Orbital Data Centres: The Honest Case

The debate about putting data centres in orbit has split cleanly along political lines. On one side: reckless billionaire hubris, poisoning the last commons for server racks. On the other: government-hating environmentalists blocking the future of American computing. Both are wrong in the same way. They started from a conclusion and worked backwards. Here is what the physics, the environmental science, and the geopolitics actually say.

"The question is not whether orbital compute has a footprint. Everything has a footprint. The question is whether it is smaller than the footprint of not doing it, and whether we are honest about both sides of that ledger."

---

The Physics Is Genuinely Compelling

Start with what is not in dispute. The sun delivers 1,361 watts per square metre to a surface in low Earth orbit. No clouds, no atmosphere absorbing 30%, no night in a sun-synchronous orbit. On the ground, a well-sited solar farm produces an average of 150 to 200 watts per square metre after accounting for weather, angle, and the rotation of the Earth. That is not a marginal advantage for orbital solar. It is a structural one, and no amount of political opinion changes the underlying physics.

SpaceX's AI1 satellite, revealed in June 2026, targets 150 kilowatts of compute power from a 70-metre wingspan solar array. To match that power output from rooftop solar on the ground requires roughly 750 to 1,000 square metres of panels and a cooperative sun. Meanwhile a single large terrestrial hyperscale data centre burns between three and five million gallons of water per day for cooling, consumes hundreds of acres of land, takes years to permit, and draws from a grid that is already struggling to keep pace with AI demand in most of the developed world. Orbital compute uses no water. It clears no land. The heat rejection is a radiator panel facing a 2.7 Kelvin vacuum. These are real advantages, not marketing claims.

The AI compute problem is also real. Training frontier models already costs tens of millions of dollars in electricity. The next generation will cost more. The terrestrial grid is not growing fast enough to absorb that demand, particularly in the regions where data centre density is already highest. Orbital compute does not ask permission from a utility company. The energy source has been running for five billion years.

---

What the Environmental Critics Get Right

The critics who say this technology has no atmospheric footprint are wrong. It does. NOAA's SABRE mission flew high-altitude measurement flights and found aerospace alloy signatures in more than 10% of stratospheric particles sampled, at a time when roughly 8,700 satellites were in orbit. The same research projects that number rising above 50% if 50,000 more satellites are added. These are real measurements, not projections.

The reentry chemistry is more complex than the aluminium headline suggests. SABRE detected more than 20 distinct elements in those stratospheric particles. The unstudied materials riding along with the aluminium airframe are a genuine concern at scale:

• Aluminium oxide (alumina) — the most studied; at approved constellation scale (roughly 60,000 satellites), projected at 18 times the natural meteoric baseline. Peer-reviewed models put ozone depletion risk anywhere from negligible to 9%. That range spans "nothing" to "something real" and nobody knows where it lands.
• Cobalt — from battery cathodes; already entering the stratosphere at 3.5 tonnes per year, equivalent to around 350 EV battery packs annually; stratospheric chemistry essentially unstudied
• Arsenic — from gallium arsenide solar cells; likely recondenses as arsenic trioxide; nobody has modelled what that does in the stratosphere at scale because until now nobody needed to
• Carbon soot — from carbon fibre structural panels; unlike car exhaust soot which washes out in days, stratospheric soot persists for years and has roughly 540 times the climate impact per gram; projected to triple by 2029 even on current constellation plans

The structural problem is the 20 to 30 year lag. Material injected into the mesosphere during reentry takes decades to drift down to ozone-relevant stratospheric altitudes. The consequences of today's launch cadence will not be fully felt until the mid-2040s at the earliest. Decisions made now about scale and design are banking a debt whose invoice arrives after most of the infrastructure is already built.

---

What the Environmental Critics Get Wrong

The atmospheric effects at current scale are genuinely trace-level. This is not a defence of the technology at full buildout. It is a statement of fact about where we are now. Ninety-three tonnes of material entered the stratosphere from satellite reentries in 2024. The stratosphere is very large. The concentrations are detectable by sensitive instruments and essentially undetectable by any biological system. The SABRE findings are scientifically significant as a baseline measurement and not currently a public health concern.

The physical shadowing argument fails completely. At full buildout of every constellation currently filed with the FCC, roughly 1.24 million satellites, the combined solar panel area is approximately 744 square kilometres. Against Earth's 127.5 million square kilometre sunlight interception disk, that is 0.00058%, six parts per million. The threshold where atmospheric scientists start caring about sunlight reduction is around 0.1%, a level associated with major volcanic eruptions. Orbital compute at any plausible scale does not approach that threshold.

The comparison must also be honest. Terrestrial data centres are not clean. A large hyperscale facility emits between 100,000 and 500,000 tonnes of CO2 equivalent per year depending on its grid. The global data centre sector consumes 200 to 250 terawatt-hours of electricity annually, growing at 15 to 20% per year, most of it from grids that are not yet fully renewable. US data centres withdraw over 600 billion litres of cooling water per year. If the environmental argument against orbital compute is serious, it has to engage with what the terrestrial alternative actually costs. That comparison is not as clear-cut as either side pretends.

---

What the Boosters Get Wrong

The pro-orbital argument has three significant holes that its proponents tend to wave past.

Launch carbon is not in the accounting. Every satellite going up requires a rocket. Even Starship, the most efficient heavy launch vehicle ever built, burns hundreds of tonnes of methane per flight. A constellation with a 5 to 7 year satellite lifespan requires continuous replacement launches indefinitely. Nobody has published a full lifecycle carbon analysis for orbital data centres that includes manufacturing, launch, operations, reentry, and replacement. That analysis might still favour orbital compute. It might not. The honest position is that we do not know yet, and the boosters are presenting half the ledger.

The scale contradiction is real. The pro-orbital argument says current atmospheric effects are trace-level, which is true. It also says we should build a million satellites to unlock unlimited compute. You cannot have both without acknowledging that the atmospheric effects scale with the fleet. At 60,000 satellites the alumina input is modelled at 18 times natural. At one million satellites there is no peer-reviewed model, but linear extrapolation puts it around 8,000 times natural. "The effects are small now" and "we should build a million of these" require the same sentence to complete honestly: "and those effects will not be small then."

The Kessler cascade risk is not paranoid delusion. At a million satellites distributed across the 500 to 800 kilometre orbital band, a single collision cascade could render those shells unusable for decades. GPS, weather satellites, military reconnaissance, and all future orbital infrastructure depend on those shells remaining navigable. A Kessler event is not recoverable on any human timescale. It is a tail risk with a civilisational-scale downside, and the pro-orbital camp dismisses it too quickly. The risk does not make the technology wrong. It makes the question of how we manage orbital density genuinely important in a way that deserves more than a footnote.

---

The Geopolitics, Honestly

China's Guowang programme is a state-backed LEO constellation with filed capacity for 12,992 satellites. It is real, it is funded, and it is not subject to the planning objections, environmental reviews, or competitive pressures that shape Western commercial space development. The argument that the US and its allies should slow down orbital compute development on atmospheric grounds while China builds without those constraints is not an environmental argument. It is a unilateral disarmament argument in scientific language.

That said, the answer to "China is building" is not "build a million satellites with no monitoring, no debris management, and no atmospheric science." The geopolitical case for speed does not override the engineering case for doing it in a way that does not destroy the orbital environment that every country, including the US military, depends on. Both things are true simultaneously.

---

What Doing It Right Actually Looks Like

Orbital data centres are probably worth building. The physics is too good and the terrestrial constraint is too real. But "worth building" is not the same as "build recklessly at million-satellite scale before the science catches up." The gap between those two positions is where the actual policy work needs to happen.

The history of transformative infrastructure is not a history of technologies with no downsides. It is a history of technologies whose benefits were real enough that humanity decided the downsides were worth managing. Cars reshaped cities in ways that took decades to understand and are still being undone. Aviation created contrail-driven climate effects that nobody modelled in 1950. Nuclear power has waste problems that remain unsolved 70 years in. In each case, the technology was worth building and the failure was not building it but in assuming the downsides would sort themselves out without deliberate management.

Orbital compute is in that tradition. The physics is right. The need is real. The environmental concerns at current scale are genuinely manageable. The concerns at million-satellite scale are not yet characterised well enough to dismiss. That is not a reason to stop. It is a reason to build the monitoring, the design standards, the debris management, and the international frameworks in parallel with the satellites, rather than deciding they can wait until after the infrastructure is already in orbit and the decisions are irreversible.

"Build it. Study it while you build. Fix what breaks. Do not lock in a million-satellite world before you understand what a million-satellite world does."