Elon Musk merges SpaceX and xAI in $1.25T mega-deal

SpaceX just swallowed xAI and X whole.

Elon Musk announced the merger of his rocket company with his AI venture, creating a combined entity valued at $1.25 trillion, according to Bloomberg. The deal, completed on Monday as reported by CNBC, crowns the merged operation as the world’s most valuable private company ahead of an anticipated SpaceX IPO later this year.

Musk’s pitch? Earth simply does not have enough electricity to keep feeding frontier AI models, so the only logical solution is to launch data centers into orbit.

Taking the compute cluster to orbit

The merger formalizes an audacious vision that began taking shape last week, when SpaceX filed an application with the FCC to launch a constellation of up to 1 million data center satellites. FCC chairman Brendan Carr accepted the filing for public comment and unusually shared it directly on X.

Speaking on John Collison’s Cheeky Pint podcast with guest Dwarkesh Patel, Musk doubled down on his math. He claimed that solar panels in orbit capture roughly five times more power than on Earth due to unfiltered sunlight and zero day-night disruption.

Musk predicted that within 30 to 36 months, space will become the most economically compelling place to generate AI compute. He went even further, predicting that five years from now, the industry will launch and operate more compute in space annually than the cumulative total on Earth.

According to Musk, the profits generated by orbital compute will eventually fund self-growing bases on the Moon and human civilization on Mars.

The engineering reality check

Naturally, skepticism is already mounting across the tech sector.

During the podcast, Patel pointed out the glaring holes in the pure-power comparison: electricity is far from the only line item in running an AI cluster. Launch costs, vacuum thermal management, and orbital radiation pose massive hurdles. Crucially, if enterprise GPUs fail during large-scale model training runs—a routine occurrence in terrestrial clusters—servicing hardware thousands of miles above Earth is effectively impossible today.

There is also the corporate baggage. The deal ropes in X, which is already under intense regulatory scrutiny and investigation in the EU over Grok generating nonconsensual sexualized deepfakes.

Financially, however, the structure makes total sense for Musk. Reuters reported that SpaceX generated about $8 billion in profit last year, and Tesla disclosed a $2 billion investment into xAI just last week.

Why it matters

Musk is engineering the ultimate captive customer loop. xAI is an expensive model developer burning billions on training runs, while SpaceX builds the heavy-lift rockets. If orbital data centers become an accepted thesis, SpaceX essentially manufactures its own endless launch demand, justifying a massive valuation bump as it approaches an IPO.

For enterprise developers and everyday users, there is zero immediate utility here. You will not be querying satellite-hosted Grok instances anytime soon, and the engineering barriers to keeping thousands of GPUs running reliably in microgravity are immense. We think tying a sensitive national-security launch provider to X's regulatory dumpster fires in Europe is a noticeable liability risk.

Yet for the broader AI race, this signals just how desperate the search for power has become. With global data center demand projected to hit 200 gigawatts by 2030, tech giants are hitting hard limits on terrestrial grids. Musk might sound borderline unhinged talking about sentient suns and orbital server farms, but he is forcing the entire industry to confront the reality that terrestrial infrastructure is running out of road.