Nature sub-journal: Building a data center in space with infinite energy. Zhejiang University confirms its feasibility for the first time.

A new study by Zhejiang University and Nanyang Technological University, Singapore aims to explore the feasibility of a space carbon-neutral data center. The space environment offers two unique advantages: abundant solar energy can provide clean and stable power for computing devices, and the deep-space environment close to absolute zero provides ideal heat dissipation conditions for server waste heat. We propose two implementation plans: one is to integrate AI accelerators on remote sensing satellites to build an "orbital edge data center" and process data directly at the source of collection; the other is to form a constellation of computing satellites to create an "orbital cloud data center", which has the ability to process space data and undertake ground computing tasks. At the same time, we have also established a full-life cycle carbon efficiency evaluation system for the space cloud data center.

Space technology and information technology are facing increasingly prominent sustainability pressures.

On the one hand, low Earth orbit is being rapidly occupied by large-scale satellite constellations. These satellites continuously generate massive amounts of "space-native data" in fields such as communication, remote sensing, and meteorological monitoring, with the scale reaching dozens of terabytes (TB) per satellite per day.

On the other hand, the rapid development of technologies such as artificial intelligence (AI) and high-performance computing (HPC) has driven the global wave of building energy-intensive data centers, leading to a sharp increase in their power consumption and carbon emission footprint.

The traditional "bent pipe" data processing model, which involves downloading all space data to ground data centers for processing, not only introduces significant communication delays, which is unfavorable for applications with high real-time requirements such as disaster emergency response, but more importantly, it further exacerbates the already heavy energy and environmental burdens of ground data centers.

Therefore, it is crucial to explore a new type of computing infrastructure architecture that can process data at the source and fundamentally reduce carbon emissions.

In response to the coexisting challenges of the explosive growth of low Earth orbit data and the high carbon footprint of ground data centers, Zhejiang University has for the first time systematically proposed a complete technical framework and evaluation system for building a carbon-neutral data center in space.

Paper link: https://www.nature.com/articles/s41928-025-01476-1

The feasibility of this technical framework is based on two sustainability advantages provided by the unique physical environment of space.

Firstly, there is nearly infinite solar energy resources. The solar energy density in Earth's orbit is much higher than that on the Earth's surface and is not severely affected by weather and day-night cycles. This enables high-efficiency solar panels to provide continuous, stable, and carbon-free power supply for high-power computing devices.

Secondly, there are extreme heat dissipation conditions. The deep-space background close to absolute zero (about -270°C) forms a natural giant heat sink, allowing the waste heat generated by computing devices to be efficiently radiated directly into the universe through radiative cooling technology. This completely eliminates the dependence on the complex, energy-consuming, and water-consuming cooling systems (such as cooling towers and chillers) of ground data centers, achieving "zero water consumption", "zero energy consumption", and "zero carbon emissions" for computing heat dissipation.

Core Framework

The core contribution of this research is the construction of a hierarchical "orbital data center" technical framework, which consists of two complementary components.

Orbital Edge Data Center

This framework aims to upgrade existing data collection satellites into computing nodes with on-board intelligent data processing capabilities.

Its technical approach is to highly integrate data sensors (such as high-resolution cameras), AI accelerators optimized for specific fields (such as VPUs, NPUs, GPUs, FPGAs, etc.), a sufficient scale of solar arrays, and an efficient thermal management subsystem for the space environment (such as a radiative cooler) on the satellite platform.

Its core value lies in achieving "on-orbit information extraction", that is, completing most of the computing and analysis tasks at the source of data generation, and only transmitting the screened and processed high-value, low-capacity information products (such as recognition results and characteristic parameters) down, thereby greatly alleviating the pressure on the scarce downlink bandwidth and minimizing data transmission delays.

Orbital Cloud Data Center

This framework envisions a more ambitious distributed orbital computing satellite constellation.

It consists of a constellation of computing satellites deployed in low Earth orbit. Each computing satellite is essentially a "space server" equipped with multiple high-performance general-purpose servers (including standard components such as CPUs, GPUs, memory, storage, and operating systems) and high-speed broadband communication capabilities.

Such an orbital cloud platform has dual service capabilities:

One is to serve as an aggregated computing power pool to process complex computing tasks (such as large-scale multi-source remote sensing data fusion analysis) from various orbital edge data centers that exceed their own processing capabilities.

The other is to serve as a carbon-neutral outsourced computing platform for ground computing tasks. Users can dynamically choose to offload computing tasks to the orbital cloud with lower carbon emissions or ground clouds in specific regions through a "carbon-aware" intelligent scheduling system, thereby optimizing the overall carbon efficiency.

Evaluation Innovation: Full-Life Cycle Carbon Utilization Efficiency Model Title

In order to scientifically and fairly evaluate the environmental benefits of the orbital data center and avoid the one-sidedness of only focusing on carbon emissions during the operation phase, the research team has innovatively proposed the evaluation index of "full-life cycle carbon utilization efficiency".

This model extends the evaluation boundary to the entire value chain, covering carbon emissions from all links such as the manufacturing of computing satellites and launch vehicles, the rocket launch process, on-orbit operation (with zero power carbon intensity), and end-of-life disposal.

Preliminary modeling analysis using this model shows that although the orbital data center will generate significant "one-time" carbon emissions during the manufacturing and launch phases, thanks to its continuous carbon-neutral advantage during on-orbit operation, from a full-life cycle perspective, its carbon efficiency is expected to exceed that of ground data centers relying on medium-carbon-intensity power grids and gradually approach the level of advanced ground data centers powered by all renewable energy sources.

Challenges and Outlook: Technological Breakthroughs and a New Paradigm for Future Green Computing Power

The research also objectively analyzes the technological and economic challenges in realizing this blueprint.

The primary technological bottleneck is the severe threat of the space radiation environment to the reliability of commercial high-performance servers. Future development depends on the maturity of radiation-hardened technologies dedicated to space data centers, which may increase the cost, complexity, and power consumption of the system.

Secondly, there is the economic feasibility challenge. Currently, the costs of satellite platforms, high-performance servers, and launch services remain high. Although the framework envisions that computing satellites have broadband communication capabilities to improve cost-effectiveness, the initial large capital investment is still the main obstacle to large-scale deployment.

Nevertheless, small-scale orbital edge computing has entered the technology verification and initial commercial exploration stage, showing a feasible technological path.

As a more transformative long-term goal, the development of the orbital cloud data center is expected to drive the coordinated breakthroughs of space energy, thermal control, communication, and high-performance computing technologies in extreme environments, laying a solid foundation for the ultimate construction of a next-generation green computing power infrastructure with global coverage, low latency, and environmental friendliness.

Summary

This perspective article published in Nature Electronics is significant in that it goes beyond the discussion of a single technology and for the first time systematically outlines the complete architecture blueprint of space-based computing infrastructure and establishes a corresponding full-life cycle environmental benefit evaluation methodology.

It not only provides an innovative solution for the efficient processing of the growing space data but also offers a forward-looking idea for fundamentally solving the high-carbon dilemma of ground data centers, pointing an important direction for the future development of sustainable computing technologies.

First Author

The first author and co-corresponding author, Ablimit Aili, Ph.D., is currently a distinguished researcher at the Yangtze River Delta Smart Oasis Innovation Center of Zhejiang University (since June 2024). He obtained his Ph.D. in Mechanical Engineering from the University of Colorado Boulder (2017–2021) and was later selected as a "President's Postdoctoral Fellow" at Nanyang Technological University, Singapore (2022–2024). He was included in the talent introduction program of Zhejiang Province in 2024. His research interests cover multiple fields such as engineering thermophysics, green data centers, green buildings, and smart microgrids.

Professor Yonggang Wen, the Associate Provost of Nanyang Technological University, serves as the co-corresponding author.

References

https://www.nature.com/articles/s41928-025-01476-1

This article is from the WeChat official account "New Intelligence Yuan", author: LRST. Republished by 36Kr with permission.