Elon Musk reportedly tried buying chip factories from established manufacturers before deciding to build Terafab, adding semiconductors to his expanding portfolio of future deliveries.
According to The Elec, Musk considered acquiring facilities from Intel, TSMC and Micron, but switched towards partnerships after being turned down. It seems even the world’s richest man cannot simply put a leading semiconductor factory in his shopping basket and select next-day delivery.
Terafab is now talking to about 10 Korean suppliers, discussing equipment, chemicals, and piping. Lam Research and Applied Materials reportedly have dedicated teams working on the project. Suppliers describe unusually quick responses and engineers with authority to make decisions.
“I’ve never seen this level of execution speed elsewhere in the semiconductor industry,” one equipment supplier told The Elec.
Industry observers expect an initial facility capable of processing 100,000 wafers monthly before 2030. Musk’s eventual ambition is a million wafers a month.
Those are substantial manufacturing targets. Read them alongside his established habit of treating the calendar as an opening negotiating position. Take the second-generation Tesla Roadster. Revealed in November 2017, it attracted reservations with production promised for 2020. Nearly nine years after that unveiling, customers are still waiting. Following another demonstration this month, production is now scheduled for 2028. By then, the original announcement will be old enough to start secondary school.
The car has gained thrusters, and its delivery date has been pushed back another two years.
In April 2024, Musk said Tesla’s humanoid robot could be ready for sale by the end of 2025. Later that year, he touted an eventual price of $20,000–$30,000 for a machine capable of everyday tasks. By January 2026, the language had shifted to production, “hopefully” starting toward the end of the year, with an initially “agonisingly slow” ramp. The affordable domestic robot has proved considerably better at moving deadlines than doing the washing-up.
For Terafab, Musk now targets a working SRAM test chip around May or June 2027 and volume production in the third quarter of 2028. He has at least supplied the appropriate qualification.
“I put big error bars on this because we don’t know if we will actually achieve this or not,” Musk said during SpaceX’s earnings call.
Asked whether the venture had ordered ASML’s essential EUV lithography machines, he did not answer directly. That is a fairly important shopping-list item to leave hanging while discussing a manufacturing revolution.
The ambition extends well beyond chips. Speaking to Walter Isaacson, Musk suggested digital superintelligence could increase Earth’s economic output by roughly 20–30 per cent. Add robots, space access and a lunar mass driver, however, and the forecast becomes suitably astronomical.
With those capabilities, Musk said, “you can get to, I think potentially, a million times of the size of the Earth economy”.
