April 23 - Tesla chief Elon Musk said on Wednesday that the company intends to use Intel’s next-generation 14A process to produce chips at the Terafab project, an advanced artificial-intelligence chip complex he is planning in Texas.
Terafab is conceived as a multi-factory campus that will serve the semiconductor needs of several Musk-led companies. The plan calls for two advanced chip plants on a large Austin-area site - one factory to supply chips for Tesla vehicles and Optimus humanoid robots, and a second tailored to AI data center workloads intended for space applications.
“We either build the Terafab or we don’t have the chips,” Musk said during a presentation in Austin in March, noting that existing global chip production would cover only a fraction of future demand from his companies. He acknowledged gratitude to current suppliers, naming Samsung, TSMC and Micron, but said demand from his ventures would ultimately eclipse worldwide output. Musk did not provide a firm timeline for when Terafab would be completed.
Location, early scale and funding
Musk said the initial research fabrication facility will be established on Tesla’s Giga Texas campus in the Austin area. He described the near-term build as a research fab intended to try out ideas rather than deliver full production volumes. That element of the program is expected to cost about $3 billion and be capable of producing "maybe a few thousand wafers per month," he said.
According to the plan Musk outlined, Tesla will handle the research fab while SpaceX would take an early role in the larger-scale Terafab build-out. Beyond those assignments, Musk said the project team needs to determine how to proceed with the remainder of the deployment. He reiterated an earlier claim that, at full scale, Terafab would generate roughly one terawatt of computing capacity a year - compared with around half a terawatt generated across the United States today.
Independent estimates cited in prior reporting place the capital required to produce one terawatt of annual compute between $5 trillion and $13 trillion, though Musk did not provide a capital estimate for the full-scale Terafab in his comments on Wednesday.
Technology partner and supply-chain outreach
Musk confirmed Tesla intends to use Intel’s 14A manufacturing node for chips produced at Terafab. Intel’s involvement would represent a major customer win for its 14A technology and its contract manufacturing ambitions. Musk said he expects Intel’s 14A process to be "probably fairly mature or ready for prime time" by the time Terafab reaches scale and that using Intel’s process "seems like the right move."
Intel announced earlier in the month that it would join the Terafab project, bringing established manufacturing expertise. Musk’s staff have contacted a range of chip-equipment and materials suppliers as part of early planning. Firms approached for price quotes and delivery timelines included makers of photomasks, substrates, etchers, deposition equipment, cleaning systems and test machinery, according to reporting on supplier outreach.
Separately, Reuters reported that SpaceX is planning to design and build its own GPUs for AI model training as part of the broader effort.
The key unknowns
While Musk has specified target end uses - automotive chips, processors for humanoid robots and chips for space-based data centers - several crucial details remain unspecified:
- Who will finance the expensive chipmaking equipment required for full-scale production;
- Who will operate the large-scale fabs once built;
- When the facilities beyond the initial research fab will come online.
Musk’s history of announcing ambitious projects that later encounter delays or are scaled back is a contextual reality he acknowledged indirectly by providing no firm timeline for Terafab’s broader deployment.
Implications for semiconductor and AI ecosystems
If Terafab advances beyond the research phase, it would represent a significant new demand source for semiconductor capital equipment and materials, and would reshape capacity needs for chips that serve electric vehicles, robotics and AI training. For now, however, the program remains at an exploratory stage with concrete commercial terms and operational responsibilities yet to be defined.