SpaceX Orbital Data Centre Bid Faces a $51bn Cost Gap Against Terrestrial Alternatives
Musk's 1-million-satellite AI computing plan confronts prohibitive economics and near-zero regulatory approval prospects, leaving European carbon demand tied to ground-based data centre growth.
SpaceX has proposed launching up to 1 million satellites to create a solar-powered computing network in low-Earth orbit, positioning the project as infrastructure for artificial intelligence data centres operating entirely outside the terrestrial grid. The scale has no precedent, and for European energy markets currently assessing how much grid capacity and carbon headroom AI infrastructure will consume over the next decade, the proposal is too speculative to price in.6,5
For traders in ICE EUA Dec-rolling allowances, the route from Musk's orbital plan to European power demand runs through economics that are stark. AI data centres are a growing driver of European gas-fired generation during peak periods. A significant migration of computing load off-planet would reduce one of the few demand-side growth stories currently supporting EUA prices. The numbers do not suggest that happens soon.1,3
The Economist put the cost gap plainly. Building a 1GW data centre on Earth and running it for five years costs $15.9bn. An equivalent orbital facility, under optimistic assumptions (a launch cost of $500 per kilogram, specific power of 37 watts per kilogram, a satellite cost of $22 per watt, and orbits keeping satellites in sunlight 98% of the time) would cost $51.1bn. More than three times the terrestrial figure. Those totals exclude the cost of beaming power back to Earth.1
SpaceX's current commercial pricing makes even those assumptions generous. The Falcon Heavy delivers payloads at around $1,500 per kilogram; the Falcon 9 runs $3,400 per kilogram, the Economist reported. Getting to $500 per kilogram requires cost reductions that have not materialised.1
Views on the timeline diverge sharply. Musk has said the concept could be feasible "within two years, maybe three at the latest." Sam Altman of OpenAI called it "ridiculous...we are not there yet." Google plans to test the concept but has published no deployment timeline.1
Regulatory approval is a separate problem. Foreign Policy reported that the prospect of international bodies approving 1 million new low-Earth orbit satellites is effectively zero.5
None of this changes the immediate pressure on European grids. Lex Coors, president of the European Data Centre Association, told the EU in Brussels (2026-06-22) that Europe must prioritise AI infrastructure over its near-term climate ambitions or risk conceding technology sovereignty to China. That is the dilemma European carbon policy is navigating: not the removal of data centre load from the grid, but its rapid and contentious addition.3
Ground-based data centre delivery is already slipping. Sightline Climate estimated that 30 to 50% of global capacity due to come online this year could be delayed, up from 26% in 2025. Those delays compress near-term power demand growth and reduce gas-to-power dispatch across European hubs, with corresponding effects on EUA demand. Orbital computing contributes nothing to that dynamic.1
SpaceX has been raising capital aggressively regardless. Its Nasdaq IPO (2026-06-17) valued the company at $2.1 trillion at the end of its first trading day. The company subsequently began selling investment-grade bonds for the first time, with Bank of America, Citigroup, Goldman Sachs, JPMorgan and Morgan Stanley among the underwriters, OilPrice.com reported. Analysts at Oppenheimer & Co expected debt to be the primary funding source, supplemented by roughly $40bn of additional equity, per a note dated (2026-06-18).4,2
The European Commission's response to the data centre lobby's push to deprioritise climate commitments for AI connectivity is the more proximate catalyst for carbon markets. That response is expected within months. The pace at which new data centre load gets connected to European grids will follow from that decision — shaping gas-fired dispatch and EUA demand through the rest of the decade far more directly than any satellite launch schedule.3,1