Google plans to launch a satellite carrying a single artificial intelligence (AI) server into orbit on Thursday, marking an initial step toward establishing data centers in space, a concept supported by leading technology figures but met with skepticism by some experts. The satellite, developed in collaboration with SpaceX and Planet Labs, houses a server equipped with four specialized tensor processing units (TPUs) — the type of chips Google uses to power its AI applications such as chatbots.

The initiative aims to eventually deploy multiple interconnected servers in a sun-synchronous orbit, where the satellites can continuously harvest solar energy without Earth blocking sunlight. By operating in space, these data centers could theoretically alleviate constraints on land availability and environmental impacts associated with large terrestrial data centers, which tech companies have heavily invested in to support expanding AI workloads.

Blaise Agüera y Arcas, Google’s vice president who spearheaded the project, has highlighted concerns over the environmental costs and limited capacity on Earth to meet the surging demand for electricity and real estate required by conventional data centers. “If you look at the exponent — how this is going to go over the next decade or two — it starts to look alarming,” Agüera y Arcas said.

Nonetheless, numerous technical and economic challenges remain. Launching hardware capable of withstanding the intense forces of rocket ascent, along with the harsh space environment, poses significant risks to the satellites’ delicate computer chips. Outside Earth’s atmosphere, radiation can disrupt computations critical to AI functions, requiring robust solutions to ensure reliability.

Thermal management presents an additional difficulty. While space is cold, the vacuum environment prevents conventional cooling methods based on air or liquid convection from being used. Instead, engineers must rely on radiative cooling techniques, where heat is emitted as electromagnetic radiation — a method that poses design complexities, particularly since the servers must remain operational while continuously exposed to solar radiation.

Despite these hurdles, Google and other prominent technology leaders, including SpaceX’s Elon Musk, Blue Origin’s Jeff Bezos, OpenAI’s Sam Altman, and Nvidia’s Jensen Huang, envision space-based AI infrastructure as a potentially cost-effective way to power future technologies. Musk recently predicted that operational AI in space could become economically viable within five years.

However, experts caution that broad deployment of space data centers remains a long-term prospect. Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University, emphasized that this is not a matter of simply lifting traditional data centers into orbit, but rather deploying smaller, highly efficient satellites networked together.

Astronomers have also raised concerns about the growing satellite population interfering with observational research and adding to the problem of orbital debris. Although Google intends to deorbit satellites at the end of their operational life to burn up in Earth’s atmosphere, scientists note that remnants can still persist and potentially contaminate the upper atmosphere.

Google acknowledges that significant scientific and engineering advances are needed before space-based data centers can become practical or economically feasible. James Manyika, Google’s senior vice president for research, said the timeline for large-scale deployment could extend a decade or longer. The upcoming satellite launch is therefore seen principally as a critical experiment in testing the viability of operating AI workloads beyond Earth.