Meta wants to power its AI data centers with energy beamed from space
Meta has agreed with Overview Energy to reserve up to 1 GW of power that satellites will collect in orbit and beam to Earth at night. The first demonstration launch is planned for 2028, with commercial delivery in 2030. For the company, it is part of a broader race for electricity for AI data centers, where conventional renewables and batteries are no longer enough. *Meta has been designated an extremist organization and is banned in Russia.
AI-processed from Bloomberg Tech; edited by Hamidun News
Meta is ready to buy more than just new solar generation: it wants energy that satellites collect in orbit and beam to Earth at night. The company has reserved up to 1 GW of capacity from Overview Energy. It is a bet that the electricity shortage facing AI data centers will have to be closed not only with nuclear, gas and ground-based renewables, but also with technologies that still sit on the border between engineering and experiment.
The scheme Meta and Overview are promoting works like this: satellites in geosynchronous orbit above the equator receive almost continuous sunlight, convert it and send it to Earth as low-intensity near-infrared radiation. Existing solar power plants are supposed to receive this stream. The idea is to use them not only by day but also at night, without building new large sites and without separate heavy grid infrastructure.
If the approach works, the capacity utilization of already-built solar facilities would rise sharply. The deal covers up to 1 GW, comparable to the output of a large nuclear reactor. Overview Energy plans its first orbital demonstration launch for 2028, and commercial delivery of energy to the US grid could, under the current plan, begin in 2030.
The financial terms of the partnership have not been disclosed. Overview itself emerged from stealth only in late 2025 and has since positioned itself as the developer of a system that does not build dedicated ground receivers for microwaves or high-power lasers, but instead tries to use existing solar infrastructure as the endpoint for receiving energy. For Meta, this is not an exotic experiment for headlines but a continuation of a hard race for electricity for AI.
According to the company and industry estimates, its data centers consumed more than 18,000 GWh of electricity in 2024, roughly what 1.7 million American households use in a year. Against that backdrop, Meta has already contracted more than 30 GW of clean and renewable generation, as well as 7.
7 GW of nuclear power through agreements with several suppliers. On the same day, the company separately announced a reservation of up to 1 GW / 100 GWh of ultra-long-duration energy storage from Noon Energy, with a pilot project of 25 MW / 2.5 GWh by 2028.
The reason for the interest is simple: standard solar and wind generation matches poorly with the operating profile of AI infrastructure, which needs power 24/7. Batteries help, but at hyperscaler scale they are expensive, slow and require a lot of land and materials. New nuclear and geothermal projects are more reliable, but they take years to build and run into permitting.
Space-based solar looks like a third path: it promises round-the-clock energy delivery and a faster link to existing solar plants. But it remains an early technology with a long list of open questions, from the efficiency of the whole chain and capital costs to regulation, orbital logistics and the impact of additional satellites on the space environment. What this means: the main problem for AI is no longer just access to chips, but access to guaranteed electricity in gigantic volumes.
Reserving capacity from Overview Energy does not mean that Meta will actually receive space-based electricity at industrial scale within a few years. But it shows that the largest tech companies are willing to fund even pre-commercial energy technologies if they shorten the path to new gigawatts. In the coming years, competitive advantage will increasingly be determined not only by model quality, but by who first secures a reliable energy base for training and running it.
*Meta has been designated an extremist organization and is banned in Russia.
- Nov 2, 2026Meta's Muse assistant keeps dossiers on loved ones and gets root on Ubuntu by default
- Nov 2, 2026The Verge reviewer builds a $50 Meta Muse Charm alternative with Claude Code
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
- Oct 31, 2026Open-source LLM gateway KTMS 1.0 plugs your own GPUs into the cloud and survives outages
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.