SpaceX's Orbital Data Centers Just Crossed From Speculation to Strategy. Here's What Enterprise Tech Leaders Need to Know.

For years, orbital data centers were a thought experiment that surfaced at AI conferences and got dismissed by serious infrastructure people. That changed on June 19, when the Washington Post reported that SpaceX's space-based computing plans have moved from speculative to something enterprise technology leaders are now factoring into long-term planning, according to the Washington Post's reporting on the shift in industry sentiment.

The pitch is straightforward. Terrestrial data centers are running out of physical space, water, and grid capacity to support AI's compute demand. SpaceX argues the fix is to stop building on Earth at all.

What SpaceX Has Actually Built

According to Tom's Hardware, which reviewed Elon Musk's detailed video specification of the AI1 satellite, the design includes a 70-meter deployed wingspan, an average compute payload of 120 kW, a peak of 150 kW, and a density of 70 kW per ton. Musk compared one AI1 satellite's compute draw to a single Nvidia GB300 rack on the ground, describing the satellite as essentially one data center rack relocated into orbit.

Per the same Tom's Hardware report, the reveal of AI1's specifications came in the same week SpaceX's IPO priced and began trading publicly. The satellite design draws heavily on hardware SpaceX already manufactures for Starlink V3, which the company says lets it move from concept to production faster than a from-scratch space platform would allow.

According to Data Center Dynamics' coverage of the FCC filing, SpaceX filed plans with the Federal Communications Commission in January for a large-scale orbital data center satellite constellation, describing a system that would operate with what the filing called unprecedented computing capacity to power advanced AI models and the applications that rely on them.

Tom's Hardware also reported that SpaceX has signed at least one commercial compute deal tied to this infrastructure, naming Google as a counterparty, though the specific financial terms have not been independently confirmed.

Why Enterprise Buyers Are Starting to Pay Attention

The Washington Post framing matters here. Per that report, the significant development is not that SpaceX has solved orbital computing, but that credible people in the data center industry are now taking the proposal seriously rather than dismissing it outright.

The economic logic behind that shift is not abstract. According to analysis published by CarbonCredits.com citing International Energy Agency data, data centers consumed roughly 415 terawatt-hours of electricity globally in 2024, about 1.5% of total global electricity use, with demand growing approximately 12% annually over five years. Under aggressive growth scenarios referenced in that analysis, the figure could exceed 1,000 TWh by 2026. That demand curve is widely cited as the reason utilities and local communities have pushed back on new terrestrial data center construction over power draw and water use.

SpaceX is not alone in pursuing this thesis. According to MIT Technology Review's reporting, Jeff Bezos has predicted gigawatt-scale data centers in space within ten or more years, citing Blue Origin's capabilities, and former Google CEO Eric Schmidt acquired rocket company Relativity Space specifically to pursue orbital computing. The same report notes that Google has its own program targeting a test constellation of compute satellites, and that the startup Starcloud deployed an Nvidia H100-class system in orbit in 2025, becoming the first company to train a large language model and run a version of Google Gemini in space.

When multiple separate, well-capitalized players, including SpaceX, Blue Origin, Google, and Relativity Space, are independently pursuing the same infrastructure bet, enterprise buyers should treat the category as live rather than speculative, even where the underlying economics remain unproven.

The Skepticism Enterprise Buyers Should Weigh Equally

Not every credible voice agrees. Tom's Hardware reported that OpenAI CEO Sam Altman called orbital data centers "ridiculous" earlier this year, a notable dismissal from the leader of a company with significant stake in how AI compute capacity gets built next.

The technical objections are substantive. Per MIT Technology Review and the Wikipedia entry on space-based data centers, cooling in orbit is limited to radiative heat dissipation only, since there is no atmosphere for convection, making thermal management meaningfully harder than on Earth. Launch costs remain the dominant expense. A Google feasibility study, cited in the Wikipedia summary, concluded that orbital data centers only become cost-competitive with terrestrial alternatives if launch costs to low Earth orbit fall to roughly $200 per kilogram, a threshold the study projects might be reached around 2035, contingent on SpaceX's Starship scaling significantly in annual launch cadence by then.

Space sustainability experts cited by MIT Technology Review have also raised concern that a very large constellation of orbital compute satellites, replaced regularly to keep pace with chip generations, would meaningfully increase atmospheric reentry of debris compared to current rates.

What This Means for Enterprise Technology Strategy

None of this changes near-term infrastructure decisions. Orbital compute at meaningful enterprise scale remains years, possibly a decade, from commercial viability by every credible technical assessment cited above, including SpaceX's own cautious framing, reported by Yahoo Finance, that AI1 represents the company's first attempt and that no promises are being made about deployment timing or scale.

What has changed, according to the Washington Post's reporting, is the credibility threshold inside the industry. A year ago, orbital data centers were treated as a footnote in AI infrastructure conversations. Today, multiple well-funded companies are deploying capital against the thesis, one has working hardware already in orbit, and mainstream technology press coverage has shifted from questioning whether this is real to documenting that experienced infrastructure people are now taking it seriously. For enterprise technology leaders building five and ten-year infrastructure roadmaps, that shift in credibility within the industry is the relevant signal, regardless of where individual leaders land on the underlying engineering feasibility debate.

The terrestrial data center constraints driving this conversation, grid capacity, water availability, community opposition, and power costs, are well documented and are not resolving on their own. Whether the eventual answer is orbital compute, nuclear-powered terrestrial facilities, or an approach not yet proposed, the underlying pressure on AI infrastructure capacity is the strategic variable every enterprise technology buyer needs to be tracking now.