SpaceX goes public. The pitch isn't what you think.

SpaceX hit the Nasdaq this morning under the ticker SPCX, and if you've been following the story assuming this is about rockets and Martian ambitions, you've rather missed the actual pitch.

The offering raised roughly $75 billion and values the company at somewhere north of $1.75 trillion. To put that in context: that's more than the UK economy produces in an entire year, and the fundraise alone dwarfs every US IPO combined during the peak of the dot-com bubble. It is, almost certainly, the largest public market debut in history.

A significant portion of the reason investors are piling in has very little to do with Falcon 9, Starship, or putting humans on Mars. It has everything to do with AI, and specifically a plan to solve AI's exploding energy problem by building data centres in space.

I'm not joking.

The power problem is real

AI compute is hungry in a way that's genuinely difficult to overstate. Global data centre power consumption is expected to nearly double to around 1,000 terawatt-hours by the end of this decade. Planning permission for new data centres is becoming contested. Grid connections take years to secure. The electricity infrastructure was not designed for this, and the problem is getting worse faster than the solutions are arriving.

Musk's answer, delivered with the confidence of a man who has already landed booster rockets on drone ships in the middle of the ocean, is essentially: fine, we'll go somewhere where it's always sunny.

His reasoning is straightforward. On Earth, you're power-constrained. Grid connections are slow. Solar panels only work half the time. Cooling data centres costs a fortune. In space, solar power runs continuously, there's no overheating to manage (you just radiate heat away into the void), and nobody is going to complain about the planning application.

It's theoretically elegant. Whether it's practically achievable in the timescale Musk is promising is a different conversation, and we'll get there.

What they're actually building

SpaceX has already unveiled the AI1 satellite. It spans over 70 metres when deployed, wider than a Boeing 747, and carries a 150 kilowatt compute payload. The company has filed FCC plans for a constellation of up to one million satellites, announced an 11-million-square-foot factory in Texas dedicated to manufacturing them, and set a target of one gigawatt of orbital AI compute capacity by late 2027.

Earlier this year, Musk merged SpaceX with xAI, his AI company. This turned out to be useful context for the IPO filing, which quietly revealed that Anthropic, one of the leading AI labs and a direct competitor to xAI, has signed a contract to pay $1.25 billion per month for 300 megawatts of xAI data centre capacity. Per month. The broader AI industry is already handing Musk extraordinary sums just for the terrestrial infrastructure. The orbital version is the next chapter of the pitch.

What the experts actually think

Here is where the reality check arrives. The expert consensus is, to be diplomatic about it, deeply sceptical.

A Carnegie Mellon engineering professor describes Musk's 2027 timeline as "an optimistic interpretation." A Georgetown University analyst puts it more plainly: "This is something people are cynical about because it's just technologically not feasible at the moment."

The numbers explain the gap. To replicate a 100-megawatt data centre in space, a moderately sized one by current standards, you'd need a facility 500 to 1,000 times larger than the AI1 satellite. Heat dissipation through radiation works at small scales but becomes a genuinely hard engineering problem as you scale up. Google analysis suggests launch costs need to fall by at least a factor of five, to around $200 per kilogram, before any of this makes economic sense.

So: technically possible in principle, commercially unproven in practice, and 2027 is carrying a lot of weight as a delivery date.

Why this matters if you're in the IT Channel

The energy problem isn't going away regardless of whether orbital data centres materialise on schedule or get quietly repositioned as a "2032, subject to regulatory approval" target. If you're advising clients on cloud strategy, AI workloads, or infrastructure planning, the power question is already affecting decisions in ways that weren't on anyone's agenda two years ago. Data centre availability, grid capacity, and cooling costs are becoming real factors in enterprise buying conversations, not just issues for the facilities team.

Microsoft, Google, and Amazon are all signing nuclear power deals and competing for grid connections precisely because the demand curve for AI compute is running well ahead of the infrastructure to support it. The SpaceX pitch, outlandish as parts of it are, is trying to solve a real problem.

Whether it works matters less to most of us in the channel than what happens to the energy costs and availability of cloud compute over the next five years. Understanding why this money is being raised, and the scale of the problem it's trying to address, is worth your time even if you have no intention of buying SPCX shares.

One last thing

The irony of solving Earth's energy crisis by launching satellites into orbit using rockets that burn through significant quantities of kerosene per flight appears to occur to very few of the people enthusiastically wiring money into this IPO. Either the numbers genuinely work at scale and I'm missing something, or we are witnessing one of the greatest feats of collective investor confidence since somebody convinced people to buy pets.com.

Ask me again in 2028 how the gigawatt of orbital compute is coming along - maybe then we'll have 'Sharks with frikking laser beams'.