On October 21, 2026, Tesla filed its Q3 2026 Update with the SEC. Total revenue was $28.0 billion, GAAP operating income $0.7 billion and GAAP net income $0.8 billion. Tesla now describes its AI work as “Super Intelligence (SI),” and the letter says the quarter showed “strong momentum on many projects spanning Super Intelligence (SI), robotics and renewable energy.”
On robotics, Tesla reported that “Optimus production began in Fremont on our high-volume lines,” calling them “the first production-quality bots that we have manufactured.” On autonomy, it began deploying Cybercab into its Robotaxi fleet in Austin on September 3, reached nearly two million cumulative unsupervised Robotaxi miles, and said active FSD subscriptions rose to 1.74 million. FSD (Supervised) is now approved in nine EU countries.
On compute, the letter lists Cortex 1 (more than 105 MW) and Cortex 2 (more than 120 MW) in production in Texas and says: “We plan to reach 400 MW of SI training compute by the end of the year.” It also reports continued construction of an Austin chip fab, with equipment commissioning planned for 2027. Tesla links its energy storage business to AI directly, saying Megapack demand “is increasingly aligned with SI infrastructure demand.”
Why it matters: Tesla is one of the few companies trying to train AI for physical machines at data-center scale while also building those machines. Reporting training compute in megawatts rather than GPU counts, and starting humanoid robot production on a high-volume line, are concrete markers of how far that bet has moved from demos.
What it does not show: Tesla gives no Optimus unit volumes, no revenue from robots or Robotaxi, and no safety statistics for unsupervised miles. The 400 MW figure is a plan, not installed capacity, and Tesla’s FSD (Supervised) still requires active driver supervision.