Tech Leaders Look to Space as a Solution for A.I. Data Center Limits on Earth

Some technology and space industry leaders say the growth of artificial intelligence could soon outstrip available land and energy on Earth, and they are proposing data centers in orbit as a solution. Google announced Project Suncatcher in November and plans test launches beginning in 2027.

Elon Musk has said space data centers could be the cheapest way to train A.I. within about five years. Supporters of the idea include Jeff Bezos, Sam Altman and Jensen Huang, among others. Investments in terrestrial data centers are already enormous, with major companies and nations spending billions; the article says OpenAI alone has committed $1.4 trillion to such projects.

At the same time, many land-based sites are running into power and community limits, with complaints that data centers raise utility bills and strain water supplies. Proponents argue space-based facilities offer near-constant sunlight for power, fewer regulations and fewer neighbors to oppose them.

Designs from startups such as Starcloud envision large satellite-like systems with server clusters surrounded by miles of solar panels. Experts disagree about timing and feasibility. Launch costs are currently high—about $8,000 per kilogram, or roughly $2,000 per kilogram on SpaceX at the cheapest rates—and proponents say prices must fall to about $200 per kilogram for the economics to work.

Key Topics

Tech, Google, Project Suncatcher, Elon Musk, Spacex, Openai