Tech giants have a space problem. Literally. The explosion of AI models like Gemini and Llama needs more computing power than earth’s grids can reliably supply. Power-hungry data centres now account for about 1% of global electricity use—roughly as much as the whole country of Sweden. That’s why companies like Google and SpaceX are eyeing a radical fix: moving servers into orbit.

Inside Google’s X lab in Mountain View, California, engineers have quietly tested concepts for satellite-cooled data centres. According to three people familiar with the project, the team is exploring how to use SpaceX’s Starlink satellite network to beam data up and down. The goal? Cut the energy drain from cooling massive server farms, which can swallow up to 40% of a data centre’s power budget.

SpaceX isn’t just listening—it’s pitching in. Elon Musk confirmed last month that Starlink’s lasers are being tested for high-speed data links between satellites, a key step for orbiting data centres. These links could slash latency below 20 milliseconds, fast enough for real-time AI workloads like robotics or autonomous driving.

How orbital data centres would work

The basic idea isn’t new. NASA and the military have used satellites for decades to process data in space. But today’s AI models need thousands of chips packed tightly together—impossible to launch in one rocket. So Google’s team is designing modular, stackable server racks that can be sent up in batches and assembled by robots in low Earth orbit.

Cooling is the trickiest part. On Earth, companies use giant water-cooled towers or immersion tanks filled with mineral oil. In space, there’s no air, so heat radiates away naturally—but slowly. Google’s solution? Heat pipes that transfer warmth to satellite radiators, then beam it down as infrared energy. Early tests show this could cut cooling costs by up to 90% compared to ground data centres.

The catch: radiation and maintenance

Space isn’t kind to electronics. Cosmic rays and solar flares can fry unshielded chips. Google’s engineers are testing radiation-hardened chips from companies like NVIDIA and AMD, but costs are sky-high. A single hardened AI accelerator can cost 10 times more than its Earth-bound cousin.

Then there’s maintenance. No astronauts will swap out fried hard drives in orbit. The plan hinges on robot swarms that dock with server satellites for repairs. Amazon’s Project Kuiper has already tested robotic servicing arms on its prototype satellites, giving Google a head start.

Who else is looking up?

Google isn’t alone. Microsoft has partnered with Loon (yes, the defunct balloon internet company) to study high-altitude data centres. Meanwhile, IBM is testing liquid-metal cooling for satellites that could handle AI workloads. Even NASA is rumored to be exploring orbital supercomputing for Mars missions.

The race isn’t just about power. Orbital data centres could slash latency for global AI services. Today, a query sent to a data centre in Iowa takes 50 milliseconds to reach a user in Tokyo. A satellite link could cut that to 20 milliseconds—fast enough for real-time robotics or augmented reality.

What happens next

Don’t expect a SpaceX rocket carrying Google servers next week. The project is still in the whiteboard phase, with no firm timeline. But the energy crisis is accelerating the timeline. Last year, data centres in Virginia’s Loudoun County, the so-called “Data Centre Alley,” drew more power than the entire country of Kenya.

Regulators are starting to notice. The European Commission just proposed rules forcing tech giants to disclose energy use from AI workloads. Meanwhile, California is considering a moratorium on new data centres until they prove they can run on 100% renewable energy.

If orbital data centres work, they could rewrite the rules of the internet. But first, Google and SpaceX have to solve radiation, maintenance, and the little problem of launching thousands of tons of hardware into space.

For now, it’s all still a thought experiment—one that might just save the planet from melting down under the weight of its own AI dreams.

What You Need to Know

  • Source: France 24
  • Published: May 17, 2026 at 13:37 UTC
  • Category: World
  • Topics: #france24 · #world-news · #europe · #science · #space · #are-google

Read the Full Story

This is a curated summary. For the complete article, original data, quotes and full analysis:

Read the full story on France 24 →

All reporting rights belong to the respective author(s) at France 24. GlobalBR News summarizes publicly available content to help readers discover the most relevant global news.

Curated by GlobalBR News · May 17, 2026

🇧🇷 Resumo em Português

A corrida tecnológica para driblar os limites da Terra já chegou ao espaço: gigantes como Google e SpaceX investigam a possibilidade de instalar data centers em órbita para alimentar os supercomputadores de inteligência artificial, uma solução que promete não só reduzir custos, mas também o calor gerado por essas estruturas colossais.

A ideia, ainda em fase de testes e viabilidade, busca contornar o gargalo energético que afeta os data centers terrestres, responsáveis por até 1% do consumo global de eletricidade. No Brasil, onde a matriz energética é majoritariamente limpa, mas a demanda por armazenamento de dados cresce exponencialmente — especialmente com a expansão do 5G e da computação em nuvem —, a proposta soa como uma alternativa futurista, mas com implicações práticas imediatas. Além disso, a redução na emissão de calor poderia aliviar a pressão sobre sistemas de refrigeração cada vez mais caros e poluentes, um problema que afeta desde São Paulo até as regiões mais remotas do país.

Se viável, a solução poderia redefinir a geopolítica da tecnologia, colocando nações com capacidade espacial — como os EUA e a China — em vantagem, enquanto o Brasil teria de acelerar investimentos em pesquisa para não ficar para trás nessa nova fronteira.

🇪🇸 Resumen en Español

La carrera por dominar el futuro digital lleva a gigantes tecnológicos como Google y SpaceX a mirar más allá de la atmósfera, donde los centros de datos flotantes podrían ser la solución a los desafíos energéticos de la inteligencia artificial. Ambas compañías analizan seriamente la posibilidad de instalar servidores en órbita, un proyecto que, de materializarse, revolucionaría no solo la computación en nube, sino también la forma en que el mundo consume energía.

La idea, aunque aún en fase experimental, responde a un problema acuciante: los centros de datos tradicionales, esenciales para el desarrollo de la IA, consumen cantidades ingentes de electricidad y generan un calor difícil de disipar, lo que tensiona las redes eléctricas globales. En un contexto donde la demanda de procesamiento de datos crece exponencialmente, especialmente en regiones hispanohablantes con economías emergentes, esta innovación podría aliviar la presión sobre infraestructuras locales y reducir costes a largo plazo. Además, al ubicarse en el espacio, se evitarían los límites físicos de la Tierra, abriendo la puerta a una nueva era de eficiencia energética, aunque no está exenta de desafíos técnicos y regulatorios que aún requieren solución.