Elon Musk has sketched a three-part plan for TeraFab: Tesla would design chips, SpaceX would handle mass production, and Intel’s 14A process could become the manufacturing target once it is ready for volume use. The pitch is ambitious even by Musk standards, but it also reflects a simple reality in semiconductor work: building a brand-new advanced process from scratch takes time, money, and an unhealthy amount of patience.
Musk said the Intel 14A node looks like the ”right choice” for the TeraFab project and added that TeraFab and Intel already have a good relationship. He did not say whether this would mean direct licensing, so Intel’s exact role is still murky. That ambiguity is not ideal, but it is common enough in early-stage chip partnerships, where the headline often runs ahead of the legal paperwork.
Tesla’s $3 billion research fab in Texas
In the near term, Tesla’s first step is a research fab at Gigafactory Texas, which Musk said could cost about $3 billion to build. The site would be sized for several thousand silicon wafers a month, enough for experimentation, process tuning, and proof-of-concept work, but nowhere near the kind of output needed for real scale. Think of it less as a factory and more as a very expensive laboratory that smells faintly of future lawsuits.
That approach is familiar in chipmaking. New process technologies usually need a pilot phase before anyone trusts them with high-volume production, and companies from Intel to TSMC have long used smaller internal ramp-ups to shake out defects before going big. TeraFab appears to be trying to buy time by borrowing an existing advanced node instead of inventing one first.
Why Intel 14A is the shortcut
Using a licensed process is the fast lane to advanced manufacturing, especially for a newcomer that wants leading-edge capability without spending years building the recipe itself. Intel 14A is attractive for that reason alone: if it matures on schedule, TeraFab could jump straight to a modern node rather than waiting for a homegrown process to catch up. The catch, of course, is that ”ready for production” is doing a lot of work in that sentence.
- Tesla: chip design and early pilot production
- SpaceX: mass production and scale-up
- Intel 14A: possible process foundation for TeraFab
If the plan survives contact with reality, the upside is obvious: Musk gets more control over silicon for his companies, and Tesla avoids relying entirely on outside foundries for strategic chips. The harder question is whether a design-first, scale-later model can move quickly enough in a business where the competition is already spending billions to stay one node ahead.
The open question around Intel’s role
For now, the biggest missing piece is the commercial structure. A partnership, a license, a process transfer, or something else entirely would all imply very different levels of control, risk, and cost. The closer TeraFab gets to real hardware, the less vague that arrangement can stay.