CAPE CANAVERAL, Nov. 14 (Reuters) — Elon Musk’s SpaceX announced plans to deploy the world’s first space-based AI data center cluster, leveraging Starship’s launch capabilities and upgraded Starlink V3 satellites to deliver 1 terawatt (TW) of annual AI computing power by 2026, the company revealed in a technical briefing on Monday .
The breakthrough hinges on Starlink V3 satellites, which boast 10x greater capacity than their predecessors at 1 Tbps per satellite and support million-level GPU parallel computing via high-speed laser links . Each 1.5-2 ton satellite is equipped with radiation-hardened chips and 3D-printed heat sinks, enabling operation in extreme temperatures ranging from -270℃ to 120℃. "We’re moving computing from concrete bunkers on Earth to the vacuum of space—where energy is abundant and cooling is free," said SpaceX’s director of satellite engineering.
The space-based cluster offers transformative advantages over terrestrial data centers. Its lifecycle costs are just 1/10 of ground facilities, eliminating expenses for land, grid expansion, and cooling systems while cutting carbon emissions by 90% . Starship’s capacity to launch 60 V3 satellites per mission will enable 12 deployment flights in 2026, forming a low-Earth orbit network with 0.3ms latency—50x faster than ground-based data centers.
Industry rivals are racing to match the innovation. Amazon plans a 3.2-gigawatt orbital computing network by 2030 via its Kuiper Project, while China’s Guoxing Aerospace and Zhejiang Lab have launched the "Three-Body Computing Constellation," a 12-satellite network delivering 744 TOPS of 算力 . Startup Starcloud already demonstrated feasibility with its Starcloud-1 satellite, which uses NVIDIA H100 GPUs to achieve 100x faster inference than ground systems.
Real-world applications are rapidly emerging. Climate scientists anticipate 1km-grid precision in global weather modeling using the space cluster’s real-time atmospheric data processing. For deep space exploration, the network will provide localized 算力 for lunar bases and future Mars missions, reducing reliance on Earth-based communication . Even commercial sectors stand to benefit: orbital rendering farms could enable 8K/120Hz holographic communication for the metaverse.
Challenges remain, including radiation doses 25x higher than Chernobyl’s post-disaster levels, requiring silicon carbide chips and polyethylene shielding that add 300% to hardware costs . Satellites also face 45-minute daily shadow periods, demanding lithium-sulfur batteries that occupy 40% of satellite weight. "We’re developing AI-driven self-healing systems to address these gaps," the engineering director noted.
Regulators are scrutinizing the technology: the EU is evaluating compliance with the AI Act, particularly for on-satellite processing of sensitive data . Despite hurdles, the market potential is staggering—analysts project the global space computing market will reach $420 billion by 2030, accounting for 35% of total 算力 demand .
Image Suggestion: Artist’s rendering of the Starlink V3 constellation—glowing satellites connected by laser links, with insets showing a Starship launch (top) and Earth-based data center comparison infographic (bottom).
You must be logged in to post a comment.