I met the number on June 12, 2026, the same way you meet a wave you didn't see coming. SpaceX began trading on the Nasdaq under the ticker SPCX, priced at $135 per share, raising $75 billion in one go, the largest IPO in the history of recorded capital markets, bigger then Saudi Aramco in 2019, bigger then anything, really, and the crowd went absolutely mental about it, as crowds do, but I kept thinking, sitting there with the numbers spread in front of me, I kept thinking: they are buying the wrong story.

Right...

Here is the thing, and I'm going to try say this plainly because the S-1 says it in language that wanders and buries and insulates itself in clauses but the core of it is this: the orbital AI data center thesis is buried on page after page of the filing, and it is not a footnote, it is the forward motion, the gravity, the reason the company is priced at roughly $1.77 trillion when Morningstar's bear case fair value sits somewhere between $600 and $800 billion for the Starlink business alone, which is the only bit of the operation that actually turns a profit, which is, I don't know, fine, I suppose.

Starlink generated $11.4 billion in revenue in 2025 at a 63% EBITDA margin with 10.3 million subscribers across more than a hundred countries by the end of Q1 this year, and that is by any measure a real business, a serious business, the kind that justifies a serious valuation; but here is where you have to slow down and do the maths because at $1.77 trillion, the $600 to $800 billion Starlink business accounts for less than half the price you are paying today, and the remaining trillion or so is a wager on Starship at commercial scale, on the xAI division burning through $6.36 billion in operating losses annually since the February 2026 merger, on a plan to put data centers into orbit by 2028, where there is no atmosphere to cool the servers and also you have to get there on a rocket, which is speculative in the way that most things worth caring about are speculative, which is to say: seriously.

We must face what is actually being sold here.

We must confront the fact that AI capex in SpaceX's own filing constitutes 76% of total group capital expenditure in Q1 2026.

We shall not look away from the figure: $7.7 billion spent on xAI infrastructure in a single quarter, buried under the headline of a rocket company, competing with OpenAI and Google with a unit that has never yet turned a profit, and the valuation prices all of that in at day one, leaving nothing in the margin for error.

Where the stars wait and the servers spin, where the cold void and the compute begin, and the satellites sings the silence thin.

Look, the meter's a bit rough and I know it. But space-based compute really does sound like something out of science fiction until you read the actual S-1 filing and there it is in plain corporate prose: orbital data centers as a primary growth driver, utilizing the solar energy efficiency and the vacuum cooling of space to run next-generation AI workloads, currently planned for 2028, completely unproven, and magnificently, TERRIFYINGLY expensive.

Thirty percent of shares was reserved for retail investors through brokers like Robinhood and Fidelity. Which means the person buying a little slice of SPCX on their phone this week, the one who loved watching rocket landings on YouTube and thinks this is a bet on Musk's brilliance and the future of space travel, they are also, without necessarily knowing it, buying into one of the aggressive AI infrastructure plays in capital market history, a company now absorbing the losses of a gigawatt-scale data center operation that is burning cash at a rate that the profitable satellite internet business can not on its own offset.

The xAI division alone absorbed a $250 billion all-stock acquisition in February and carries cumulative total losses sitting at $41.3 billion.

That is a lot of money for a company that also happens to build rockets!

The valuation at $135 a share represents approximately 94 times 2025 revenue. Those are not conventional multiples. Their not even venture multiples, their something else, their a statement of religious conviction about what computing in low Earth orbit will eventually be worth, and maybe, maybe that conviction is right, but the conviction is not the proof.

I am not saying do not buy it. I am saying: know what you are buying. Because the orbital compute thesis might be the single large capital-formation story of the next decade, where silicon floats above the stratosphere and the economics of cooling and energy change entirely, and if that is true then $135 looks cheap and the bears simply failed to imagine the possible.

But the possible is not the same as the probable, and nobody at the retail level is being handed that distinction clearly.