AI Stops Being Virtual When Meta Starts Buying Power From Space
AI Stops Being Virtual When Meta Starts Buying Power From Space
News summary: SpaceNews reported April 27 that Meta signed an agreement with Overview Energy, a startup developing space-based solar power systems, to transmit up to one gigawatt of power for Meta data centers. Overview says the power would come from satellites collecting solar energy in geosynchronous orbit and beaming it to ground infrastructure with near-infrared wide-beam lasers. The company is targeting an in-space demonstration in 2028 and commercial service as soon as 2030. Meta framed the deal as part of a broader effort to diversify energy sources for AI data centers, alongside geothermal, nuclear and battery-storage agreements. The political point is not that orbital solar is guaranteed to work on that schedule. It is that the AI buildout has already moved from software hype into hard infrastructure competition: power contracts, siting fights, grid constraints and higher stakes for ordinary ratepayers.
Commentary
The most important sentence in this story is not the futuristic one about satellites beaming power back to Earth. It is the boring one: Meta needs more electricity for data centers, and the normal power system is not giving the company enough certainty fast enough. That is where the AI story becomes political.
For most people, artificial intelligence still sounds like a software product. It arrives as a chatbot, a search result, an ad tool, a workplace assistant or a feature buried inside an app. But the real AI race is becoming a competition for physical inputs: land, water, chips, transmission lines, substations, batteries, nuclear plants, geothermal wells and now, apparently, orbit. When a platform company starts signing power deals that sound like science fiction, the signal is simple. The abundance story has run into a scarcity bottleneck.
That does not mean space-based solar is fake. It means the incentives around it deserve scrutiny. Overview Energy may ultimately prove that beaming power to existing solar facilities can extend the usefulness of terrestrial infrastructure after dark. Meta may be correctly placing a portfolio bet across several energy sources because no single grid solution can keep up with the compute curve. But citizens should notice what is being normalized. The biggest technology companies are no longer just buying electricity from the grid. They are trying to shape the future grid around their own demand schedule.
That matters because power is not a private abstraction. Transmission capacity is shared. Local opposition to data centers is already tied to water use, land use, noise, tax abatements and household power bills. If AI companies get priority access to scarce clean power, someone else absorbs the constraint. It may show up as higher utility bills, slower industrial projects, emergency rate cases, or political pressure to subsidize infrastructure in the name of national competitiveness.
The media version of this story will lean into the wow factor: lasers from space, solar at night, satellites as power plants. The more useful frame is the incentive map. A company that believes AI demand will compound over years cannot leave electricity to quarterly politics and local utility commissions. It wants optionality before the public argument has fully caught up. That is rational from Meta’s perspective. It is also exactly why the public needs the argument now, not after the contracts are locked in and the costs are socialized.
There is a two-to-three-year horizon hiding inside the headline. Overview is aiming for a 2028 demonstration and possible service by 2030. That overlaps with the period when AI data centers, grid upgrades and ratepayer fights are likely to become much more visible. The question is not whether every space-power promise lands on time. The question is why the hyperscalers are acting as if ordinary energy procurement is no longer enough.
This is the first inning of AI’s physical infrastructure politics. Compute is not virtual. Power is not infinite. And when the largest platforms start reaching beyond the existing grid, voters should ask who gets the upside, who gets the reliability, and who gets the bill.