Title: Google's Space Datacenter Plan: Ambitious or Absurd?
Google wants to put datacenters in space. It's a bold claim, aiming to launch trial equipment by early 2027. Project Suncatcher envisions constellations of solar-powered satellites, packed with Google's TPUs (processors optimized for AI). The rationale? To meet the surging demand for AI while minimizing the environmental impact of terrestrial datacenters. Google plans to put datacentres in space to meet demand for AI
But let's break down the numbers, because the devil, as always, is in the details. Google claims that by the mid-2030s, space-based datacenter running costs could be comparable to Earth-based ones, thanks to falling launch prices. Could be is doing a lot of heavy lifting there. What discount rate are they using to project those savings? What's the assumed learning curve for on-orbit maintenance? These are not trivial questions.
Orbital Ambitions vs. Earthly Realities
The argument hinges on solar panel productivity – supposedly eight times higher in space. That's a compelling figure, but it ignores a few crucial factors. First, those panels will degrade over time, particularly in the harsh radiation environment of low Earth orbit (LEO). Second, you still need to get the power down to Earth, and beaming energy through optical links (as Google proposes) introduces its own inefficiencies and potential for interference.
And this is the part of the analysis I find genuinely puzzling: the CO2 emissions from launches. Google acknowledges that each rocket launch emits "hundreds of tonnes of CO2." Philip Johnston, co-founder of Starcloud (which is launching Nvidia AI chips into space), claims a ten-fold carbon dioxide savings over the datacenter's life. But how did he get there? What’s the lifecycle carbon footprint model they are using? What’s the assumed lifespan of the space-based datacenter versus a terrestrial one? What about the carbon cost of building the satellites in the first place? These omissions make the environmental argument feel, at best, incomplete.
Elon Musk is also in the mix, planning to scale up space-based datacenters via Starlink and SpaceX. Nvidia is partnering with Starcloud to launch AI chips. The trend is clear: a space-based computing gold rush is on. But is it driven by genuine efficiency or by the allure of escaping earthly constraints?
One major hurdle is space debris. With over 10,000 satellites already in orbit, the EU is proposing stricter regulations on satellite operators to address space pollution. The US is pushing back, calling the EU's proposed space law "unacceptable" and arguing that it could stifle innovation. It's a classic regulatory standoff: environmental concerns versus technological progress. The US stance, in this case, seems particularly short-sighted. Ignoring the long-term consequences of orbital overcrowding is a recipe for disaster – a tragedy of the commons on a cosmic scale.
The Space Force Angle
Meanwhile, the US Space Force is grappling with its own challenges. A Mitchell Institute report urges the Space Force to prioritize "dynamic space operations" (DSO) to enhance the resilience of space systems. This includes on-orbit refueling, in-space assembly, and even "decoys" to confuse adversaries. The report highlights China's growing space capabilities and the need for the US to maintain its edge. From on-orbit to launch, Space Force needs more focus on dynamic space operations
The Space Force's focus on DSO reveals a fundamental tension. Is space primarily a domain for scientific exploration and commercial innovation, or is it a strategic high ground in a new era of great power competition? The answer, of course, is both. But the increasing militarization of space adds another layer of complexity to the datacenter equation. What happens to Google's satellites if they become targets in a future conflict? What are the cybersecurity risks of having critical infrastructure in orbit? These questions demand answers, and the current discourse is noticeably silent on them.
The Hype Doesn't Match the Data
Google's Project Suncatcher is a fascinating concept, but the available data doesn't fully support the company's claims of environmental and economic viability. Significant engineering challenges remain, as Google itself admits: thermal management, high-bandwidth communications, and system reliability. Until these challenges are addressed with concrete data – not just optimistic projections – the idea of space-based datacenters remains more science fiction than sound investment.