China wants to build an AI data center in space

China Central Television (CCTV) reported on January 29 that China Aerospace Science and Technology Corporation (CASC) is committed to “building a gigawatt-capacity smart digital infrastructure in outer space” within the next 5 years. The new data centers are expected to integrate cloud computing, edge computing and terminal capabilities, and achieve deep integration of computing power, storage capacity and transmission bandwidth, allowing efficient processing of data from Earth.

Reuters Evaluation, this move poses a competitive challenge to Elon Musk’s plan. SpaceX, the American aerospace company founded by Musk, wants to use the proceeds from its IPO expected to be worth $25 billion this year to develop an AI data center in orbit, thereby solving energy limitations on Earth.

At the World Economic Forum (WEF) taking place in Davos (Switzerland) last week, Musk said SpaceX will launch a solar-powered AI data center in space within the next two to three years, asserting this is the place with the lowest cost to operate AI. According to this billionaire, placing solar cells in orbit can generate 5 times more electricity than on the ground.

According to last month’s CASC policy document, China also plans to shift the energy “burden” of AI operations to orbit, using gigawatt solar power centers to create an industrial-scale “Space Cloud” by 2030. The document identifies the integration of space solar power with AI computing as a key goal of China’s 15th Five-Year Plan (phase 2026-2030).

Musk is also said to be preparing to merge SpaceX into artificial intelligence company xAI or car company Tesla. According to ReutersSpaceX could merge with xAI before the IPO, bringing together products such as Grok chatbot, X social network, Starlink satellites and SpaceX rockets under one corporation. This move could create new momentum for plans to build data centers in outer space.

In addition to SpaceX, billionaire Jeff Bezos’ aerospace company Blue Origin is also developing technology for orbital AI data centers. Bezos predicts they can “beat” sub-Earth facilities in terms of cost within 10-20 years by continuously exploiting solar energy and radiating heat directly into space.

Space data centers can consist of many satellites connected together to handle the huge computing needs of AI models, for example Grok or ChatGPT, in the context of centers on Earth consuming too much energy and operating costs are increasing. Proponents say building the center above the atmosphere allows for near-constant solar energy harvesting, while eliminating the cooling burden that accounts for much of the cost of ground facilities.

However, experts warn that it will take many years for this idea to be commercialized, citing risks from space debris, cosmic radiation, maintenance difficulties and launch costs. Deutsche Bank expects the first small-scale deployments could take place in 2027-2028, helping to test both the technology and costs. If successful, larger systems, consisting of hundreds or thousands of satellites, could appear in the 2030s.