Volume production at Terafab could begin in the third or fourth quarter of 2028, Elon Musk told analysts on SpaceX’s earnings call this week. He called it “an unbelievably fast schedule for semiconductor production” and then conceded the obvious: “We don’t know if we can actually achieve it.”
The plan runs in stages. Engineers expect a first test run built on static random-access memory circuits at a research site in Austin around May or June 2027, the standard way to prove out a new process flow. The volume line at the main Texas plant follows about a year later.
What changed is supplier behaviour. “I thank ASML and other suppliers for moving up equipment supply so we can achieve it,” Musk said. Asked directly whether extreme ultraviolet lithography had been ordered, he did not answer. ASML has already told investors its own output plans will not cover all demand in 2027 and 2028, so any accelerated delivery to a newcomer comes at someone else’s expense.
The venture is meant to hold logic, memory, packaging and test under one roof and to eventually produce about a terawatt of computing capacity a year. Intel joined in April without disclosing what it contributes. Musk has said he will run the site himself rather than hand it to TSMC.
The nearest beneficiaries may be smaller Korean toolmakers. HPSP is said to be positioning high-pressure hydrogen annealing equipment for test work, and Hanmi Semiconductor is chasing orders. The sites sit far from the Texas ground being graded, and the fabs whose wafers fuel today’s AI buildout will still be running at full tilt before Terafab’s does.
