WASHINGTON, Jan 29 - A proposed merger between Elon Musk’s SpaceX and xAI, reported exclusively on Thursday, could accelerate Musk’s long‑running pitch to move large portions of AI computing off Earth and into orbit. The idea has attracted renewed scrutiny as major technology players weigh how to meet soaring AI compute demand.
Space‑based data centers remain an experimental concept rather than an established industry. The design under discussion envisions hundreds of satellites, each powered by solar arrays, networked together in orbit to shoulder the intensive computing workloads generated by AI models like xAI’s Grok or OpenAI’s ChatGPT. Backers argue operating above the atmosphere would provide nearly continuous solar energy and sharply reduce the cooling challenges that dominate the operating expense of terrestrial data centers.
These potential advantages speak directly to core cost drivers for data‑center operators. On the ground, large AI models require vast amounts of electricity and sophisticated cooling systems; those two categories account for a meaningful portion of total cost of ownership. In orbit, advocates contend, steady access to sunlight and the ability to radiate heat into space could change these economics.
At the same time, engineers and space specialists emphasize significant technical and commercial obstacles remain. The challenges cited include the proliferation of space debris, protecting delicate electronics from cosmic radiation, the practical limits on performing in‑person repairs once hardware is on orbit, and the expense of launches. Those constraints mean commercial viability is not imminent, and would require both technological advances and favorable economics to be realized at scale.
Deutsche Bank projects that initial, small‑scale orbital data center demonstrations could occur in 2027–28 to test both technology and business models. If those missions validate performance and cost assumptions, wider constellations - potentially numbering in the hundreds or thousands - could follow in the 2030s, according to that assessment.
Why would Musk pursue this? SpaceX is the most active commercial launcher and satellite operator today, having placed thousands of satellites in orbit for its Starlink internet service. From an asset and capability perspective, SpaceX is well positioned to field the rocket lift and on‑orbit deployment capacity an orbital compute system would require. Musk has argued the economics are clear: "It’s a no‑brainer building solar‑power data centers in space ... the lowest‑cost place to put AI will be space, and that will be true within two years, three at the latest," he said at the World Economic Forum in Davos earlier this month.
There is also a financial angle: SpaceX is reportedly weighing an initial public offering this year that could place the company's valuation above $1 trillion. Part of any proceeds, according to reporting, might be directed toward developing satellite systems designed to host AI compute.
Other major industry players are pursuing parallel approaches. Jeff Bezos’s Blue Origin has explored concepts for orbiting AI compute resources, advancing the view that large gigawatt‑class data centers in space could, over time, undercut the unit costs of Earth‑bound facilities through uninterrupted solar power and more efficient heat rejection to space.
Commercial demonstrations are already under way. Nvidia‑backed Starcloud launched a satellite called Starcloud‑1 on a Falcon 9 last month that carries an Nvidia H100 chip - described as the most powerful AI processor yet placed in orbit - and is running Google’s open‑source Gemma model as a proof of concept. The firm envisions a modular "hypercluster" made up of many satellites, collectively delivering around five gigawatts of computing capacity - a level comparable to several hyperscale terrestrial data centers.
Google is advancing a related research program called Project Suncatcher, which aims to link solar‑powered satellites fitted with its Tensor Processing Units into an orbital AI cloud. Google plans an initial prototype launch with partner Planet Labs around 2027.
State actors are also mobilizing. China has set out plans to establish a "Space Cloud," deploying space‑based artificial intelligence data centers over the next five years, according to state media. The country’s principal space contractor, China Aerospace Science and Technology Corporation, has pledged to build gigawatt‑class space digital‑intelligence infrastructure as part of a five‑year development plan.
The concept remains a long‑horizon bet. Early missions will need to demonstrate that orbital systems can deliver reliable compute at a cost that challenges terrestrial incumbents, while also addressing the unique operational and regulatory risks of on‑orbit infrastructure. For investors and corporate planners, the trajectory will be shaped by the outcome of those first demonstrations, the pace of launch‑cost reductions, and progress in safeguarding hardware against the space environment.
Summary
The proposed SpaceX‑xAI merger has brought fresh focus to efforts to host AI compute in orbit. Proponents point to near‑constant solar power and reduced cooling needs as potential cost advantages; critics note the technical, operational and economic hurdles that still need to be solved. Early test deployments are expected in the late 2020s, with large‑scale constellations contingent on those results.