SpaceX’s latest moonshot isn’t a rocket; it’s a semiconductor empire in the making — and if Elon Musk gets his way, it could become one of the most expensive industrial bets in history.
For more than a year, Musk has been warning that the true choke point for AI and robotics isn’t clever algorithms, it’s chips — and the power and factories needed to make them. Existing semiconductor manufacturers, he’s argued, simply aren’t moving fast enough to feed his growing constellation of companies.
Out of that frustration came a simple, blunt ultimatum: "We either build the Terafab or we don’t have the chips, and we need the chips, so we build the Terafab".
That line, buried in county filings and later reported in the tech press, is the philosophical seed of what’s now known as Terafab — a proposed mega-fab that aims not just to supply SpaceX, Tesla, and xAI, but to redefine how much compute one corporate ecosystem can own.
The first concrete glimpse of Terafab arrived via a decidedly unglamorous route: a public proposal posted on the Grimes County, Texas website. SpaceX, identified as the space company that also houses Musk’s AI venture xAI, laid out plans for a semiconductor manufacturing and advanced computing facility in the county.
The numbers were brazen even by Musk standards. SpaceX said it was considering spending $55 billion, at least initially, to build a semiconductor factory in Grimes County, with total spending that "may" eventually reach $119 billion. The filing described the project as a “multi-phase, next-generation, vertically integrated semiconductor manufacturing and advanced computing fabrication facility.”
This wasn’t just another data center or chip line; “multi-phase” and “vertically integrated” signaled an intent to own as much of the stack as possible — from wafer to finished compute modules.
Musk had already sketched the outlines: the project, dubbed Terafab, would lean on Tesla’s resources and rope in a heavyweight partner from the old guard of chips — Intel. Together, they plan to develop chips for AI servers, satellites, a proposed space-based data center, autonomous Tesla vehicles, and robots.
The broader strategic aim was clear: Muskworld doesn’t want to rent AI compute from the cloud oligopoly; it wants to manufacture the compute itself.
Within a day, more details spilled out through additional reporting. A public hearing notice — this time tied to a request for tax breaks for the project — shed light on scope and geography.
SpaceX plans to invest at least $55 billion into its “Terafab” chip plant in Austin, Texas. According to that filing, if additional phases are built, the total investment could balloon to $119 billion.
So is Terafab in Austin or in rural Grimes County northwest of Houston? The answer, for now, seems to be “both” — or at least, “it’s complicated.” One filing anchors the project to Grimes County; another pegs a massive Terafab plant to Austin and explicitly notes the request for local tax incentives.
Musk himself added another twist. In a post on X, he said Grimes County was only one of several locations under consideration for the factory, suggesting a now-familiar playbook: float a mega-project, let local governments compete, and negotiate for incentives.
What’s not ambiguous is the ambition. When Musk first announced Terafab back in March, he touted plans for the factory to produce enough chips to support up to 200 gigawatts per year of computing power on Earth, and up to one terawatt in space. That’s a scale that doesn’t just compete with today’s hyperscale data centers; it dwarfs them.
One of the more surprising beats in this timeline came in April, when Intel publicly embraced a supporting role. The legacy chip titan announced it would help design and build Terafab, casting the partnership as proof of its own relevance in the AI era.
Intel put it bluntly: “Our ability to design, fabricate, and package ultra-high-performance chips at scale will help accelerate Terafab’s aim to produce 1 TW/year of compute to power future advances in AI and robotics.”
For Intel, this is a chance to hitch its manufacturing comeback to the most headline-grabbing industrialist on the planet. For Musk, it’s a shortcut to world-class process technology and packaging at a time when building a bleeding-edge fab from scratch typically takes a decade and unfathomable capital.
Behind all the filings and partnerships sits a brute-force calculation: Musk’s companies are compute-hungry in a way that’s hard to overstate.
Semiconductor manufacturers, Musk argues, “aren’t making chips quickly enough for his companies’ artificial intelligence and robotics needs.” Hence the Terafab threat: if the market won’t supply chips at the required scale and pace, he’ll build the supply himself.
The stakes are already massive. The combined entity of SpaceX and xAI is said to have a valuation of $1.25 trillion and is expected to go public in June — a listing that would make Terafab not just an industrial bet, but a defining narrative for public investors weighing Musk’s long-term vision.
While Terafab remains on paper, Musk’s ecosystem is already quietly building the downstream infrastructure that would drink from this future chip firehose.
SpaceX is expanding its data center footprint on Earth and currently operates a “Colossus” data center in Memphis, Tennessee. That facility recently signed an agreement to power Anthropic’s AI models on Earth, positioning SpaceX not just as a rocket company or satellite provider, but as an emerging member of the AI infrastructure club.
The Texas plant, once built, will be run by SpaceX and Tesla and make chips for both companies, with uses spanning AI, robotics, and space-based data centers. In other words, Terafab is designed as the upstream engine for a vertically integrated compute stack: from silicon, to data centers, to orbit.
From one vantage point, Terafab looks like classic Musk maximalism: a response to supply constraints by attempting to internalize an entire global industry. The language of the filings — “next-generation,” “vertically integrated,” “multi-phase” — reads like a blueprint for a private, Musk-controlled alternative to TSMC and Nvidia’s current grip on the AI chip supply chain.
Supporters argue that someone has to try. With AI demand exploding, governments struggling to onshore advanced manufacturing, and chip lead times still tight, a $55–119 billion bet on new fabrication capacity in the U.S. could be exactly the kind of overreach that drags the rest of the sector forward.
Critics, however, see familiar red flags:
At the same time, Terafab slots neatly into a larger industrial-policy story. The U.S. has spent the last few years trying to claw back semiconductor manufacturing via the CHIPS Act and allied corporate subsidies. Musk’s plan — part public spectacle, part hard infrastructure — could either become the crown jewel of that push or its most spectacular cautionary tale.
For now, Terafab exists as filings, quotes, and a handful of aggressive targets:
Public hearings over tax breaks and local permitting will be the next tangible checkpoints. Then come land acquisition, construction timelines, environmental reviews, and the inevitable chorus of local boosters and skeptics.
Musk has framed the decision as binary — build Terafab, or stall his AI and robotics ambitions for lack of chips. The real story, unfolding now in Texas filings and corporate boardrooms, is whether the rest of the ecosystem — governments, partners, investors — is willing to bankroll a factory-sized answer to that challenge.
