On March 21, 2026, Elon Musk revealed Terafab, a groundbreaking semiconductor fabrication project jointly developed by Tesla, xAI, and SpaceX. The initiative aims to build a vertically integrated facility capable of producing more than one terawatt (one trillion watts) of AI compute capacity per year, addressing the surging demand for AI-optimized chips across enterprise and space applications (en.wikipedia.org).

The facility will consolidate the entire chip production lifecycle—design, lithography, fabrication, memory production, advanced packaging, and testing—under one roof. A prototype “Advanced Technology Fabrication” plant will be established at Tesla’s GigaTexas site in Austin, Texas, enabling rapid iteration: “make a chip, test it, revise the mask, and repeat without shipping wafers between sites,” a capability currently unmatched in the industry (en.wikipedia.org).

Terafab targets cutting-edge 2-nanometer process technology, with initial output projected at 100,000 wafers per month, scaling to 1 million wafer starts per month at full capacity. The long-term ambition is to produce between 100 and 200 billion custom AI and memory chips per year, serving both edge inference needs—such as Tesla vehicles and Optimus humanoid robots—and AI infrastructure for orbital systems (en.wikipedia.org).

Musk framed Terafab as essential to humanity’s future, stating, “We either build the Terafab, or we don’t have the chips, and we need the chips, so we build the Terafab.” He emphasized that existing global fabrication capacity—currently producing only about 2% of what Tesla and SpaceX will require—is insufficient to support their AI ambitions (en.wikipedia.org).

This announcement marks one of the most ambitious enterprise AI infrastructure projects to date. By vertically integrating chip production and targeting unprecedented compute scale, Terafab could reshape the semiconductor landscape and redefine how AI compute is provisioned for both terrestrial and extraterrestrial applications.

Why it matters for enterprise AI:

  • Scale and autonomy: Terafab promises unmatched compute capacity, reducing reliance on external foundries and supply chains.
  • Innovation acceleration: On-site rapid iteration could dramatically shorten chip development cycles.
  • Strategic control: Owning the full stack—from design to packaging—gives Tesla and SpaceX unprecedented control over AI hardware tailored to their specific needs.
  • Broader implications: If successful, Terafab could set a new standard for enterprise AI infrastructure, influencing how other organizations approach compute sovereignty and vertical integration.

As of March 23, 2026, Terafab remains in the early stages, with prototype operations planned in Austin. The full-scale facility’s location and timeline remain undisclosed, but the project’s scale and ambition already signal a potential paradigm shift in enterprise AI hardware strategy.