Google and SpaceX are in advanced discussions about launching AI data centers into orbit, according to a report from Bloomberg citing the Wall Street Journal on May 12, 2026. The idea, which sounded fanciful even three years ago, is now central to SpaceX's pitch for its planned $1.75 trillion IPO later this year. Investors are being told that orbit may be the cheapest place on the future map of AI compute.

For Google, the conversation builds on its existing moonshot. TechCrunch notes that the company announced Project Suncatcher last year, with the aim of launching prototype satellites that test how machine-learning workloads behave in space by 2027. Google also already owns about 6.1 percent of SpaceX, so a deeper partnership is more an evolution than a surprise.

Why orbit, why now

The argument for space-based compute is mostly thermodynamic. Cooling is the single largest energy cost in a modern hyperscale data centre, and solar power in low Earth orbit is both more constant and roughly eight times stronger than at the surface. If panels and radiators can be scaled, advocates argue, the energy and water demands that are now straining municipal grids around AI campuses could be sidestepped entirely.

The argument against is that nothing else about it is easy. Radiation degrades silicon faster in orbit than on the ground. Servicing failed hardware in space is expensive at best and impossible at worst. Round-trip latency to ground stations limits which workloads make sense up there. And the launch math, even with reusable rockets, has to clear a much higher bar than running another fibre line to a desert facility in Arizona.

One piece of a wider SpaceX compute push

Anthropic has separately signed a deal to use the full compute capacity at SpaceX's terrestrial Colossus 1 data centre, including more than 300 megawatts and over 220,000 Nvidia GPUs (see our earlier coverage of Anthropic's parallel deal with Akamai). Between Suncatcher, Colossus, and a possible orbital deployment, SpaceX is positioning itself less as a launch provider for the AI industry and more as one of its largest infrastructure landlords.

Sources

  1. i. www.bloomberg.com
  2. ii. techcrunch.com
  3. iii. finance.yahoo.com
  4. iv. www.technobezz.com
  5. v. english.cw.com.tw

Commentarii · 0

Add · a · Comment