The race to build orbital data centers is missing its biggest variable: power
Elon Musk has projected that orbital platforms will become the lowest‑cost environment for running artificial‑intelligence workloads within the next 36 months, while LoneStar disclosed plans for a lunar‑based data centre and NVIDIA highlighted its Vera Rubin Space‑1 demonstrator. The statements, made in recent industry briefings, converge on a shared premise: moving compute off‑planet could reshape the economics of AI services. All three announcements emerged in the same week of public statements, underscoring a rapid escalation of interest in space‑borne data processing.
The discussion of power requirements dominates the debate. McKinsey estimates $6.7 trillion in data‑center spending will be needed by 2030, with hyperscalers collectively allocating roughly $410 billion in 2025 and projecting $650‑$700 billion for 2026—a 60 percent year‑over‑year rise. The International Energy Agency warns that global data‑centre electricity use could double to 945 terawatt‑hours by 2030, matching Japan’s entire grid, while turbine manufacturers report orders filled through that year. Solar conversion on orbit can generate five times the output of comparable Earth‑based panels because of uninterrupted sunlight and the absence of atmospheric losses. Silicon cells have approached their 29‑30 percent Shockley‑Queisser ceiling, prompting companies such as Tandem PV, a Beyond Earth Ventures portfolio firm, to develop perovskite‑silicon tandem modules that combine record efficiency with space‑qualified durability. For shadowed or high‑inclination orbits, radioisotope power systems from Zeno Power—also backed by Beyond Earth—are being tested for seabed and lunar use, offering continuous, grid‑independent electricity. Fusion propulsion, pursued by NASA and DARPA, promises specific‑impulse values orders of magnitude higher than chemical rockets, potentially lowering the cost of delivering massive payloads to orbit and enabling larger power‑intensive infrastructures.
Industry stakeholders are being urged to treat energy as a foundational element rather than a downstream add‑on. Hyperscalers such as Microsoft, Google, Amazon, Alphabet and Meta have already signed multi‑gigawatt nuclear power‑purchase agreements and, in Google’s case, acquired renewable‑energy developer Intersect Power to secure solar and storage capacity. The article recommends that investors evaluate space portfolios for exposure to power‑hardware firms, and that data‑centre procurement processes incorporate dedicated requests for orbital power suppliers, mirroring terrestrial standards like ASHRAE. The viability of Musk’s 36‑month timeline will hinge on the readiness of these solar, nuclear and fusion power solutions, as the computing hardware appears set to arrive before the supporting energy infrastructure.
