The ultimate battleground for computing power breakthroughs is not in Silicon Valley, but in outer space orbit.

```

When the power demands of ground data centers approach their physical limits, tech giants have realized that the next trillion-dollar computing gold mine has shifted from crowded power grids to the silent orbits of space.

This concept, once belonging to the realm of science fiction, has recently become a market focus due to intensive announcements and plans by heavyweight figures such as SpaceX founder Elon Musk, Amazon founder Jeff Bezos, and NVIDIA CEO Jensen Huang.

According to a deep research report published on December 25 by Zhou Tianle and other analysts of the Guotai Haitong Securities Industry Research Center, space computing power is not simply about sending servers into space, but about a paradigm shift from "sensing in space, computing on the ground" to "sensing and computing all in space". Facing the dual rigid constraints of surging power demand and cooling difficulties on the ground, leveraging the endless solar energy and natural cooling environment of space has become the key solution to the computing power bottleneck.

In the latest industry developments, this enthusiasm has already turned into practical action. Wallstreetcn reports that Google is planning to use its TPU architecture to build a distributed satellite cluster, while the startup Starcloud has announced the successful training of a space-based large language model on satellites equipped with NVIDIA GPUs.

The logic behind this trend is not only a technological vision, but also reflects a reshaping of capital expenditure expectations: instead of contending with ever-increasing power costs and regulatory resistance on the ground, it makes sense to exploit the resource advantages of space.

Driven by Physical Bottlenecks: Why Space?

The expansion of computing power on the ground is facing two major physical rigid constraints: energy and cooling.

According to the International Energy Agency (IEA), in 2024 global data centers will consume a total of 415 terawatt hours, with that number expected to double by 2030.

With the soaring demand for AI large model training, ground power grid construction is facing an “intergenerational gap”; green energy capable of being dispatched takes a long time to build, making it hard to match AI’s rapid growth. Reports from Morgan Stanley suggest that in the coming years the power shortfall for U.S. data centers could reach 20%.

Meanwhile, high-density chips are leading to soaring cooling costs. New-generation chips such as NVIDIA's GB200 keep increasing their heat flux density; traditional air cooling is already at its limit, and while liquid cooling has improved matters, it faces challenges regarding water consumption and system complexity.

By contrast, the space environment provides a perfect solution. Space has intensive solar energy of up to 1360 W/m², unaffected by day or night or weather, enabling 24/7 continuous power supply. More critically, the cosmic background temperature is as low as 3K (about -270°C), providing an infinite “heat sink” for passive radiative cooling, achieving zero water consumption and zero energy consumption for heat dissipation.

“Space’s unique abundance of solar energy can support round-the-clock power for on-orbit data centers, and the -270°C deep cold space environment is ideal for passive heat dissipation, solving both energy supply and cooling bottlenecks faced on the ground.”

Moreover, what truly stirs capital markets is the enormous cost difference between ground and space.

According to Lumen Orbit’s white paper, a 40MW data center cluster's ten-year energy cost would be $140 million on the ground, but only $2 million in space (the cost of solar arrays). This fundamental change in cost structure gives space computing a crushing long-term economic advantage. In this respect, the energy cost ratio of ground to space is about 70 to 1.

On earth, cooling systems typically mean huge water consumption and power waste; but in space, “passive radiative cooling technology is a zero-energy, zero-carbon passive cooling method that uses full-band infrared radiation to send heat directly into deep space.”

Giant-led Differentiated Exploration

In the U.S. market, the development of space computing power shows a clear pattern of giant-led initiatives. The report notes: “Early exploration and capability building of space computing by global leaders is gradually forming large-scale commercial diffusion.”

Starcloud is pioneering “computing power as a service in orbit.”

As a pioneer, Starcloud is focused on providing on-orbit AI computing services. Its test satellite Starcloud-1 is equipped with NVIDIA’s H100 GPU, and has completed on-orbit lightweight large language model training and remote sensing image pre-processing verification. Its goal is to build a 5GW space data center and to complete a 40MW-scale facility by 2030.

Google is extending from its cloud computing architecture.

Its "Sun Catcher" project is not just satellite launches; it aims to use self-developed TPUs to build distributed satellite clusters, with a focus on software scheduling and inter-satellite networking. The report argues that Google is aiming to “define future standards for space computing,” migrating its massive cloud computing and AI ecosystem into orbit.

SpaceX plays the role of infrastructure base.

Relying on the Starlink constellation, SpaceX has built the world’s only infrastructure capable of carrying large-scale computing power in orbit. Currently, this computing power is mainly used for internal services like inter-satellite link management and traffic scheduling, but its high-power satellite platform (Starlink V3) and low-cost launch capabilities (Falcon 9 and Starship) lay the physical foundation for future large-scale computing deployment.

Vertically Integrated Industrial System

The U.S. has already built a vertically integrated industrial system in space computing, from basic chips to top-level services, led by dominant companies.

At the chip level, the U.S. has taken the lead in achieving the stable on-orbit operation of commercial AI chips (COTS). NVIDIA’s Jetson series and HPE’s Spaceborne Computer project have proven that commercial GPUs, with redundancy and protection design, can adapt to the radiation environment of space. This allows mature CUDA ecosystems and AI models on the ground to be directly migrated into orbit, creating a hard-to-replicate hardware-software ecosystem barrier.

At the infrastructure level, SpaceX solves the problems of “sending computing to space” and “networking in orbit” through its mastery of high-power satellite platforms, reusable launch systems, and ultra-large-scale constellation networks. High-frequency, low-cost launch capabilities make deploying larger, higher-power computing payloads (such as server-class equipment) economically feasible.

Furthermore, the U.S. government provides ongoing financial and market support for industry development through risk-sharing mechanisms (like NASA’s procurement contracts) and diversified commercial demand (commercial remote sensing, cloud services).

China’s Approach: Systematic Development Under National Strategy

Unlike the U.S., where commercial giants take the lead, China’s development of space computing power demonstrates clear national strategic guidance, with a dual-track pattern of “dedicated computing constellations + intelligent remote sensing constellations”.

Dedicated computing constellations are aimed at building a pure space-based computing network. Represented by the “Three-Body Computing Constellation” project, the first group of 12 satellites was launched in May 2025. Each satellite provides up to 744 TOPS of computing power, achieves intra-orbit interconnection with 100 Gbps laser links, and is equipped with a space-based distributed operating system to tackle challenges in high-performance onboard computing and high-speed inter-satellite networking.

Intelligent remote sensing constellations are the mainstream path for large-scale applications. Taking the “Oriental Insight” constellation as an example, by loading intelligent processing units onto remote sensing satellites, “on-orbit sensing and real-time interpretation” is achieved. For example, in disaster monitoring, satellites can process data directly and deliver results, reducing response times from hours to minutes.

At the policy level, from the 14th Five-Year Plan to the "Action Plan for Promoting High-Quality and Safe Development of Commercial Space (2025-2027)", China is advancing the evolution of space computing from technical verification into systematic deployment through top-level design and local industry collaboration (for example, Beijing’s space data center construction plan).

Risk Warning and DisclaimerThe market involves risk, and investment should be done with caution. This article does not constitute personal investment advice and does not take into account the investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular situation. Investment is at your own risk. ```