Elon Musk said on October 20 that ASML and other chip equipment suppliers are speeding up deliveries to Terafab, the Tesla and SpaceX chip factory planned for Gibbons Creek, Texas.
He made the claim on SpaceX's Q3 2026 earnings call, which was broadcast live on X. He is targeting volume production by the third or fourth quarter of 2028, and he called that timeline highly aspirational with "big error bars." SpaceX plans to make a first experimental test chip at its research fab in Austin around Q2 2027. Intel joined the venture in April 2026, and Terafab will use EUV (extreme ultraviolet) lithography tools on Intel's 14A process. Applied Materials, Lam Research, KLA and Tokyo Electron are named suppliers, and Tesla and SpaceX are in talks with about 10 South Korean equipment makers.
The timing is the open question. ASML is the only commercial supplier of EUV machines, which cost $400 million each. Unofficial sources put Terafab's first tool at mid-2027. A mid-2027 first tool would leave about 12 to 18 months before the 2028 volume target. It is unclear whether Musk's accelerated schedule improves on that date. The expedite claim also comes from Musk, speaking about suppliers he thanked on the call.
The compute goals are just as ambitious. Musk expects 2.5 to 3 GW of AI computing capacity active by the end of 2026 and is highly confident about adding 10 GW in 2027. He said he would be surprised if they miss 13 GW by the end of 2027. A 440,000-GPU Nvidia Vera Rubin cluster is planned for the end of Q2 2027. By comparison, the restructured 10 GW Project Stargate and Meta's 5 GW Hyperion are both expected to take until the 2030s.
Terafab's stated goal is 1 TW of AI computing power per year. Gartner puts total global AI-optimized data center capacity at roughly 44 GW as of late 2026, so 1 TW is more than 22 times that. Global AI compute has grown about 3.3x to 3.4x a year since 2022, according to EpochAI. That is fast, but Terafab asks for a much bigger jump.