SpaceX unveiled this week the factory that will build its orbital data centers — an eleven-million-square-foot facility it has nicknamed the Gigasat, intended to produce more than a thousand satellites a year by late 2027. The satellites are the point. Each is a compute node with a wingspan of seventy meters, wider than a Boeing 747, drawing a hundred and fifty kilowatts of power from the constant sunlight of space, and the company’s filed plan calls for up to one million of them — a constellation of orbital AI processors aiming at a hundred gigawatts of compute capacity by 2030, powered entirely by the sun, beyond the reach of any grid, any zoning board, any community that might object. The buildout that pitched tents in Ohio fields because the Earth could not power it has now drawn its logical conclusion. If the planet cannot supply the compute, the compute will leave the planet.
The Logic of Leaving
The reasoning, once stated, is difficult to refute on its own terms, which is what makes it unsettling. The binding constraint on the buildout is power, and power on Earth is scarce, slow to build, contested by the communities that must host its generation, and interrupted by night and weather. In orbit, at the right altitude, the sun never sets; the solar power is constant, free of the grid, free of the ratepayer revolt, free of the zoning board and the environmental review and the neighbors who do not want a jet engine running outside their homes. A satellite in sunlight is a power plant with no community to object and no night to interrupt it. If the problem is that the Earth cannot supply enough uninterrupted power without a fight, the orbital solution is not insane. It is the elimination of the fight by the elimination of the Earth.
This is the same constraint I have traced through every physical form it has taken — the tents and jet engines in the Ohio fields, the reach for orbital solar, the migration to France’s reactors — followed now to its terminus. Each prior response tried to find more power on Earth: build faster, burn fuel, go where the grid is strongest. The satellites abandon the premise. They concede that the Earth-bound search for power is a losing race against the compute’s appetite, and they relocate the entire operation to the one place where power is unlimited and unowned. It is the most extreme answer the buildout has produced to the question that has driven all the others, and it is extreme not because it is irrational but because the rational answer to “the Earth cannot power this” turns out to be “then we will not power it on Earth.”
The scale is the part that should arrest you, because a million is not a number the mind processes as real. The current orbital population, after seventy years of spaceflight by every nation combined, is in the low tens of thousands of active satellites. SpaceX proposes, alone, to multiply that by a factor of many tens — to place in orbit more machines than humanity has launched in its entire history, several times over, each one wider than an airliner, in pursuit of compute for a technology whose demand is assumed to keep doubling. The plan does not describe an addition to the orbital environment. It describes the replacement of the orbital environment with an industrial one, owned substantially by a single company, for the purpose of feeding a single appetite.
The Commons Above
The debris problem is the one the company waves away, and it is the one that should not be waved away, because orbit is a commons and commons are destroyed in a particular, well-understood way. Low orbit has a carrying capacity; beyond a certain density of objects, the probability of collision rises, and each collision produces fragments that raise the probability further, until a cascade of collisions can render entire orbital bands unusable for generations — a runaway the physicists named decades ago and the regulators have never seriously prepared for. A million satellites is a density no one has modeled at scale, proposed by a company that points to its collision-avoidance software as reassurance, as though the answer to multiplying the orbital population by an order of magnitude were a better autopilot.
And the deeper pattern is the privatization of a commons that belongs, by treaty and by nature, to everyone. Orbit is not American, not corporate; it is the shared sky above all nations, governed by agreements written when satellites were rare and a single company placing a million of them was unimaginable. SpaceX is proposing to occupy a vast share of that commons for its own industrial purpose, ahead of any framework capable of governing the occupation, on the reliable principle that it is easier to ask forgiveness than permission and that the facts on the ground — or in this case the facts in the sky — will be established before the rules to constrain them exist. The grid said no on Earth, so the buildout is enclosing the sky, where no one has yet been empowered to say no at all.
There is a grim elegance to the choice of orbit, because it escapes every form of resistance this record has documented at once. The ratepayer cannot be charged for a power plant in space. The community cannot block a data center it cannot see. The zoning board has no jurisdiction over the constant sun. The environmental review does not reach the vacuum. Every mechanism by which the Earth-bound public has slowed or contested the buildout — the bill, the veto, the moratorium, the protest — depends on the buildout being somewhere a public can reach it, and a satellite is nowhere a public can reach. The migration to orbit is not only a search for power. It is a flight from accountability, to the one place where the people affected by the technology have no standing to object to it.
What This Means
The plan may not be built at the scale announced; a million is a filing, not a fleet, and the history of grand orbital ambitions is mostly a history of grand orbital ambitions. The economics of launching, maintaining, and cooling that much hardware in space are formidable, and the constellation that actually flies may be a fraction of the one on paper. But the direction is the signal, independent of the final count, and the direction is unmistakable: the buildout has concluded that the Earth is a constraint to be escaped rather than a home to be accommodated, and it has identified the sky as the next territory to enclose. Even a tenth of a million satellites would be the largest industrial occupation of orbit in history, undertaken by a private company, ahead of any law.
What the satellites make visible is the absolute priority the buildout has assigned to compute over every other consideration, including the integrity of the shared environment, terrestrial and celestial alike. The technology is treated as an end so overriding that the Earth’s grids, the Earth’s water, the Earth’s communities, and now the Earth’s orbit are all merely obstacles between the buildout and the power it needs — to be engineered around, charged to others, or left behind. A civilization that will fill the sky with a million machines rather than slow the growth of its compute has revealed where, in its hierarchy of values, the compute actually sits, which is above the sky.
I would observe only what the night sky will become if this is built, because it is the kind of consequence that arrives without anyone having chosen it. For the whole of human history, a person who looked up saw the same stars their ancestors saw, the one view that no industry had touched, the commons that belonged to everyone precisely because no one could enclose it. A million reflective satellites in low orbit would end that — would streak the dark with moving lights, scatter the telescopes’ view of the cosmos, and convert the oldest shared inheritance of the species into the infrastructure of a single company’s compute. No one will have voted for it. It will simply appear, one launch at a time, until the night that every generation before us looked up into is gone, replaced by the glittering exhaust of a machine that needed somewhere to think, and chose the sky.