Elon Musk has picked Intel’s 14A node for his TeraFab daydream.
Speaking to analysts at Tesla’s earnings call, the plan is to lean on 14A once TeraFab has real capacity and the process is no longer half-baked.
Musk said: “We plan to use Intel’s 14A process, which is state-of-the-art and in fact not yet totally complete… it probably will be fairly mature or ready for prime time by the time TeraFab ramps.”
He piled on the compliments, too, saying, “14A seems like the right move, and we have a great relationship with Intel, a lot of respect for the CEO, the CTO, and the new team there.”
The chatter points to TeraFab taking Chipzilla’s process tech and dropping it into a fab that runs itself, which smells a lot like licensing without anyone saying the L-word out loud. Musk has not mentioned licensing per se, so Chipzilla’s role still looks like a big blank space waiting to be negotiated.
Musk sketched a split job sheet between his two flagship outfits, with Tesla running the pilot line and SpaceX taking the brutal, high-volume slog. That tidy division on a slide does not magically make it painless in the real world.
In the near term, Tesla wants a semiconductor R&D facility at its Texas campus, priced at around $3 billion. The site is meant to be a small pilot line producing a few thousand wafers per month while it explores manufacturing ideas and tests whether they withstand the rigours of production reality.
Tesla chief executive Elon Musk: “This is something we expect to be a $3 billion initiative and support a few thousand wafers per month. It is intended to explore ideas for improving the fundamental technology of chip manufacturing, as well as some of the new physics we would like to test. We also want to test whether something is working in the production [environment]. You need kind of like a few thousand wafer starts a month to make sure that the production process is sound.”
From the sound of it, 14A is not the star of the Texas pilot line anyway, at least not early on. The point is experimentation, validation and seeing what breaks before anyone bets a mass-production facility on it.
For the bigger leap, SpaceX is expected to build the proper high-volume manufacturing site. Coordinating between Tesla and SpaceX is the sort of corporate tangle that invites board scrutiny and conflict-of-interest reviews, which tends to slow everything to a crawl.
“That is basically what we have figured out thus far, is Tesla’s doing the research fab, SpaceX doing the initial part of the large-scale TeraFab, and then we’ve got to figure out the rest,” Musk said.
If TeraFab does end up licensing a process node, it gets a shortcut compared with inventing a leading-edge manufacturing stack from scratch, which can eat a decade. For Chipzilla, it could mean fresh cash and a partner willing to spend silly money on tools and facilities.
The industry has done this dance before, when schedules slip, or ambitions outpace engineering reality. GlobalFoundries licensed Samsung’s 14nm-class process tech in 2014 and 2015 after its own 14nm XM node stalled, and the first clearly attributable, high-volume GF-built 14nm products showed up in 2016 with AMD’s Radeon RX 400-series Polaris GPUs.
