Select Language:
A potential merger between Elon Musk’s SpaceX and xAI, reported exclusively by Reuters on Thursday, could energize Musk’s initiative to deploy satellite data centers into orbit. This move aligns with his broader strategy to lead in the rapidly expanding AI competition against dominant tech firms such as Alphabet’s Google, Meta, and OpenAI.
Here’s what we understand about AI computing via satellites:
What are space-based AI data centers?
Space-based data centers are still in the developmental stage but are envisioned to depend on hundreds of solar-powered satellites arranged in orbit to meet the massive processing needs of AI platforms like xAI’s Grok or OpenAI’s ChatGPT. These would be especially beneficial as terrestrial data centers face rising energy costs. Supporters argue that operating above Earth’s atmosphere provides near-constant solar energy and bypasses the extensive cooling requirements that drive costs for ground-based facilities, potentially making AI operations more efficient.
However, experts and space engineers warn that commercial viability is still years away. Concerns include the risks posed by space debris, shielding hardware from cosmic radiation, limited options for in-space maintenance, and the high costs associated with satellite launches. Deutsche Bank anticipates initial small-scale orbital data center tests around 2027-2028 to evaluate both the technology and the economics. If successful, larger satellite networks—possibly in the hundreds or thousands—could be developed in the 2030s.
Why is Musk pursuing this?
As the most successful rocket manufacturer historically, SpaceX has launched thousands of satellites as part of its Starlink broadband network. If satellite-based AI computing becomes the future, SpaceX is uniquely positioned to operate or facilitate the deployment of satellite clusters equipped for AI processing. Musk has indicated that space is the logical choice for data centers due to lower costs. At the World Economic Forum in Davos earlier this month, he stated, “Building solar-powered data centers in space is a no-brainer… within two to three years, space will be the cheapest place to run AI.”
SpaceX is rumored to consider an IPO this year, which could value the company—including its rockets and satellite infrastructure—at over $1 trillion. Some of the proceeds might be allocated toward developing AI-focused satellite data centers, according to sources.
What are Musk’s competitors doing?
Blue Origin, founded by Jeff Bezos, has been working on space-based AI data center technologies as well. Bezos predicts that “gigawatt-scale data centers” orbiting Earth could become cheaper than terrestrial counterparts within 10 to 20 years by harnessing abundant solar power and radiating excess heat into space.
Nvidia-backed Starcloud has already demonstrated this approach with its Starcloud-1 satellite, launched last month on a Falcon 9 rocket. The satellite is equipped with an Nvidia H100—claimed to be the most powerful AI chip deployed in space—and is training and operating Google’s open-source Gemma model. The company envisions creating a modular “hypercluster” of satellites capable of delivering around five gigawatts of computing power, akin to several massive data centers combined.
ChatGPT
Perplexity
Gemini AI
Grok AI
