is deliberate. The filing landed four days after Rivian beat its own second quarter delivery outlook and raised full year guidance, and hours after the stock gained 8.11% in a single session. It’s also the raise EVXL said was coming when CEO RJ Scaringe told analysts in February that Rivian would be opportunistic about raising capital.

Rivian filed to sell 75 million new Class A shares Monday evening and, in the same breath, told investors the second quarter went well: preliminary revenue of $1.55 billion to $1.65 billion, up from $1.30 billion a year earlier, with cash climbing to roughly $5.3 billion. The company attributes the growth to higher deliveries, with average selling prices dented by a heavier mix of commercial vans, alongside more software development work and regulatory credit sales, according to the filing. KPMG has not audited the preliminary numbers.

The cash figure deserves a second look. The quarter that lifted Rivian’s balance from $4.8 billion to roughly $5.3 billion also included Volkswagen‘s $1 billion equity payment, received April 30 after winter testing milestones. Strip that out and the arithmetic implies operations and capital spending still consumed several hundred million dollars in three months. The new shares add roughly 6% to the share count.

The DOE Loan Requires Rivian’s Equity Before Government Money Flows

The Department of Energy loan behind Rivian’s Georgia factory works on a you-first basis: before the company can draw on the up to $4.5 billion facility, expected in early 2027, Rivian must satisfy equity contribution requirements written into the loan agreement it signed with the government. The agreement Rivian signed in January 2025 spells out base and contingent equity commitments and a completion guarantee that Rivian, as sponsor, owes the project, and the DOE must see evidence those obligations are met before advancing funds.

The loan itself has already changed shape once. Rivian and the DOE restructured it in April from the original up to $6.6 billion to up to $4.5 billion, or $4,006 million of principal plus $494 million of capitalized interest, after the company raised the Georgia plant’s initial capacity by 50% to 300,000 vehicles a year. Production there begins in late 2028. Monday’s share sale is Rivian putting its side of the money on the table first.

The Timing Follows The Playbook Scaringe Described In February

Rivian launched this sale four days after reporting 12,194 second quarter deliveries, a beat of its own outlook that came with raised full year guidance of 65,000 to 70,000 vehicles, and hours after the stock closed at $20.14, up 8.11% on the day.

Scaringe told analysts on the fourth quarter call that Rivian would be “opportunistic with regards to raising additional capital.” As EVXL wrote in February, that was a polite way of saying more dilution was coming, and we expected at least one more raise before positive cash flow. Selling into a rally, with a delivery beat still fresh and the R2 ramping even after lease prices pushed some early buyers away, is what that word means in practice.

EVXL’s Take

I rarely get to point at a five-month-old call this cleanly. In February I wrote that Scaringe’s capital raise language meant more dilution was coming, and that at least one more raise would land before Rivian reached positive cash flow. Here it is. Credit where it’s due: the execution is textbook. You don’t sell 75 million shares when you’re desperate. You sell them the week your stock jumps on a delivery beat, with a healthy Q2 preview stapled to the prospectus as a sweetener.

The part shareholders should sit with is the you-first clause. The DOE loan requires Rivian’s own equity in the Georgia project before the first federal dollar moves, and that equity comes from whoever buys these 75 million shares. It’s a fair trade for low-cost financing, but call it what it is: taxpayers lend after shareholders pay. Watch the July 30 earnings call. If management hints this raise isn’t the last one before Georgia opens, believe them. The last hint aged perfectly.

Sources: StreetInsider, U.S. Securities and Exchange Commission.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.