Tech Billionaires Set Sights on Space to Power the AI Revolution
As artificial intelligence demands surge, tech giants are increasingly looking beyond Earth to address the escalating energy needs of massive data centers. The idea of server farms in space, once relegated to science fiction, is rapidly gaining traction, with companies actively pursuing this ambitious vision. But is building a digital infrastructure in orbit a realistic solution, or a costly and complex fantasy?
The current trajectory of AI growth is placing unprecedented strain on global electricity grids. A single Microsoft AI data center, slated to begin operations in 2026, is projected to consume as much power as the entire city of Zurich. This demand is onyl expected to grow as AI permeates more aspects of daily life, with 800 million people using ChatGPT weekly and millions more experimenting with AI-powered video applications and web browsers. Companies like OpenAI, Meta, Amazon, Google, and Elon Musk’s XAI are all investing heavily in expanding their data center capacity, leading some experts to question the sustainability of this growth.
The “Hyperscaler” challenge
These tech giants, frequently enough referred to as “hyperscalers,” are embarking on an unprecedented expansion of their IT infrastructure to manage the immense data generated by their AI services. Jeff Bezos believes that within the next decade or two,these “AI factories” will be constructed not on Earth,but in space. OpenAI CEO Sam Altman shares this vision, and Eric Schmidt, former head of Google and a prominent investor, has even acquired a rocket startup specifically to launch AI supercomputers into orbit. Elon Musk, thru SpaceX, is also actively pursuing the development of orbital data centers.
However, not all experts are convinced that Earth will run out of power for AI. Some advocate for increased reliance on nuclear energy, while others question the fundamental benefits of AI and its long-term commercial viability. Despite these reservations, AI optimists maintain that the technology’s growth is unstoppable and that harnessing the abundant solar energy available in space will become essential in the long run.
Philip Johnston, CEO of the startup Starcloud, predicts a dramatic increase in electricity demand. “In the next three years alone,AI will require the electricity output of ten new nuclear power plants,” he stated. This rapid expansion is already contributing to rising electricity prices in the United States.
Starcloud: Pioneering the Orbital Data Center
Starcloud is aiming to build an orbital data center powered by a massive solar power plant by approximately 2035. The project will require multiple rocket launches to deliver the servers, solar panels, and robots necessary for assembly and maintenance in orbit. The proposed data center is expected to require 5 gigawatts of electricity – equivalent to the output of five commercial nuclear power plants.
To maximize energy capture, the satellites will need to maintain constant alignment with the sun, orbiting in a way that prevents them from falling into Earth’s shadow.A significant engineering challenge lies in cooling the computers, as AI processors generate considerable heat and the vacuum of space lacks the air needed for conventional heat dissipation. Starcloud plans to address this with a massive radiator designed to radiate excess heat into the surrounding environment. According to Johnston, this radiator will be ten times lighter and a hundred times cheaper than existing technologies, though specific technical details remain closely guarded trade secrets.
Another critical concern is protecting the processors from cosmic radiation and micrometeoroids.Starcloud plans to use advanced shielding materials and redundant systems to mitigate these risks.
The Race to Space
While Starcloud is at the forefront, a race to establish this infrastructure is already underway. Experts like Malcolm Macdonald, a space expert at the University of Strathclyde, caution against expecting immediate profitability. Though, he acknowledges that SpaceX’s Starship coudl change the economic equation.
Space researcher Michael gschweitl from ETH Zurich believes that data centers in space are feasible, noting that small-scale space computers are already well-developed for analyzing data from Earth observation satellites. He emphasizes that building a 5-gigawatt data center like the one Starcloud envisions is physically possible, but presents a significant engineering challenge. Gschweitl predicts that such a project is more likely to be undertaken by a large, established space company rather than a startup.
The challenges are immense,but the potential rewards – access to abundant solar energy and a solution to the growing energy demands of AI – are driving innovation and investment in this bold new frontier.
