As the power demands of terrestrial data centers approach physical limits, tech giants have realized that the next trillion-dollar computing power goldmine is shifting from crowded power grids to the silent orbits of space. This once science-fiction concept has recently become a market focus due to concentrated statements and strategic moves by heavyweight figures such as SpaceX founder Elon Musk, Amazon founder Jeff Bezos, and NVIDIA CEO Jensen Huang. According to a deep research report released on December 25 by an analyst team including Zhou Tianle from Guotai Haitong Securities Industrial Research Center, space-based computing is not merely about launching servers into space, but represents a paradigm shift from "space sensing, terrestrial computing" to "space sensing, space computing." Faced with the dual rigid constraints of surging terrestrial power demands and cooling difficulties, leveraging space's boundless solar energy and natural cooling environment has become a key solution to break through the computing power bottleneck. In the latest industry developments, this enthusiasm has translated into concrete action. Google plans to utilize its TPU architecture to build distributed satellite clusters, while startup Starcloud announced it successfully trained a large language model on a satellite equipped with NVIDIA GPUs. The logic behind this trend is not just a technological vision but also reflects a reshaping of capital expenditure expectations: rather than grappling with increasingly high electricity costs and regulatory resistance on Earth, it is more advantageous to utilize the resource benefits of space.
Driven by Physical Bottlenecks: Why Space? The expansion of terrestrial computing power is facing two major physical rigid constraints: energy and cooling. According to International Energy Agency statistics, global data center electricity consumption reached 415 TWh in 2024, and this figure is expected to double by 2030.
With the explosive growth in AI large model training demands, terrestrial power grid construction faces a "generational gap." The construction cycle for dispatchable green energy is long, making it difficult to match the rapid pace of AI demand. A Morgan Stanley report points out that the U.S. power gap for data centers could reach 20% in the coming years. Simultaneously, the cooling costs associated with high-density chips are substantial. The thermal flux density of next-generation chips like NVIDIA's GB200 continues to increase, pushing traditional air cooling to its limits. While liquid cooling technology offers improvements, it faces challenges related to water consumption and system complexity.
In contrast, the space environment offers a perfect solution. Space possesses a solar energy density as high as 1360 W/m², unaffected by day-night cycles or weather, enabling 24/7 continuous power supply. More critically, the cosmic background temperature is a mere 3K (approximately -270°C), providing an infinite "heat sink" for passive radiative cooling, achieving zero water consumption and zero energy cooling.
"Space's uniquely abundant solar energy can support 24-hour continuous power generation for orbital data centers, and the deep cold environment of space at -270°C is ideal for passive cooling, simultaneously resolving the two major terrestrial bottlenecks of energy and heat dissipation."
Furthermore, what truly captures the capital markets' attention is the vast cost differential between terrestrial and space operations. According to calculations in a Lumen Orbit white paper, the ten-year energy cost for a 40MW data center cluster is $140 million on Earth, but only $2 million in space. This fundamental change in cost structure gives space-based computing an overwhelming long-term economic advantage. In this regard, the energy cost ratio between terrestrial and space is approximately 70:1.
On Earth, cooling systems often entail massive water consumption and power waste; whereas in space, "passive radiative cooling technology is a zero-energy, zero-carbon emission passive cooling method that uses full-band infrared radiation to expel heat directly into the deep cosmos." Differentiated Exploration Led by Giants In the U.S. market, the development of space computing exhibits a distinct characteristic of being led by giants. The report notes: "The early exploration and capability building of space computing, led by global leaders, is gradually forming the basis for large-scale commercial diffusion."
Starcloud is pioneering "Orbital Computing as a Service." As a first mover, Starcloud clearly focuses on providing orbital AI computing services. Its test satellite, Starcloud-1, equipped with an NVIDIA H100 GPU, has successfully completed in-orbit training of lightweight large language models and verification of remote sensing image preprocessing. Its goal is to establish a 5GW space data center and build a 40MW-scale facility by 2030.
Google is extending from its cloud computing ecosystem. Its "Project Sun Catcher" is not just about launching satellites but involves plans to use its self-developed TPUs to build distributed satellite clusters, emphasizing software scheduling and inter-satellite networking. The report analyzes that Google aims to "define the future standard for space computing," replicating its vast cloud computing and AI ecosystem in orbit.
SpaceX plays the role of infrastructure foundation. Leveraging the Starlink constellation, SpaceX has built the world's only infrastructure with scalable orbital computing capacity. Although its current computing power is primarily used for internal services like inter-satellite link management and traffic scheduling, its high-power satellite platform and low-cost launch capabilities lay the physical foundation for future large-scale computing deployment.
A Vertically Integrated Industrial System The United States has already established a vertically integrated industrial system in the space computing field, spanning from underlying chips to top-level services, led by industry giants. At the chip layer, the U.S. has率先 achieved stable in-orbit operation of Commercial Off-The-Shelf AI chips. NVIDIA's Jetson series and HPE's Spaceborne Computer project have demonstrated that commercial GPUs can adapt to the space radiation environment with software redundancy and protective design. This allows the mature CUDA ecosystem and AI models from Earth to be directly migrated to orbit, creating a difficult-to-replicate hardware and software ecosystem barrier.
At the infrastructure layer, SpaceX, by controlling high-power satellite platforms, reusable launch systems, and mega-constellation networks, has solved the challenges of getting computing power "into space" and "networked." High-frequency, low-cost launch capabilities make deploying higher-power, heavier computing payloads an economically viable proposition.
Furthermore, the U.S. government, through risk-sharing mechanisms and diversified commercial demand, provides continuous funding and market support for industrial development.
The Chinese Path: Systematic Development Guided by National Strategy Unlike the U.S.'s commercial giant-led approach, China's space computing development exhibits a clear characteristic of being guided by national strategy, forming a dual-track pattern of "dedicated computing constellations + intelligent remote sensing constellations." Dedicated computing constellations aim to build pure space-based computing networks. Represented by the "Three-Body Computing Constellation," which successfully deployed its first 12 satellites into orbit in May 2025. Each satellite boasts a computing power of 744 TOPS and achieves full-orbit interconnection via 100Gbps laser links, equipped with a space-based distributed operating system designed to solve challenges in satellite-based high-performance computing and high-speed inter-satellite connectivity.
Intelligent remote sensing constellations represent the mainstream path for scaled application. Exemplified by the "Eastern Wisdom Eye" constellation, by loading intelligent processing units onto remote sensing satellites, it achieves "in-orbit perception, real-time analysis." For instance, in disaster monitoring, satellites can process data directly and downlink results, compressing 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 Spaceflight (2025-2027)," China is promoting the evolution of space computing from technological verification to systematic deployment through top-level design and regional industrial coordination.
Comments