The guy who said he'd colonize Mars is now trying to outspend Intel's entire market cap on a chip factory—and the bill just doubled Wall Street's estimate.
The Summary
- SpaceX filed a public notice revealing the Terafab project will cost $55 billion initially, potentially reaching $119 billion with additional phases—more than double Morgan Stanley's $45 billion estimate.
- The facility aims to produce AI chips for Tesla's robotaxis and Optimus robots, creating vertical integration across Musk's companies: SpaceX, xAI, and Tesla.
- Total investment could exceed what SpaceX plans to raise in its record-setting IPO, signaling capital requirements that dwarf typical semiconductor projects.
- Grimes County commissioners meet June 3 to vote on property tax abatements for the Gibbons Creek Reservoir site, 100 miles from Austin.
The Signal
Musk announced Terafab in March as a joint venture between his three major companies. Now we know what it actually costs to play chipmaker at hyperscale. The $55 billion starting price puts this in rarefied air. TSMC's Arizona fabs are projected at $40 billion total. Intel's Ohio project clocks in around $20 billion. Musk is proposing to spend more than both, combined, on phase one alone.
The public filing describes this as a "multi-phase, next-generation, vertically integrated semiconductor manufacturing and advanced computing fabrication facility." Translation: Musk wants to own the entire stack, from silicon to inference, for his AI ambitions. Tesla's Full Self-Driving and Optimus humanoid robot both need compute. So does xAI's Grok model. Buying from Nvidia means paying Nvidia margins and waiting in Nvidia's queue.
"SpaceX proposes construction that would represent a transformative investment in domestic semiconductor manufacturing capacity."
But here's where the math gets uncomfortable. Bloomberg notes the total investment could exceed what SpaceX aims to raise in its IPO, which would already be the largest on record. If the IPO targets $80-100 billion in fresh capital and Terafab eats $55-119 billion, the allocation problem becomes obvious. Either SpaceX is betting its entire public market debut on this chip play, or Musk plans to fund phases with revenue from Starlink and launch services, or he's counting on tax abatements and federal CHIPS Act money to close the gap.
The location choice is deliberate. Gibbons Creek Reservoir sits near College Station, home to Texas A&M. That's a talent pipeline for engineers. It's also rural enough that land is cheap and county commissioners are motivated. The June 3 vote on property tax abatements will test how badly Texas wants this. A $119 billion project would be the largest private construction effort in state history.
Key capital questions:
- Where does $55-119 billion actually come from if not the IPO?
- How much of this qualifies for CHIPS Act subsidies designed to reshore manufacturing?
- What happens to SpaceX's core rocket business if capital gets redirected to fabs?
Morgan Stanley's original $45 billion estimate now looks quaint. Either their model missed the scope, or the scope expanded after Musk saw what vertical integration actually requires. Modern leading-edge fabs cost $15-20 billion per facility. If Terafab includes multiple nodes, packaging, and research infrastructure, $119 billion starts to make sense. It also starts to look like the kind of bet that defines a company, or sinks it.
The timeline matters too. Semiconductor fabs take 4-6 years to build and ramp. Tesla's robotaxi ambitions and Optimus production timelines assume chip supply in 2-3 years. That gap means Terafab isn't solving Musk's near-term compute problem. It's a hedge against a future where AI agents need so much silicon that even Nvidia's capacity isn't enough.
The Implication
Watch the June 3 Grimes County vote. If commissioners approve abatements, construction likely starts this year. If they balk, Musk either renegotiates or moves the project, and the $55 billion number was just an opening bid to extract maximum concessions.
For the broader agent economy, this matters because it signals what the top players think compute will cost in 2030. Musk is essentially saying: the only way to build millions of humanoid robots and city-scale robotaxi fleets is to own the foundry. If he's right, other AI-first companies will follow. If he's wrong, SpaceX shareholders will spend the next decade paying for the most expensive strategic error in tech history.