SpaceX and Tesla Terafab concept. Credit: SpaceX / Tesla
Elon Musk told the All-In Summit that SpaceX and Tesla’s Terafab bet is not optional: the companies either build the in-house chip plant or fail to scale. Speaking with SpaceX President Gwynne Shotwell during the Los Angeles session, he said the Terafab idea “sort of did come to me in a dream,” then gave the hard rationale around chip supply and long-term compute demand.
The All-In Summit is the annual conference run by the hosts of the All-In podcast (Chamath, Jason Calacanis, David Sacks, and Friedberg); this year’s event ran September 13-15, 2026 in Los Angeles.
“It’s either build TeraFab or fail to scale,” Musk said. “Those are the two options.” He tied the project to concern that chips from Taiwan might not stay available for unspecified reasons, and to a broader capacity gap: existing fabs are maxed out while AI scaling needs logic, memory, and packaging for server centers plus edge compute in humanoid robots, cars, and Mars systems. Shotwell also noted SpaceX is already doing chip packaging, a bottleneck hosts flagged as scarce worldwide.
That framing lands on top of the public buildout already underway. SpaceX and Tesla’s $16.8 billion Terafab groundbreaking in Grimes County set the Texas plant story, and SpaceX later sued over release of Terafab tax-application records after Texas open-records pressure. Musk described a crawl-walk-run path that starts with an Austin R&D fab collaboration on the Giga Texas campus, aiming to make something useful by the end of 2027 before true scale.
Readers tracking the plant should separate the All-In strategy talk from the Texas permitting and records fight. The Summit line is Musk’s clearest recent statement that Terafab is insurance against overseas chip risk and a prerequisite for the compute loads Tesla and SpaceX say they need.
