SpaceX's Orbital Data Centers Threaten to Reshape Nuclear Industry's AI Boom

The content of this article does not represent the views, policies, opinions, positions, or official statements of the U.S. Navy, the Department of Defense, the federal government, or any military or government organization.

WASHINGTON, When SpaceX filed an application with the Federal Communications Commission on Jan. 30, 2026, proposing up to one million solar-powered satellites to serve as orbital data centers, it raised a question the nuclear energy industry has been reluctant to answer out loud: if artificial intelligence infrastructure leaves Earth, what happens to the nuclear demand that was supposed to rescue it?

The question is not theoretical. SpaceX unveiled the first hardware in the orbital data center program, a satellite called AI1, in June 2026. The vehicle carries a 70-meter solar array generating 150 kilowatts of peak power and relies on the vacuum of space for cooling, according to company descriptions and Data Center Dynamics reporting.

The timing disrupts a narrative that had been building for two years. Big technology companies, facing severe power constraints on the ground, had begun signing deals with nuclear startups at a pace the industry had not seen in decades. Microsoft restarted operations at Three Mile Island. Meta committed to a 1.2-gigawatt nuclear campus in Ohio with Oklo. OpenAI's Stargate project outlined plans to pair data centers with small modular reactor baseload power.

Now, with the orbital alternative entering hardware development, the question is whether those commitments hold.

The Terrestrial Power Crisis

The pressure driving both the nuclear deals and the orbital computing concept stems from the same source: AI infrastructure is consuming electricity faster than the grid can supply it.

Goldman Sachs projected U.S. data center power demand will double from 31 gigawatts in 2025 to 66 gigawatts by 2027. In the first quarter of 2026 alone, an estimated $130 billion in data center projects were blocked or delayed by power permitting constraints, according to industry analyses reported by SpaceNews and Data Center Dynamics.

The Electric Power Research Institute projects data centers could account for nine percent of U.S. electricity generation by 2030, up from an estimated four percent today.

Small modular reactors, compact nuclear plants sited near demand centers, emerged as a candidate solution. Big tech collectively committed to more than 10 gigawatts of new nuclear capacity. Oklo, NuScale, X-Energy, and TerraPower all saw deal pipelines expand, according to IAEA and IDTechEx reporting.

The power crisis on the ground is also what made orbital computing begin to look attractive. If sufficient power cannot be secured on Earth, and low Earth orbit offers near-constant solar with passive cooling, the physics of the alternative become worth examining.

What SpaceX Is Proposing

The AI1 satellite design is more credible than most prior orbital computing proposals because SpaceX has already demonstrated the underlying technologies.

The company's reusable Falcon 9 rocket provides launch at costs no current competitor matches. The Starlink constellation, serving approximately 3.7 million subscribers in more than 70 countries, already operates the laser inter-satellite communication links that orbital data centers would require. The February 2026 merger with xAI, creating a combined entity called SpaceXAI valued at $1.25 trillion, placed AI model development under the same corporate structure as launch and satellite operations, according to Futurum Group and KeepTrack reporting.

Each AI1 satellite carries machine-learning accelerators. Thermal management relies on radiative heat dissipation into space rather than the liquid cooling and air-handling systems that consume significant power in ground-based facilities.

Market analysts tracking the sector estimate the orbital data center market, projected at approximately $1.77 billion in value by 2029, could expand to $39 billion by 2035 at a 67.4 percent compound annual growth rate, according to industry forecasts. SpaceX indicated prototype AI1 satellites could launch in early 2027.

Two Scenarios for the Nuclear Industry

The impact on nuclear demand depends on which of two paths the orbital computing story takes.

In the first scenario, orbital computing reaches meaningful scale by 2028 or 2029. A share of AI training and inference workload migrates off the terrestrial grid. Technology companies that had planned nuclear deals for data center baseload power reduce or defer those commitments. Companies oriented primarily toward large utility projects would face more exposure. Companies targeting smaller, more diversified applications may be less affected.

In the second scenario, orbital computing encounters delays or cost overruns. Sam Altman, chief executive of OpenAI, called orbital data centers "ridiculous" in June 2026 and said they "are not something that's going to matter at scale this decade," according to TechCrunch and Fox Business. Thermal management in orbit, limited to radiative dissipation with no atmospheric convection, presents real engineering constraints at scale. If orbital systems slip past 2030, the terrestrial power crisis does not wait. Nuclear demand would intensify, not diminish.

The Motley Fool noted in a July 13, 2026 analysis that SpaceX's orbital data center moves were "something NuScale Power investors should be watching", capturing the uncertainty without resolving it.

The Case for Complementarity

Critics of a zero-sum framing point out that orbital and nuclear power are not necessarily in competition.

For deep space missions, polar orbits with extended eclipse periods, and lunar surface operations, nuclear radioisotope power systems have no viable solar alternative. NASA has relied on them for decades. If AI workloads eventually operate in those environments, nuclear power becomes an enabler of space computing, not a competitor to it.

Ground infrastructure supporting orbital computing, satellite manufacturing, launch facilities, mission control, ground stations, all require terrestrial power now, not in 2035.

The nuclear industry's real vulnerability, analysts suggest, is not orbital data centers specifically. It is the risk that any credible alternative reduces the urgency of nuclear buildout long enough to dissipate the regulatory and financial momentum the sector has built. Commercial small modular reactor timelines in North America remain long. The window of favorable market conditions may not remain indefinitely open.

What Leaders Need to Know

For those working in energy policy, infrastructure investment, technology strategy, and national security, the orbital data center story demands attention regardless of its speculative dimensions.

SpaceX does not file FCC applications as a theoretical exercise. The AI1 hardware exists. The xAI merger has closed. The financial incentive, bypassing a terrestrial power bottleneck that actively blocked $130 billion in data center investment in a single quarter, is substantial, according to SpaceNews reporting.

The decisions being made now by nuclear developers, technology companies, regulators, and capital markets are being made under genuine uncertainty about which infrastructure path prevails. That uncertainty is the normal condition for decisions that matter. It makes getting the analysis right more important, not less.

---

Sources: FCC filing DA-26-113A1; SpaceNews; Data Center Dynamics; CNBC (June 21, 2026); IAEA Bulletin; IDTechEx; The Motley Fool (July 13, 2026); Carbon Credits; TechCrunch (July 13, 2026); Fox Business. This article is based on publicly available reporting, regulatory filings, and industry analysis. Nothing here constitutes investment advice.