Data Center Construction Shifts To Space? Beijing Proposes Orbiting Computing Power As Aerospace Opens A New Narrative

The Battlefield Of Computing Power May Extend Beyond The Ground Into Space. On November 27, At The “Smartly Painting The Starry Sky, Winning Power In Space — Promotion Meeting On Space Data Center Construction,” It Was Announced That Beijing Plans To Build And Operate A Centralized Large‑Scale Data Center System Exceeding One Gigawatt (GW) In The 700–800 Kilometer Dawn‑Dusk Orbit, Aiming To Move Large‑Scale AI Computing Power Into Space.

According To The Planning Scheme Released At The Meeting, The Data Center System Will Comprise Space Computing, Relay Transmission, And Ground Control Subsystems.

Construction Will Proceed In Three Phases: From 2025 To 2027, Breakthroughs In Key Technologies Such As Energy And Cooling Will Be Achieved, Iterative Test Satellites Will Be Developed, And The First Computing Constellation Will Be Built; From 2028 To 2030, Breakthroughs In On‑Orbit Assembly And Construction Will Reduce Build And Operating Costs, Leading To The Second Computing Constellation; From 2031 To 2035, Satellites Will Enter Mass Production And Networked Launch, With On‑Orbit Docking To Form A Large‑Scale Space Data Center.

Space Computing Is A Technology That Deploys Data Centers And Computing Capacity Into Orbital Space, Using Satellites And Onboard Hardware For In‑Orbit Data Processing. High‑Speed Inter‑Satellite Laser Communications Enable Data Transmission And Real‑Time Processing, With Results Relayed Back To Earth.

With The Announcement Of Large Projects Such As “Stargate” And Continuous Upward Revisions To CSP Vendors’ Capital Expenditures, Explosive Growth In Computing Demand Has Become A Certainty. However, Power Generation, Transmission, And Cooling Requirements Have Long Constrained The Deployment Of Large AI Data Centers.

Unique Conditions In Space — Such As “Perpetual Solar Energy” And A Vacuum Environment — Theoretically Provide Energy And Cooling Advantages Unavailable To Ground‑Based Facilities.

At A Previous Saudi Investment Forum, Elon Musk Predicted: “Within Five Years, AI Computing In Space Will Be Cheaper Than On Earth.” He Argued This Would Be Enabled By “Free” Solar Energy And Easier Cooling Technologies: “In Space, You Can Use Continuous Solar Energy, You Don’t Actually Need Batteries Because Space Is Always Sunny, And Solar Panels Will Be Cheaper Since You Don’t Need Glass Or Frames. Cooling Is Simply Radiative Cooling.”

Earlier This Month, Google Proposed The “Suncatcher” Project, Planning To Launch Two Prototype Satellites Equipped With Trillium‑Generation TPUs In Early 2027, Directly Deploying AI Computing Power Into Space.

In Addition, Amazon Founder Jeff Bezos And Alphabet CEO Sundar Pichai Both Detailed Plans To Build Data Centers In Orbit. Bezos Predicted That Within 10 To 20 Years, Humanity Will Be Able To Construct Gigawatt‑Level Data Centers In Space.

Space Computing Projects Are Already Underway Internationally, With Nvidia Participating. U.S. Startup StarCloud Plans To Build The First Gigawatt‑Level Space Data Center For In‑Orbit AI Computing, Supported By Nvidia’s Startup Accelerator Program. Canadian Photovoltaic And Energy Storage Company PowerBank And Singapore Aerospace Firm Orbit AI Will Launch A Satellite Named DeStarlink Genesis‑1 In December To Validate The Deployment Of Nvidia Computing Chips In Low Earth Orbit.

Domestically, China’s First Fully Interconnected Space Computing Satellite Constellation Has Officially Entered The Networking Stage, Marking The First Launch Of The “Three‑Body Computing Constellation” Led By Zhijiang Laboratory. Beijing Xingkong Institute Orbital Chengguang Completed Its First Financing Round, With The Core Task Of Deploying Computing Satellites In The Dawn‑Dusk Orbit To Form A Space Data Center. ADA Space And Zhejiang Lab Jointly Launched The “Three‑Body Computing Constellation,” Which Has Already Sent The First Batch Of 12 AI Satellites Into Orbit. Guoxing Aerospace Initiated The “Star Computing Plan” In November 2024 And Successfully Launched The World’s First Space Computing Constellation In May 2025, Achieving In‑Orbit Cluster Computing Power Of 5 POPS.

The Current Push By Industry Giants To Move Computing Power Into Space Is Inseparable From Advances In Aerospace. Progress In Reusable Rocket Technology Has Greatly Reduced Launch Costs, Transforming Space Computing From “Economically Unfeasible” To “Commercially Calculable.”

A Large Data Center Cannot Be Built In A Single Launch; Instead, Like Building Blocks, Multiple Launches Must Deliver Modules — Computing Units, Energy Systems, Cooling Devices — Into Orbit For Subsequent Assembly.

Previously, Launching A Single Satellite Could Cost Hundreds Of Millions Or Even Billions Of U.S. Dollars, And Sending Multi‑Ton Data Center Modules Into Orbit Rendered Projects Economically Nonviable. With SpaceX And Others Reducing Launch Costs By Tens Of Times Through Rocket Recovery, Investors And Companies Are Seriously Evaluating Whether Long‑Term Returns From Space Data Centers Can Cover Total Costs.

For Nations, If Space Computing Proves Feasible, Orbital Paths Will Become “Mining Zones” For Computing Power, And Competition Over Chips Will Escalate Into “Orbital Resource Competition.”

Technically, Data Transmission And Chip Radiation Resistance Still Face Tests. A Single Satellite Can Generate Terabytes Of Data Per Day, But Limited Ground Station Coverage And Spectrum Bandwidth Often Restrict Actual Downlinked Data To Less Than 10%.

Even In Geosynchronous Orbit, Constructing Large AI Data Centers Faces Severe Challenges: Megawatt‑Scale GPU Clusters Require Enormous Radiative Cooling Wings To Rely Solely On Infrared Emission. This Implies Each Gigawatt‑Level System Would Need Tens Of Thousands Of Square Meters Of Deployable Structures — Far Beyond Any Spacecraft Achievements To Date.