During Tesla's most recent earnings call, Elon Musk disclosed that TeraFab — his nascent semiconductor venture — intends to build its AI chips on Intel's 14A process node, in what would almost certainly require a technology licensing arrangement with the struggling American chipmaker. Musk split responsibilities between two of his companies: Tesla will construct and operate a pilot research fabrication line at its Gigafactory Texas campus, while SpaceX will own the high-volume manufacturing side when TeraFab eventually scales.
A bet on a node that isn't finished yet
Musk was unusually candid about the speculative nature of the plan. "We plan to use Intel's 14A process, which is state-of-the-art and in fact not yet totally complete," he told analysts and investors on the call. His reasoning is that by the time TeraFab reaches meaningful production capacity — which he placed later in this decade — Intel's 14A should have matured into a reliable, high-yield process. "By the time TeraFab scales up, 14A will be probably fairly mature or ready for prime time," he added, citing what he described as a strong relationship with Intel's current leadership team.
The 14A node sits at the leading edge of Intel's process roadmap, representing the company's 1.4-nanometer-class generation. Crucially, Musk stopped short of using the word "licensing" — meaning the precise commercial structure of the Intel-TeraFab relationship remains undefined publicly. Whether Intel licenses the process recipes outright, provides process development support, or takes an equity stake in TeraFab has not been confirmed by either party. What Musk did confirm is the intent to run the node at a TeraFab-operated facility rather than simply outsourcing production to Intel's own foundry.
Texas pilot line, then SpaceX scale-up
The nearer-term piece is a $3 billion semiconductor research facility at Gigafactory Texas. Musk described it as a pilot line capable of processing a few thousand wafers per month — modest by commercial fab standards, but adequate for what he framed as an experimental mandate. "It is really intended to try out ideas for improving the fundamental technology of how chips are made, some of new physics we would like to test out," he said. Notably, Intel's 14A process is not expected to feature prominently at this early Texas facility; the pilot line is about process experimentation, not production.
SpaceX takes the handoff from there. Once the Texas pilot validates a workable approach, SpaceX is expected to build and operate the actual high-volume manufacturing plant. The arrangement is operationally tidy on paper, but in practice it requires both companies' boards to approve any joint initiative and both to clear conflict-of-interest reviews — a governance layer that Musk himself acknowledged will slow progress.
What licensing Intel's process actually means in practice
Licensing a process node is not without precedent, though the track record is mixed. GlobalFoundries licensed Samsung's 14nm-class technology in 2014 and 2015 after failing to complete its own 14nm node. The first clearly attributable high-volume products built on that licensed process — AMD's Radeon RX 400-series Polaris GPUs — didn't arrive until 2016, and yields remained a persistent concern during the ramp. Rapidus, the Japanese consortium, provides a more recent parallel: it licensed IBM's 2nm technology to anchor its domestic fab ambitions, a project still deep in development.
The comparison matters because modern process technologies are categorically more complex than those from a decade ago. Porting Intel's 14A — with its high-numerical-aperture EUV requirements and Intel-specific tooling — to an entirely new fab operated by an organization with no prior semiconductor manufacturing experience would rank among the most technically demanding industrial undertakings attempted outside of an established chipmaker. Musk's framing of 14A as "not yet totally complete" only sharpens that challenge: TeraFab would be licensing a process still under active development, then attempting to replicate it at scale in a facility that doesn't yet exist.
For Intel, the calculus looks different. The company has spent several years repositioning itself as a contract manufacturer under its Intel Foundry Services banner, with mixed commercial results so far. A licensing deal with TeraFab — backed by the capital of Musk's broader enterprise — would represent an unconventional but potentially lucrative revenue stream that doesn't require Intel to compete directly for wafer volume. Whether Intel's board views that as a lifeline or a distraction from its core foundry rebuild is a question that will become harder to avoid as TeraFab moves from earnings-call ambition toward concrete ground-breaking.
