NVIDIA just put an AI supercomputer into orbit — 25x the H100

At its GTC conference on March 16, 2026, NVIDIA announced Space Computing — a dedicated initiative to put AI processing chips directly into orbit. The centerpiece is the Space-1 Vera Rubin Module, a chip designed specifically for satellite-based AI inferencing (processing AI requests directly in space, rather than sending data to Earth for analysis). It delivers 25 times more AI compute than the H100 GPU — NVIDIA's current data center workhorse — in a form factor rated for the radiation and thermal extremes of low Earth orbit.

CEO Jensen Huang said at GTC: "Space computing, the final frontier, has arrived. Intelligence must live wherever data is generated."

Why Putting AI Into Space Actually Matters

Every minute, satellites collect enormous amounts of data: imagery, radar, thermal sensors, communications signals. Today, almost all of that raw data is transmitted back to Earth, where it waits in a queue to be processed. This round-trip delay — ranging from minutes to hours depending on satellite position and ground station availability — creates a fundamental bottleneck for time-sensitive applications.

The Three Hardware Products

NVIDIA announced three distinct hardware platforms for the space computing ecosystem:

• Space-1 Vera Rubin Module (coming 2026): The flagship orbital AI chip. 25x more AI compute than the H100 GPU (H100 = NVIDIA's current top-of-line data center GPU, used in most major AI training clusters). Designed for Low Earth Orbit environments — handles radiation, temperature extremes, and the hard vacuum of space. Built for orbital data centers (ODCs) where satellites serve as AI processing nodes in the cloud.
• NVIDIA RTX PRO 6000 Blackwell Server Edition (available now): Ground-based counterpart. Processes satellite data on Earth at up to 100x the speed of legacy CPU-based systems (CPU = a standard computer processor; GPU = graphics processor, dramatically faster for AI tasks). Enables near-real-time analysis of downlinked satellite data for organizations that aren't yet ready for full orbital processing.
• NVIDIA Jetson Orin (available now): An ultra-compact embedded module for satellites where size and power consumption are hard constraints. Enables AI on small satellites (CubeSats and similar form factors) that can't carry a full server-class GPU.

Six Companies Already Signed On

NVIDIA announced six launch partners that will build orbital AI systems on the new hardware:

• Aetherflux — orbital energy transmission and data relay
• Axiom Space — commercial space station operations
• Kepler Communications — satellite internet infrastructure
• Planet Labs PBC — Earth observation imagery (publicly traded; operates the world's largest fleet of Earth-imaging satellites)
• Sophia Space — space-based AI processing services
• Starcloud — orbital cloud computing

Planet Labs is the most recognizable name: it operates hundreds of satellites that image the entire Earth daily. Integrating NVIDIA Space-1 compute into Planet's constellation would mean AI analysis happens in orbit, reducing the latency between a photo being taken and an alert being generated from hours to seconds.

What This Means for Non-Space Industries

If you work in agriculture, infrastructure, or emergency management, orbital AI means the satellite services you rely on (crop monitoring, infrastructure inspection, disaster mapping) will become dramatically more responsive. Instead of receiving a report the next business day, you'd get alerts within minutes of an event.

If you're a developer building on geospatial data (location services, mapping, environmental data), the NVIDIA IGX Thor platform (available now) and the developer toolkit at build.nvidia.com provide the ground-based infrastructure to start building orbital-AI-ready pipelines today.

The full announcement is available on the NVIDIA Space Computing newsroom page. The Space-1 Vera Rubin Module ships later in 2026; ground-based platforms are available now.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News