Elon Musk is planning a chip factory while admitting he does not yet know how to make chips. His proposed approach for Terafab is to begin with standard equipment, then change course as bottlenecks become clear, a factory plan with a learning curve built in.
Terafab would start with standard equipment
The first step would be to use existing semiconductor manufacturing equipment, Musk said during a conversation with Tesla Owners Silicon Valley. Terafab would then modify those tools to raise throughput and, where production bottlenecks appear, move to specialized equipment designed to process wafers faster.
Musk said he still has much to learn about chip manufacturing and does not know how to make chips himself. That admission puts the project’s ambition in context. Terafab is a proposal, not a detailed account of an operating factory or a settled manufacturing process.
The longer-term plan is to bring several stages of production under one roof: making photomasks, producing chips and packaging finished parts. Musk believes putting those steps together could help the company improve its processes faster. He also acknowledged that putting the project into practice could take several years.
He said equipment used by TSMC, Samsung and Intel appears similar in many cases. That observation has strengthened his interest in developing equipment of his own, though he did not give a timeline for specialized tools.
The sequence is incremental: start with standard machinery, find the constraints, then decide where custom equipment might help. It is a practical-sounding outline, but the difficult part is in the details, and those remain to be worked out.
Musk describes several AI-related episodes
Musk also spoke about generative AI, including music videos made with several AI systems working together. He said their quality and emotional effect had impressed him.
He described having what he called “AI psychosis,” or “cyberpsychosis,” several times during 2026, estimating that it had happened around six times. He used the phrase figuratively to describe his own experiences, rather than presenting it as a formal diagnosis.
His concern about data was more concrete. Musk said knowingly false information and conflicting basic principles in training data could lead to unpredictable results. The warning sits alongside his enthusiasm for AI, which he described as approaching an extraordinary stage.
Terafab remains a plan, and Musk says he has plenty to learn before making chips. For now, the factory’s clearest starting point is the admission that the work has not yet been figured out.