Stray Dog
EN中文

Prices are Hyperliquid perpetual futures: contracts that track each stock and trade 24/7. They are not exchange prices and can differ from them, most of all while US markets are closed. Change is over 24 hours. For information only.

  • TSLA—
  • PLTR—
  • HIMS—
  • MU—
  • SPCX—
  • SOFI—
  • BTC—
…Loading prices
TSLA

Terafab reversal to build and run it alone, record California Powerwall virtual power plant…

Sources (9)

Three developments. (1) On 7 Oct Musk himself overturned the 6 Oct line that TSMC was joining Terafab. He made clear that Tesla and SpaceX will build and run it themselves, that TSMC can only be a tenant, and that the process bet rests solely on Intel's 14A.

Read more

(2) At the California demand peak on 6 Oct, 85,774 Powerwall homes sent back 601 MW, and the whole fleet peaked at 634 MW, roughly the size of a large utility-scale power plant. (3) A Cybercab in Philadelphia was in a low-speed collision, and the whole left-side door panel came off, exposing the internal structure.

Key evidence

  • Musk on X, 7 Oct, in his own words: 'We will build and run the fab. Let there be ZERO doubt about that.' TSMC may rent a part of Terafab, but that is all.
  • Intel CEO Lip-Bu Tan confirmed in Tokyo the same day that Intel remains the only foundry partner, supplying the 14A process (High-NA EUV).
  • After the news, Intel rose about 2% pre-market and TSMC's US-listed shares fell about 1.3%.
  • Terafab investment starts at $16.8 billion and underpins the compute base across four lines: FSD, Robotaxi, Optimus and xAI.
  • The earlier line on compute allocation, about 25% to Tesla Optimus and 75% to SpaceX, is unchanged.
  • At the California demand peak on 6 Oct, 85,774 Powerwall homes fed back 601 MW, and the whole Powerwall fleet peaked at 634 MW.
Show 4 more
  • Energy storage had a gross margin of 20.4% in Q2 against 16.3% for automotive, the highest of any business in the company.
  • The virtual power plant extends the one-off sale of Powerwall hardware into long-term grid services and recurring peak-shaving revenue.
  • @TeslaZoa posted on X on 7 Oct: a Philadelphia Cybercab was in a low-speed collision, and the whole left-side door panel fell off, exposing the internal structure.
  • This adds to earlier ride-experience problems, such as the 17 Sep case in Austin in which a 10-minute ride took 70 minutes, and butterfly doors and a trunk that would not close properly.

Catalysts to watch

  • Yield and progress checks on Intel's 14A (High-NA EUV)
  • A formal agreement and timetable for Terafab
  • The revenue-sharing model and expansion of the California virtual power plant
  • Growth in Powerwall and Powershare installations driving service-based storage revenue
  • Accumulating samples and quality data from Cybercab's training deployment in Philadelphia

Risks

  • Building and running it alone means Tesla and SpaceX carry the advanced-process production risk by themselves, and Intel's 14A is itself unproven
  • So far there is only a single reply from Musk on X, with no formal agreement or change to the timetable
  • Four lines compete for the same compute base, and the allocation line is unchanged
  • The 634 MW is an instantaneous peak on a single day, not sustained capacity, and VPP revenue-sharing is still small change against total revenue
  • The virtual power plant depends on California's electricity pricing mechanism and compensation incentives, so it relies heavily on the regulatory regime
Show 2 more
  • The door panel coming off in a low-speed collision is a single case, but combined with the problems in the ride experience it points to manufacturing consistency and crash tolerance of the steering-wheel-free dedicated vehicle that are yet to be proven
  • A door panel that comes off easily may also be a modular design that makes replacement easy, and more samples are needed

Sources 9

  1. Musk on X, 7 Oct
  2. TeslaNorth
  3. Pomegra
  4. Reuters
  5. Tesla official X, 6 Oct
  6. DriveTesla
  7. 24-7WallSt
  8. @TeslaZoa on X, 7 Oct
  9. NotaTeslaApp

A brief is a lead for our research, not a verified conclusion or advice.