Keeping up with artificial intelligence rivals is expensive, but it isn’t clear Elon Musk’s big idea — computing in space — is going to help the bottom line.

This week, the X boss is consolidating his business empire. His rocket and satellite company SpaceX will acquire his AI operation xAI, which runs Grok, in a $1.25 trillion merger. Musk has done it before: xAI previously acquired Twitter, now X. If SpaceX floats later this year, in what may be the biggest IPO of all time, more capital will be made available.

To sweeten the deal, he elaborated on his vision of privately owned data centres, beaming back data to anywhere on Earth, entirely free from the reach of regulators. “This marks not just the next chapter, but the next book in SpaceX and xAI’s mission: scaling to make a sentient sun to understand the Universe and extend the light of consciousness to the stars!” Musk wrote in a memo to staff.

The vast capital drain of developing and operating Grok forced his hand on the merger. xAI, which produces the Grok chatbot, reported a $1.46 billion loss last month, and burned through $8 billion in nine months last year. The monthly expenditure to maintain Grok is around $1 billion. Revenue is negligible, a problem shared by all the major US AI companies as businesses seek to turn hype into tangible productivity gains. Indeed, 95% of companies surveyed by MIT saw no return at all from investment in AI.

Musk has placed bets across the board: outpacing Waymo in robotaxis, the Chinese in humanoid robotics — yet data centres in space remain his most audacious gamble. The data centres will be hoisted into orbit via SpaceX’s large-capacity rockets, and then transmit data via Musk’s Starlink satellite network, which is owned by SpaceX.

Here, Musk is combining both Old West and New Californian myths. Space is both a new frontier to conquer, a desire expressed in William Gilpin’s declaration of manifest destiny as “Divine Task! Immortal Mission!”. It’s also a zone independent of regulators, an idea first conceptualised in John Perry Barlow’s Declaration of the Independence of Cyberspace in 1995, which is still the ur-text for techno-utopians. An orbiting Musk computing facility would be a temporary autonomous zone made possible by the tycoon’s wealth — for now, at least. No wonder the same proposition has also been advanced by Jeff Bezos, who runs a rival launch and satellite business, and OpenAI CEO Sam Altman.

However, the economic reality is brutal. Satellite connections are slower than fibre, with longer delays or “latencies”. An orbiting data centre is vastly more expensive to build than one on the M4, even given SpaceX’s formidable rocket capacity. The Falcon Heavy rocket has lowered the cost of orbital payload to below $3,000 per kg, and can send almost 64 tonnes of infrastructure into orbit at a time. More shielding is required, but so is more power.

This expense may be worth it if the output — what is generated in space — is extremely valuable. For example, zero gravity enables purer crystals to be manufactured than can be on Earth, or what British space startup Space Forge calls “the saffron of the semiconductor industry”. Space Forge has received investment from Nato to produce the material for chips.

Yet the output of a data centre is bits, an undifferentiated product, and a commodity: buyers are keenly aware of price. Unless SpaceX can create something as valuable as saffron, the appeal may be narrow. Investors will therefore look for more viable ways of recouping their capital — particularly if, as expected, AI bubble valuations fall back to earth.

---

Andrew Orlowski is a business columnist at The Daily Telegraph and has covered technology competition lawsuits for 25 years.

AndrewOrlowski

---