CryptoLite_247

CFO Bret Johnsen at Goldman:
• Closed earlier this month
• $1.11 billion a month starting December 1
• Adds roughly $13 billion of ARR
$SPCX
Today at Goldman Sachs Communacopia, SpaceX CFO Bret Johnsen sat with Eric Sheridan for about 34 minutes. Not an earnings call. No slide deck theater. He walked through the three things that actually matter from here: the near-term print, Starship turning into a production machine, and orbital compute arriving sooner than the market wants to believe.
Most people will screenshot “$100 billion ARR.” The useful part is how that number is built, why the contracts are deliberately short, and why Flight 14 is the date the CFO actually sounded excited about.
1. What he actually said about the new contract
Johnsen’s update was specific.
They signed another hosting contract earlier this month.
It starts December 1.
It is about $1.11 billion a month.
That is another ~$13 billion of ARR.
He said that contract gives them even more conviction in the year-end $100 billion ARR target.
• $100B ARR is the December month annualized. It is an exit run rate, not $100 billion of revenue already booked in 2026. Q2 revenue was about $7.8 billion for the whole quarter. The year-end target is about the December exit, not the trailing twelve months.
• Almost all of these compute contracts are structured as roughly 90 days plus a 90-day out. Call it a six-month commitment. This is not a ten-year take-or-pay.
So the $1.11B/month print is real and it is large. The quality of it depends on whether it ramps on December 1 and whether the customer stays.
2. Why this contract changes the $100B path
Johnsen framed terrestrial compute as the thing currently bending the company’s P&L.
Capacity from this talk:
• A little over 2 GW by year-end 2026
• 5–10 GW deployed in 2027
• Monetization next year: $30–$50 per watt, and they are already at the high end of that range
• Payback on new compute deployments: under one year
• NVIDIA: exclusive relationship. GPU allocation is the one piece they do not fully control
• Power, buildings, permitting: he said they would not give the 5–10 GW range without line of sight
This is not just another headline contract. SpaceX is selling compute the same way it learned to sell launch and connectivity: own the building, own the power, stand the machines up, take the customer all the way through. Sheridan even introduced the company as SpaceXAI. After the xAI combination, that is the story they want investors to hold.
Vertical integration was the first answer in the room, not a throwaway. Rockets first, because the supply chain could not move at their pace or hit their quality bar. Then Starlink: launch, satellite, customer. Now AI: they are the GC on the building, they put up the power, and they take their own models straight to consumer and enterprise.