Recently, discussions by the world's richest man, Elon Musk, about "space computing power" have reignited interest in this field.

Musk stated, "With the advent of Starship, the path to large-scale deployment of solar-powered AI satellites has finally been opened. This is, in my view, the only way to achieve the deployment of 1 terawatt (1TW) of AI computing power annually."

But what exactly is "space computing power"? What benefits would moving powerful computing capabilities to space bring?

Musk has repeatedly voiced his enthusiasm for space computing power. On October 30, he engaged in a discussion with a writer about the feasibility of "space data centers." The writer, Eric Berger, remarked, "There has been much debate about the feasibility of space data centers. While challenges exist, they seem more achievable if autonomous construction is possible." Musk responded, "Simply scale up the Starlink V3 satellites with high-speed laser links. SpaceX will do this."

On November 4, Musk further announced plans to expand the Starlink V3 satellite network to build space data centers, aiming to deploy 100GW of data center capacity annually via Starship within 4-5 years.

Notably, on November 2, U.S. space data center startup Starcloud successfully launched its first satellite, Starcloud-1, aboard a SpaceX rocket. The satellite is equipped with NVIDIA's H100 GPU and is set to run Google's open-source model, Gemma, in orbit—demonstrating that large language models can operate in space. If successful, Starcloud plans to construct larger-scale data centers in space, proposing a 5GW data center satellite powered by a 4-square-kilometer solar array. The company believes space data centers could reduce energy costs tenfold and lessen Earth's energy demands.

Meanwhile, Google announced its "Project Sun Catcher" on November 5, planning to launch two prototype satellites with Trillium TPUs by early 2027 to deploy AI computing power directly in space. A Google executive stated, "Space may become the optimal location for scaling AI computing in the future."

China has also made strides in space computing power. On May 14, the Long March 2D rocket launched a 12-satellite constellation—the world's first space computing satellite network—into a 700-kilometer sun-synchronous orbit. This marks humanity's first distributed computing network in low-Earth orbit, ushering in a new era of "space computing." The mission, centered on the "Star Computing" initiative, aims to shift from traditional "space sensing, ground computing" models to active, intelligent "space sensing, space computing" paradigms.

Wang Jian, an academician at the Chinese Academy of Engineering and director of Zhejiang Lab, emphasized that satellite interconnectivity is fundamental to building a computing constellation. Once interconnected, satellites' value is greatly amplified—reducing operational costs and improving data efficiency.

Li Chao, chief engineer of Zhejiang Lab's "Three-Body Computing Constellation," highlighted that traditional models discard over 90% of satellite data due to ground station limitations. The new approach processes data in space before transmitting results to Earth, enabling AI models to operate in orbit.

Huaxi Securities noted that space computing power's "low-cost, high-resource, large-scale" advantages offer a sustainable business model for commercial space ventures. The transition from "space sensing, ground computing" to "space sensing, space computing" is underway, with future plans envisioning a global space-based computing network.

According to Research and Markets, the global in-orbit data center market is projected to grow to $39.09 billion by 2035, with a 67.4% CAGR over a decade. As nations compete in energy efficiency, thermal management, radiation resistance, and laser-link technologies, computing satellites may become symbols of national space capabilities and digital sovereignty.