The Physical Layer

AI and energy infrastructure, explained plainly.

The Physical Layer

Research · Initiation · 2 November 2026

SpaceX Nasdaq: SPCX

SpaceX's price leaves about US$404 billion for businesses it does not yet report. Orbital compute is the largest of them.

The call No call
Price at publication
US$134.72close, 30 October 2026
Target
Noneno position at this price
Conviction
Medium

What would prove the thesis wrongThe shares close above US$175 or below US$90 on 29 October 2027, or SpaceX's reported 2027 revenue lands below US$130 billion or above US$260 billion, either of which would put my base value outside the no-call band at today's price.

What would turn this into a callA price above about US$150, where my base value would sit 25 per cent below it, or below about US$98, where it would sit 15 per cent above it; or SpaceX disclosing an AI compute satellite weighed in service at 100 kilowatts or more per tonne launched, which would break my piece's claim.

Read the full notePDF, 6 pages: thesis, valuation, scenarios, risks and sources Open the modelExcel, live formulas: change an assumption and the target moves

The call in brief

  1. SpaceX reports three businesses: launch, Starlink, and Super Intelligence, which rents out AI computing and sells Grok, X and Cursor. Orbital data centres sit in none of them and earn nothing yet.
  2. At US$134.72 SpaceX is valued at about US$1.78 trillion after net cash. On my 2027 estimates its three businesses are worth about US$1.37 trillion, which leaves about US$404 billion, 23 per cent, for everything else.
  3. That gap is about six gigawatts of orbital computing valued like SpaceX's ground computing. At the 75 kilowatts per ton SpaceX states for its first design, read as compute per metric ton, it is about 38 times the mass SpaceX put into orbit in all of 2025.
  4. My base value is about US$112, 17 per cent below the price and inside my no-call band; the bear case is US$45 and the bull case US$203.
  5. SpaceX said on 20 October that its first Starmind satellites fly in 2027, a year ahead of its prospectus, and it gives its first design a 'vehicle efficiency' of 75 kW per ton, a quarter short of its target if that is compute per metric ton.