Tesla and SpaceX Commit $16.8 Billion to 'Terafab,' a Massive Texas AI Chip Factory

Tesla and SpaceX Commit $16.8 Billion to 'Terafab,' a Massive Texas AI Chip Factory

After months of speculation and shifting numbers, Elon Musk's two flagship companies have put a place on the map. Tesla and SpaceX confirmed Thursday that "Terafab," the advanced semiconductor complex they are building together, will rise in Grimes County, Texas, about an hour northwest of Houston, backed by an initial investment of $16.8 billion. Musk called it "the largest and most valuable building on Earth by far," and the plans behind that boast are staggering: a single site spanning more than 100 million square feet, built to manufacture, package, and test the chips that will power Tesla's robots and cars and SpaceX's ambitions in orbit.

The pitch is vertical integration at a scale no chipmaker has attempted. Rather than lean on outside foundries, Terafab is designed to bring logic, memory, packaging, and testing under one roof, shortening the path from chip design to deployment. The companies say at least 3,000 people will work there, most of them local hires from Grimes and neighboring Brazos County.

What the companies are promising

In a post on its website, SpaceX framed the project in sweeping terms. "The facility will be an advanced semiconductor fab that will bridge the divide between current global chip supply and the compute demand of the future," the company wrote, adding that Terafab "will be epic in both its mission and in its sheer size, designed to build new compute at an unprecedented scale and speed."

The chips it produces are meant to feed a very specific vision. SpaceX says the fab will turn out processors "optimized for edge computing and inference for use in hardware like Tesla's Optimus robots and self-driving Cybercabs, along with high-power chips designed for operating SpaceX's space-based data centers." Tesla echoed the announcement on X, pointing to a goal of "producing over 1 terawatt of compute per year" and warning that "both Tesla and SpaceX will need far more chips than current and future global production can supply."

That terawatt figure is the animating logic behind the whole bet. Musk has argued that a future of millions of humanoid robots, driverless robotaxis, and satellite-based AI data centers requires computing power far beyond what today's production rates support. Terafab is his answer to a supply problem he insists the rest of the industry cannot solve.

The numbers keep moving

The $16.8 billion headline comes with fine print. When Tesla and SpaceX first floated Terafab in March, the figure attached to it was $25 billion. Now the same project is described as a $16.8 billion "initial phase," with future expansion phases pushing total spending much higher. SpaceX has suggested in filings that it may eventually spend as much as $119 billion across a multi-phase construction plan, though the company notably did not mention Terafab on its first earnings call earlier this week.

There is also a gap between the confident announcement and the paperwork. In May, SpaceX's IPO-related S-1 filing described Terafab as only a "general framework" with no binding commitments and no finalized split of intellectual property between the two companies. Intel has said it will contribute to the project but has been vague about exactly how. And Tesla's near-term chip supply is already locked up elsewhere: Samsung's Taylor, Texas fab is in production on Tesla's AI5 chip, and Tesla signed a $16.5 billion deal with Samsung last year for its next-generation AI6 processor, with TSMC covering the rest. That makes Terafab a long-horizon bet rather than a fix for anything Tesla needs today.

The scale claims are hard to square with existing facilities. At more than 100 million square feet, Terafab would be over ten times the floor space of TSMC's largest gigafabs. TSMC is spending roughly $165 billion to build six fabs in Arizona, and those remain a fraction of Terafab's stated footprint.

Local stakes and community pushback

The site carries history and friction. Terafab will sit on the Gibbons Creek Reservoir, which once cooled a coal-fired power plant that shut down in 2018. SpaceX says it is "committing to the use of" reservoir water rather than local groundwater, a nod to a community already uneasy about the water and power demands of the data centers spreading across Texas. A heavily attended county meeting on Wednesday drew hundreds of residents who raised concerns about the millions in tax breaks awarded to the project and a perceived lack of transparency.

Not everyone is wary. "We believe this agreement will strengthen our district, expand opportunities, and better prepare our students for their future," said Dr. Sarah Borowicz, superintendent of the Anderson-Shiro Consolidated Independent School District, in a statement released by Governor Greg Abbott's office. "There are defining moments in the life of a school district, and this is one of those moments."

Why It Matters

Terafab is the clearest sign yet that the AI boom is pulling its heaviest players into semiconductor manufacturing itself. For years, the hardware supply chain ran on a division of labor: chip designers like Nvidia and Tesla drew the blueprints, and foundries like TSMC and Samsung did the fabrication. Owning a fab is brutally expensive, slow, and technically unforgiving, which is why almost no one attempts it from scratch. Musk's move to build memory and logic in-house is a wager that controlling the entire stack, from silicon to robot, is worth tens of billions and years of risk.

The appeal of vertical integration is speed and independence. If Tesla and SpaceX genuinely need a terawatt of compute a year, no existing supplier can guarantee that capacity, and every chip bought on the open market is a chip a competitor could bid away. A captive fab, in theory, lets the companies iterate hardware and software together and insulate themselves from the geopolitics of chip supply. The counterargument is just as sharp: fabs are the hardest thing in tech to build, neither company has done it, and the same track record includes AI chips and 4680 battery cells that fell short of their billing.

What to watch next

The real test is execution. Watch for whether that $16.8 billion first phase converts into a poured foundation and a binding IP agreement between Tesla and SpaceX, and how much of the cost ultimately lands on Tesla's balance sheet at a moment when its car business has stopped growing. Intel's role, the water and power permitting fights in Grimes County, and any firm timeline for first production will reveal whether Terafab is a genuine industrial project or, as its own filings once hinted, still a framework in search of a commitment.