Musk’s TeraFab blueprint

Elon Musk unveiled the project at the defunct Seaholm Power Plant in Austin during a weekend presentation on March 21, 2026, and framed it as a joint effort across Tesla, SpaceX and xAI. (electrek.co) Musk said the facility will produce two families of chips — edge inference processors for vehicles and robots and a separate "D3" line hardened for space — and the team is targeting a 2‑nanometer process node. (cbsnews.com) Company materials and coverage say the ramp plan calls for roughly 100,000 wafer starts per month initially with an ambition to scale toward 1 million wafer starts per month, and public estimates cited annual output targets in the low‑hundreds of billions of custom chips. (electrek.co) SpaceX has separately filed with U.S. regulators for permission to deploy as many as one million orbiting data‑center satellites, and the company presented designs that emphasize long solar arrays and laser links for inter‑satellite communications. (techbuzz.ai) Coverage and industry analysts cautioned that Musk’s on‑stage “launch” was a formal project announcement rather than a ready‑to‑run fab, noting that building and qualifying leading‑edge semiconductor fabs typically takes multiple years and hundreds of billions of dollars; Forbes recorded Tesla’s small‑batch target for the new inference chip in 2026 with volume slated for 2027 but warned the term “launch” does not equal an operational fab. (forbes.com) Regulatory analysts flagged wider implications: even a small fraction of the satellite counts proposed would multiply today’s orbital traffic many times over (the European Space Agency estimates roughly 15,000 satellites currently in orbit, with Starlink accounting for over 11,000), a point likely to influence licensing and space‑traffic evaluations. (techbuzz.ai)