Recently, Elon Musk stated on X (formerly Twitter) that SpaceX will scale up its Starlink V3 satellites and begin building data centers in space to address the growing demand for computing power in the AI era.Wait… is computing power really going to space?Today, let’s dive into the topic of “space-based computing power”, and as always, we’ll break it down using AlphaEngine.

1

Musk’s Bold Move: Scaling Up V3 Satellites for Space ComputingAs the demand for computing power surges with the rapid advancement of artificial intelligence, interest in space-based data centershas skyrocketed.In May of this year, former Google CEO Eric Schmidttook the helm as CEO of Relativity Space, positioning the company to explore space-based computing.In October, Jeff Bezos, founder of Amazon, publicly stated that within the next 10 to 20 years, the company plans to build gigawatt-level data centers in space.Just recently, following a report by tech media outlet Ars Technicaon the potential of autonomous assembly technology for constructing large data centers in orbit, Musk responded on X, suggesting that Starlink satellitescould be repurposed for this goal.He wrote:

“You can do it just by scaling up the Starlink V3 satellites. These satellites are equipped with high-speed laser links, and SpaceX will be pursuing this.”

Musk’s interest in space-based computing has significantly boosted attention to this emerging field.Currently, Starlink V2 Mini satelliteshave a maximum downlink capacity of about 100 Gbps, but V3 satellitesare expected to deliver 10 times that capacity—around 1 Tbps (1,000 Gbps).SpaceX plans to launch dozens of Starlink V3 satellites per mission using its Starshiprocket, with the first such launches potentially happening in the first half of 2026.

2

The Unique Value of Space-Based Data CentersA space-based data centerrefers to modular computing infrastructure deployed in Earth’s orbit. Essentially, it involves relocating traditional data centers from the ground to space.By integrating high-performance computing payloads, these centers enable “in-space computing”—processing massive volumes of data directly in orbit as it’s generated by satellites and other platforms. This approach fundamentally bypasses the physical limitations faced by terrestrial data centers, such as energy constraints and land availability.Faced with the staggering prediction that global AI Data Center (AIDC)power demand could reach 347 GW by 2030, space-based data centers offer unique advantages.

Energy Efficiency:

By deploying high-efficiency solar panel arrays, space-based centers can generate power directly from the sun—at 5 times the energy output per unit areacompared to Earth. This allows for self-sufficient energy supply in orbit, completely eliminating dependence on terrestrial power grids.

Cooling Advantage:

The vacuum of space, particularly on the dark side of Earth where temperatures plunge to -270°C, provides highly efficient radiative cooling. This method is 3 times more effectivethan traditional cooling systems and requires no precious water resources, solving the near-limit cooling challenges faced by ground-based data centers.

3

From “Earth-Sensed, Ground-Processed” to “Space-Processed, On-Demand Delivered”Space-based data centers are pioneering a new paradigm: “in-orbit processing + on-demand downlink”, which disrupts the traditional “Earth-sensed, ground-processed”model.Under the old model, vast amounts of raw satellite data had to be sent back to Earth, but limited satellite-to-ground bandwidthmade this process slow and expensive, leading to data bottlenecks—or even data loss.Space-based centers solve this by cleaning, analyzing, and intelligently extracting insights from data while still in orbit, transmitting only the most valuable analyzed results and decision-making informationback to Earth—enabling true “in-space computing”.A prime example is the Starcloudproject, which plans to launch the first AI-powered satelliteequipped with an NVIDIA H100 chip. Its core function will be to process multiple terabytes of raw data generated daily by spacecraft and space stations.This satellite can perform real-time analysisof satellite data, covering applications such as synthetic aperture radar (SAR) interpretationand deep-space radio signal processing—effectively bypassing the bottleneck of ground-based data transmission.Similarly, the “Three-Body Computing Constellation”developed by Zhejiang Labfocuses on space-based computation. Comprising 12 computing satellites, the constellation enables full interconnectivity between satellitesand delivers complete in-orbit computing capabilities.Each satellite offers 744 TOPS(trillions of operations per second) of computing power, with laser-based inter-satellite communication speeds of up to 100 Gbps, making it ideal for real-time taskslike disaster monitoringand weather forecasting.

4

Space-Based Data Centers vs. Traditional Ground Data CentersCompared to traditional terrestrial data centers, space-based data centersdemonstrate disruptive advantagesacross key dimensions: technical architecture, cost structure, deployment model, energy efficiency, and scalability.Particularly in terms of cost, space-based solutions show significant benefits.For example, operating a 40 MW data cluster for 10 years:

  • •A traditional data centerwould incur around 167million∗∗intotalcostsoveradecade,including∗∗140 million for energy consumptionand $7 million for cooling.
  • •In contrast, a space-based solutionwould cost only about $8.2 million.
    • •The largest single costis the one-time launch expense (~$5 million),
    • •Followed by solar array deployment (~$2 million).
    • •After that, energy is provided almost for freeby sunlight.

5

From Sci-Fi to Reality: Five Major Technical ChallengesSkeptics argue that space-based computing is still science fiction, citing insurmountable technical hurdles. But what are the real challenges?

1. Radiation Hardening & Hardware Reliability

Space’s extreme radiation environment threatens computing hardware. Satellites face threats like cosmic rays, single-event upsets (SEUs), and single-event latch-ups (SELs), which can cause logic errors or permanent chip damage.Solutions include using military-grade hardened electronicsor redundant backup systems. For instance:

  • •Axiom Spaceis testing military-spec hardware.
  • •Lonestaris exploring placing lunar data centers inside underground lava tubes for radiation shielding.Redundant computing modules are also essential to prevent single points of failure.

2. Thermal Management

While space offers excellent radiative cooling, high-power chips like GPUsstill need effective heat dissipation.Since there’s no air for convection, heat pipes or fluid loopsmust transfer heat to radiators, which then emit infrared radiation. Projects like Starclouduse a mix of liquid cooling and large radiator panels.But larger radiators add weight, increasing launch costs.

3. Energy Stability

Although solar power is 2–3x more efficient in space, eclipse periods (shadow zones)remain a challenge. Satellites must rely on battery storage, whose capacity and lifespanare limiting factors.Starcloudplans to deploy a massive 5 km × 4 km solar array, but this requires breakthroughs in megastructure deployment in orbit.

4. Communication & Autonomous Maintenance

Communication between satellites and with Earth involves latency, though laser links(like those used by Starlink and Starcloud) help reduce it. However, atmospheric interferenceand signal attenuation over long distancesremain issues.Additionally, space-based centers require lightweight, containerized softwarethat can run autonomously, enabling self-healing and decision-makingwithout human intervention.

5. Launch Costs & Scalability

Reusable rockets like SpaceX’s Starshiphave reduced per-launch costs, but gigawatt-scale projects(e.g., Starcloud’s 5 GW initiative) still require large-scale constellations, keeping total costs high.Long-term, orbital congestion in low-Earth orbit (LEO)may also impact deployment locations and thermal efficiency.

6

Key Players in the Space Computing ArenaThe space computing sector is still in its early stages, with participation from both startupsand tech giants.

Notable Startups:

  • •Starcloud(formerly Lumen Orbit): A pioneer focused on building orbital data centers. It plans to launch the world’s first AI satellite equipped with NVIDIA’s H100 chip (“Cloud-0”), aiming to build a gigawatt-scale orbital data center. Its H100 chips are expected to deliver 100x the performance of those on the ISSin zero gravity.
  • •Axiom Space, Lonestar, and others are also exploring lunar and orbital data solutions.

Major Tech Companies:

  • •NVIDIA: Collaborates with Starcloud via its Inception program, planning to launch its first H100-equipped satellite in 2025.
  • •Amazon: Its Project Kuiperaims to rival Starlink, having launched its first 27 satellites in 2025 via Atlas V rockets. AWS edge computing will support future in-orbit AI data nodes.
  • •Microsoft: Partners with SpaceX on Azure Space, providing global cloud access via Starlink, and is testing satellites for U.S. government use. Azure Orbital Cloud Accessis in preview.
  • •Meta: Teamed up with NVIDIA and HP on “Space Llama”, offering AI research support on the ISS and optimizing astronaut operations in real time.
  • •SpaceX: Already a leader with its massive Starlink constellationand advanced laser inter-satellite links. (For deeper insights, see our previous in-depth analyses on SpaceX.)

7

Space Computing Industry Chain Overview

Upstream (Launch & Infrastructure):

Includes satellite manufacturersand launch service providers.

  • •Satellite makers: Maxar, Thales Alenia, Airbus Defence, Lockheed Martin
  • •Launch providers: SpaceX (Falcon series), Rocket Lab, Blue Origin, ULA, Arianespace

Midstream (Hardware & Communication):

Focuses on radiation-hardened computing hardwareand high-speed inter-satellite communication.

  • •Key players: SpaceX, OneWeb, Kepler, Hughes Network Systems
  • •Modular in-orbit infrastructure: Axiom Space, Loft Orbital, Skyloom

Downstream (Applications):

Transforms technological advantages into practical use across sectors like:

  • •Earth observation (Planet Labs)
  • •Telecommunications (Iridium, Globalstar)
  • •Autonomous driving, and more (details omitted here due to length)