Elon Musk unveiled Tesla and SpaceX‘s “Terafab” chip manufacturing project on the night of March 21, 2026, at a theatrical event inside Austin’s defunct Seaholm Power Plant — light beams shooting into the sky, Texas Governor Greg Abbott in the audience — and walked away without giving a construction timeline or confirming how the estimated $20–25 billion cost fits into Tesla’s already-strained capital plan. According to Reuters, the facility is planned for the North Campus of Giga Texas and will be jointly operated by Tesla, SpaceX, and xAI, which SpaceX acquired in an all-stock deal in February 2026. Musk called it “the most epic chip building exercise in history by far.” That’s a high bar, and he has a history of not clearing it.

Terafab Targets Two Chip Designs With Radically Different Missions

Terafab is designed to produce two distinct chip families under one roof, covering every stage of semiconductor production from design through packaging and testing — a vertical integration model no chip company has attempted at this scale. The first chip type is an inference processor for Tesla vehicles and Optimus humanoid robots, building on the current AI4 architecture. The second is a high-power chip designed specifically for operation in space, built to handle the harsher thermal environment of orbit. Musk said that chip “needs to operate at higher temperatures” and will feed SpaceX’s planned constellation of orbital AI satellites.

On Earth, Terafab is targeting 100 to 200 gigawatts of annual computing capacity. The larger ambition is a full terawatt per year directed toward space-based AI workloads. For context, Musk said at the event that current global AI compute output runs at roughly 20 gigawatts annually — and that all of Earth’s current chip output covers only about 2% of what his companies will eventually need. Terafab’s stated target of one terawatt is 50 times today’s total global AI compute figure. The facility is aiming for 2-nanometer process technology, the most advanced node now entering early commercial production. TSMC spent $165 billion and years of construction to build six Arizona fabs, which won’t reach 2nm output until 2029.

Musk’s Chip Supply Warning Has Been Building for Over a Year

The argument behind Terafab is not new. Musk has been signaling a semiconductor supply ceiling since Tesla’s Q4 2025 earnings call, warning investors that combined output from TSMC, Samsung, and Micron would hit its limit within three to four years. At the Seaholm event, he acknowledged all three suppliers by name before adding: “There’s a maximum rate at which they’re comfortable expanding. That rate is much less than we’d like.” His phrasing Saturday — “We either build the Terafab or we don’t have the chips” — was the same logic he used in December 2025 when claiming FSD unsupervised was nearly solved and that Tesla needed hundreds of gigawatts of AI chips annually to scale autonomy.

Tesla already has skin in the semiconductor supply game. The company signed a $16.5 billion deal with Samsung in July 2025 to produce AI6 chips at Samsung’s Taylor, Texas fab through December 2033, and the AI5 chip is currently manufactured by TSMC, initially in Taiwan before transitioning to TSMC’s Arizona facility. Terafab, if built, would sit alongside those agreements as a fully owned third leg of the supply chain. Whether it complements or eventually replaces those contracts is not something Musk addressed Saturday night.

Tesla’s In-House Silicon Track Record Includes One High-Profile Failure

The Terafab announcement lands in uncomfortable proximity to Tesla’s Dojo supercomputer collapse. In August 2025, Tesla disbanded its entire Dojo team. Musk called the project “an evolutionary dead end” and redirected the company’s AI compute budget toward a new Cortex supercomputer built on third-party Nvidia chips. That abrupt shutdown created a political mess in New York, where state officials had negotiated a lease extension partly on the strength of a $500 million Dojo investment commitment — and learned about the cancellation from Musk’s X post, not from Tesla. EVXL covered the fallout in detail at the time.

Dojo was later revived in January 2026, rebranded as an orbital AI compute concept tied to SpaceX’s satellite infrastructure plans. That pivot raised questions about whether Tesla’s chip roadmap follows engineering necessity or the shifting priorities of Musk’s portfolio. Terafab now represents a third major in-house silicon initiative in less than two years. The pattern is worth noting even before the first piece of construction equipment arrives on the Giga Texas North Campus.

EVXL’s Take

I’ve watched Tesla announce grand manufacturing revolutions before. Battery Day 2020 is the one that keeps coming back. Musk stood on a stage, promised 4680 cells would cut costs by 50%, and said Tesla would hit 3 TWh of annual production by 2030. Five and a half years later, the 4680 program has been a grind, and the dry electrode process required six or seven revisions before it worked at any scale. Chip fabrication at 2nm is harder than battery manufacturing — by an order of magnitude. TSMC doesn’t get to its process nodes by announcing them at a rock concert.

The devil’s advocate case is real, though. Musk does have a record of eventually delivering things that sound impossible when announced. SpaceX’s reusable rockets were dismissed the same way. And the supply math he’s describing — that his companies’ combined chip demand will outpace what any external foundry is willing to build — is not obviously wrong. If Tesla, SpaceX, and xAI’s AI compute needs compound at the rates Musk claims, owning a fab starts to make strategic sense regardless of how difficult it is to build.

But here’s the problem: Tesla’s auto business has been in a two-year sales decline. The Terafab cost, estimated at $20–25 billion, is not yet included in Tesla’s 2026 capex plan, which already exceeds $20 billion. Tesla’s CFO confirmed that Saturday. This announcement is being made at exactly the moment Tesla needs its capital focused on cars, not semiconductor fabs. Terafab feels less like a strategic necessity and more like a SpaceX pre-IPO story dressed in Tesla branding. My read: construction breaks ground in Austin by end of 2026. A functioning 2nm fab producing chips in volume? That’s a 2031 problem at the earliest — if it happens at all.

Frequently Asked Questions

What is Terafab?

Terafab is a joint chip manufacturing facility announced by Tesla, SpaceX, and xAI in Austin, Texas. Elon Musk unveiled the project on March 21, 2026, describing it as the largest semiconductor fab ever proposed. It is planned for the North Campus of Giga Texas.

What chips will Terafab produce?

Two chip types: an inference chip for Tesla vehicles and Optimus humanoid robots, and a high-power space-grade chip for SpaceX’s orbital AI satellites. Musk said the space chip must operate at higher temperatures than chips designed for terrestrial use.

How much will Terafab cost?

Estimates place the full project cost at $20–25 billion. Tesla’s CFO confirmed that figure is not yet incorporated into Tesla’s 2026 capital expenditure plan, which already exceeds $20 billion.

When will Terafab be operational?

Musk did not provide a construction or production timeline at the March 21 event. Multiple outlets, including Bloomberg and TechCrunch, noted that Musk has a history of overpromising on project schedules.

What happened to Tesla’s Dojo supercomputer project?

Tesla shut down the Dojo team in August 2025, with Musk calling it “an evolutionary dead end.” The project was partially revived in January 2026 as an orbital AI compute concept tied to SpaceX’s satellite plans. Terafab is the latest iteration of Tesla’s in-house silicon strategy.

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