I never wrote a SpaceX deep dive because the company is already picked apart every week.
Facts first, then what I think they mean for us as $SPCX owners.
On 10 September, at Goldman’s Communacopia conference, SpaceX CFO Bret Johnsen said the company closed another AI hosting deal earlier this month. This translates into about $1.11 billion a month starting 1 December 2026, or roughly $13 billion of ARR, and that gives them even more conviction on the $100 billion ARR target by year-end.
Two things get mixed up in the coverage. ARR here is an exit run-rate, meaning December annualized. It is not a claim that SpaceX will report $100 billion of 2026 revenue.
Johnsen also said almost all of these hosting deals are 90 days plus a 90-day out, so call it a six-month commit. That is not a ten-year Microsoft Azure contract. SpaceX wants the option to pull capacity back for its own models. That flexibility is a feature for them and a risk for anyone treating $13 billion as locked forever.
The compute relationships already in the public record are Anthropic at about $1.25 billion a month through May 2029, Google at about $920 million a month from October 2026 through June 2029, the new unnamed December deal, and a separate ~$6.7 billion six-month cloud deal discussed on the August call. William Blair suspects that last one is DoD. SpaceX did not name it. Capacity target from the same session is a little over 2 GW terrestrial by end-2026 and 5 to 10 GW in 2027, still on Nvidia for the GPUs.
On 25 August SpaceX and “Louisiana” announced Starbase Louisiana in Vermilion Parish. About 125,000 acres, a stated $100 billion investment, the fourth and largest launch site, five complexes, ten pads, propellant, power and housing. Construction is aimed at 2027 and the first launch is targeted as soon as 2029.
The official job figure is 3,000 direct over ten years at about $93k average. Shotwell said the site is also there to put mass in orbit for orbital data centers.
(Btw I think we should all be happy when a company creates good-paying jobs and attractive opportunities. And these are mainly field roles, not laptop remote jobs.)
But in all honesty, this is not relevant for 2026 ARR. It is basically a cadence insurance for Starship and for the orbital compute story. If you cannot launch often, you cannot build a data center in orbit, and Texas alone is not enough pads for a launch every few hours.
Louisiana’s incentive package is large. Reports put the tax package in the mid-twenties of billions, mostly a long property-tax holiday, with annual payments back to the parish instead. Watch permitting and watch whether construction actually starts in 2027.
On the SpaceX-Tesla-merger I am going to say this as my view:
I think an all-stock Tesla combination gets announced in the first half of 2027. The operational overlap is already there: Terafab with Tesla, Tesla’s disclosed SpaceX stake and framework agreement, Cybertruck spend, shared talent, Grok inside the house, and cars and robots that need cheap launch and cheap compute. Shotwell already said on IPO day that putting the companies together might make Elon’s life a little easier and that the synergies are visible.
Wall Street is split. Dan Ives has been bullish on a 2027 merge. Prediction markets have been pricing a 2027 announcement as a coin flip. Oppenheimer has argued they may be more useful apart.
If it happens, you get launch, constellation, terrestrial and orbital compute, vehicles, robots and an in-house model under one roof. That combination does not exist at one other public company. What it is worth is not a spreadsheet exercise I trust, and I will not pretend it is.
From about $2 trillion today, $10 trillion in 2030 is about a 5x. I can say the sentence but I cannot model it cleanly. It only works if the $100 billion exit run-rate becomes a much larger earned revenue base, launch cadence actually shows up in Louisiana, and the multiple does not collapse when growth slows.
On physical AI and growth:
McKinsey’s midpoint is about $2.9 trillion of US economic value by 2030 if agents and robots get into workflows rather than slide decks. They also put the global data-center build in the $7 trillion range for 2025 to 2030. Goldman’s real-economy note is that the next phase is factories, robots, power and logistics, not another chatbot. Anthropic’s own scenario work has a wide band: a modest GDP lift by 2030 on one end, and something much larger only if adoption is extreme.
Ok back to SpaceX
This is a category-defining company.
Launch is already a quasi-monopoly in the West. Starlink is the civilian and government connectivity layer that other satellite stories have to compete with. The compute business is new and the contracts are short, but the customer list is not a science-fair.
Vermilion is the attempt to make Starship a factory, and Governments already treat this “stack” as infrastructure. You do not recreate the pads, the constellation, the talent and the relationship with Washington easily, aka we have a big fat moat.
SpaceX is a lifelong hold for me the way Coca-Cola is for Warren Buffett.
I like the mission, and I think this company has to exist and adds value to humanity.
Thanks for reading
FJ
—
FJ Research covers the few category-defining companies that can dominate their industry and still multiply in the decade ahead.
This is for folks who want a concentrated book and the thinking behind it. What is in, what came out, and why.
If that is how you invest, subscribe.
Sources
Bret Johnsen, Goldman Sachs Communacopia & Technology Conference, 10 September 2026 (Yahoo Finance / Pras Subramanian, Seeking Alpha, Benzinga)
SpaceX Q2 2026 results, Form 8-K / Exhibit 99.1, 4 August 2026
Louisiana Economic Development, NYT and NOLA.com on Starbase Louisiana, 25 August 2026
Gwynne Shotwell, CNBC IPO-day interview on Tesla synergies, June 2026
McKinsey MGI, agents and robots / AI economy notes, 2025–2026
Goldman Sachs Investment Banking, Harnessing AI for the Real Economy, June 2026
Anthropic, Scenarios for our Economic Future
Positions & avg. entry: