Politics on the Ground, Compute in Orbit
I’ve never been a futurist, much less a maximal one. But Elon Musk has shown me that, in the words of the late George Allen, the future is now.
Allen did not mean the distant horizon had arrived on schedule. He meant he refused to treat the present as a waiting room. He traded future draft picks for veterans who could win games immediately because he rejected the idea that the important results belonged to some later, more convenient decade. The motto was operational. It still is.
I have spent years treating artificial intelligence the same way I treat any other powerful tool: a gun, an axe, a shovel. The tool itself carries no moral charge. Outcomes track the character and incentives of the people who pick it up. Markets adapt. Work transforms. Human judgment and the capacity to integrate new capability still decide most of what follows. The loudest voices on both the apocalyptic and utopian ends of the spectrum over-claim. They always have.
That framing still holds. What has shifted is the speed at which the physical and political conditions surrounding the tool appear able to change.
Look at the fight over data centers. The resistance is real, bipartisan, and growing. Local governments propose or pass moratoriums. Candidates from both parties distance themselves. Polling shows sharp swings against facilities that, only a short time ago, many communities courted for the tax base and the construction jobs. The factual rebuttals are available: water use comparable to other industrial or agricultural loads, noise governed by ordinance, cost-causation rules that require the large users to pay for the upgrades they trigger, measurable fiscal returns in places that have hosted them. None of it seems to reverse the political weather. Perception and incentives are outrunning the evidence.
There is another side to the same story. The electricians and the trades who actually build and maintain these facilities see steady, high-wage work that lasts. In Northern Virginia the hours have doubled and then some. Apprenticeship pipelines have expanded. In Loudoun County the facilities supply a large share of property-tax revenue and return far more in taxes than they consume in services. Blanket opposition does not erase the demand for compute. It simply decides where the capacity gets built and under what standards.
While that terrestrial argument grinds on, one company is attempting to change the physical domain altogether. SpaceX is designing and preparing to launch satellites whose primary purpose is compute rather than communications. Large solar arrays. Radiators that reject heat into the vacuum. Laser links between the satellites and down to the existing Starlink network. The first-generation designs are already public. The regulatory filings seek authority for a constellation that could eventually number in the hundreds of thousands or more. The launch vehicle that makes the mass numbers possible is the same vehicle the company has been iterating in public for years.
Starlink is the precedent that makes the claim serious. The conventional industry treated mass-market low-Earth-orbit broadband as a capital-intensive graveyard. SpaceX treated the hard problems—phased-array terminals, high-volume satellite production, constellation management, laser inter-satellite links—as engineering tasks to be driven down in cost and up in reliability. The result is a functioning global system that generates real revenue. That track record does not guarantee the next step will succeed at the scale now being discussed. It does justify assigning the effort a higher probability than pure speculation would allow.
The compensation package Musk negotiated with the SpaceX board makes the time horizon explicit. A substantial portion of his personal upside vests only if the company reaches an extraordinary valuation and plants a permanent city of a million people on Mars, or if it operates orbital data centers drawing at least a hundred terawatts, more than a thousand times the electricity consumed by every data center on Earth today. That is not conventional executive incentive design. It ties personal wealth to outcomes measured in decades and in civilization-scale units. Most boards would never accept it. This one did, after watching the same person convert a series of claims that once looked impossible into operating systems.
The strategic context sharpens the stakes. Advanced chips still matter, and the United States continues to lead in the highest-end silicon. Energy and the ability to site and power large facilities are becoming equally decisive. China faces fewer domestic political constraints on adding generation and data-center capacity. The United States carries higher friction: permitting, interconnection queues, local opposition. If that asymmetry persists, the tool remains available but the volume and reliability with which Americans can apply it become constrained by our own political economy. Orbital capacity, even at partial scale, is one way to reduce dependence on those terrestrial bottlenecks. It leverages advantages the United States currently holds: heavy reusable lift, operational experience with large constellations, and continued access to the best chips. China is pursuing its own orbital computing efforts. They remain smaller and more focused on processing satellite data in orbit for now. The race is no longer only about models and algorithms. It is also about who can reliably deliver the energy and the physical substrate.
None of this requires conversion to maximal futurism. It requires noticing a pattern. In this particular set of hands, the distance between an ambitious engineering claim and an operating reality has repeatedly closed faster than conventional forecasting assumed. Reusable orbital rockets. A global broadband constellation. Now the attempt to place meaningful compute capacity in the same regime. The claims remain extreme. The record of execution is what compresses the timeline.
Artificial intelligence is still a tool. The character of the people and the structure of the incentives will continue to shape what it produces. The live question is the speed at which the conditions of its large-scale use can be rewritten by politics on the ground or by moving a share of the capacity off the planet altogether. The future arriving now is not a guarantee of any particular outcome. It is the refusal to treat the decisive capabilities as permanently deferred. Whether that refusal produces durable American advantage or merely expensive spectacle remains a matter of agency, engineering, and sustained will. The tool does not decide that. We do.
