SpaceX Picks Texas for Its $16.8B Terafab—Now Comes the Hard Part

Kind: captions Language: en SpaceX and Tesla have chosen Grimes County, Texas for Terrafab, a planned semiconductor complex with a $16.8 billion first phase and at least 3,000 jobs. That makes Terrafab more than a presentation. But it does not mean the giant production fab is already rising from the ground or that its defining 1 terowatt target is close. The project actually has two sites. In Austin, Tesla broke ground on the advanced technology chip fab, a smaller research and process development facility where site work is visible. Roughly 100 miles east, Grimes County is the planned mass production campus. State filings schedule phase 1 spending from 2026 through 2028, but a schedule is not construction evidence. The money headlines describe different layers of the plan. $16.8 8 billion for the newlyannounced first phase, about $119 billion for a possible four-phase buildout, and a much lower contractual floor. Under the executed county agreement, SpaceX must invest at least $5 billion by 2030 and create 1,800 full-time jobs by 2035. That floor is not the same as the public ambition. Why build it? SpaceX and Tesla want custom chips for robots, autonomous vehicles, satellites, and orbital computing and tighter control over design, logic, memory, packaging, and testing. But 1 terowatt per year is not the FAB's power bill. It is the claimed combined power rating of the compute hardware produced in a year, an unusual output metric. Without a chip mix, wafer starts, yields, packaging yields, or a target date, that number cannot be translated into a conventional auditable fab plan. Independent estimates show the scale problem. Depending on assumptions, reaching the headline target could require well over 1 one leading edge fabs and trillions of dollars. Those are scenarios, not forecasts. Intel could lower the starting technology risk, but the prospectus describes a framework, not a finished foundry contract. Specific milestones and spending still require negotiation, and partners may leave. Even a viable process needs scarce lithography tools, metrology, chemicals, water, packaging capacity, and experienced engineers. ASML says projects at this scale add pressure to equipment supply. Local trade-offs will matter, too. Water rights tied to the site allow up to 3.2 billion gallons of permanent annual consumption, but that ceiling is not a usage forecast. Watch the Austin Silicon, then the process transfer. That, not the renderings, will tell us whether Terapab works.