The trillion-dollar AI data center boom is spawning a new frontier: outer space. Tech billionaires Elon Musk and Jeff Bezos, after years of rivalry in rockets and satellites, are now extending their competition to orbital AI data centers, aiming to relocate massive computing infrastructure beyond Earth.

On December 10, sources revealed that Musk’s SpaceX plans to use upgraded Starlink satellites to handle AI computing workloads, pitching the technology as part of its equity offering, potentially valuing the company at $800 billion. Meanwhile, Bezos’ Blue Origin has reportedly assembled a team to develop orbital AI data center technology for over a year.

The race isn’t limited to these titans. OpenAI CEO Sam Altman has explored acquiring a rocket operator to deploy AI computing in space, while tech giants like Alphabet are also making moves. This trend signals a potential convergence of AI and aerospace, driven by the need to address Earth’s escalating energy demands for AI.

Despite enthusiasm for limitless solar energy in space, orbital data centers face significant engineering hurdles and cost-efficiency scrutiny. Skeptics argue the technical risks are underestimated, making near-term competition with terrestrial facilities unlikely. For investors, this represents both a visionary opportunity and high uncertainty.

**Earth "Too Small" for AI Ambitions: Space as the Energy Solution** The core rationale for space-based data centers is overcoming Earth’s physical constraints, particularly the staggering power needs of AI training and inference. Proponents envision satellites packed with AI chips, powered by uninterrupted solar energy, transmitting processed data back to Earth.

"The idea of moving resource-intensive infrastructure off-planet isn’t new, but it requires lower launch and satellite costs. We’re nearing that tipping point," said Will Marshall, CEO of satellite operator Planet Labs.

Veteran tech investor Gavin Baker notes that space offers inherent advantages in energy and cooling. Satellites can harness 30% more intense and up to six times more abundant solar energy than Earth, eliminating costly battery storage. Near-absolute-zero temperatures in space also enable "free" cooling, bypassing complex terrestrial systems. Baker predicts orbital data centers could become a pivotal breakthrough within 3–4 years.

**Giants Clash: Musk’s Starship vs. Bezos’ New Glenn** Launch capability is decisive in this space race. Both Musk and Bezos are leveraging next-gen heavy-lift rockets to realize their visions.

SpaceX’s AI computing tech will reportedly be integrated into upgraded satellites designed for its fully reusable Starship rocket, which Musk claims could annually deploy 300–500 GW of solar-powered AI satellites. Meanwhile, Blue Origin’s partially reusable New Glenn rocket, with its massive payload capacity, aims to dominate bulk satellite deployments. Bezos has argued orbital data centers could outcompete Earth-based AI infrastructure in cost within two decades.

"It all comes down to launch capability," emphasized Jonny Dyer, CEO of Muon Space. Companies with reliable, low-cost, high-capacity launch systems will lead this emerging race.

**New Entrants: Altman and Alphabet’s Plays** Beyond Musk and Bezos, other tech leaders are eyeing orbital data centers. Altman reportedly considered acquiring a rocket operator to deploy AI compute in space, citing AI’s eventual power demands. Alphabet has partnered with Planet Labs to launch two test satellites with custom AI chips by early 2027, though scaling to match terrestrial data centers would require tens of thousands of satellites.

Former Alphabet CEO Eric Schmidt and startups like Aetherflux and Starcloud are also advancing plans, while IBM’s Red Hat and Axiom Space launched a computing prototype in August.

**Challenges: Cost, Tech, and Scale** Despite the promise, orbital data centers must overcome formidable barriers. Technical hurdles include temperature control, cosmic radiation shielding, and low-latency data transmission. Critics argue proponents underestimate these risks and question cost competitiveness, especially if terrestrial power constraints ease.

Baker adds that while inference tasks may find clearer applications in space, migrating full-scale AI training workloads will take far longer. The path to space-based AI compute remains arduous and uncertain.