NVIDIA GTC 2026 | NVIDIA Launches Space-1 Vera Rubin into Space to Build a True "Cloud Computing" Platform

At NVIDIA GTC 2026, Jensen Huang announced the launch of the Space Computing platform, extending artificial intelligence computing capabilities from ground data centers into orbit, aiming to establish a decentralized AI infrastructure spanning Earth and space. This new architecture pushes AI computation further from the cloud into orbital data centers.

NVIDIA introduced Space-1 Vera Rubin as the core of space computing. This module compresses data center-level processing power into a size, weight, and power envelope suitable for space environments. Compared to the NVIDIA H100 GPU, it can deliver up to 25 times the AI computing performance in space inference scenarios. This enables large models and advanced geospatial analysis to run directly in orbit, supporting real-time image processing and autonomous scientific exploration.

As data center power consumption rapidly increases, the industry is turning its attention to the nearly unlimited solar energy resources in space, attempting to establish computing capabilities directly in orbit to reduce costs associated with data transmission and energy constraints. However, this vision faces challenges such as heat dissipation in space (lack of convection), rocket launch costs, and orbital resource competition. Companies like Aetherflux, Axiom Space, Kepler Communications, Planet, Sophia Space, and Starcloud are also mentioned.

NVIDIA Launches Space-1 Vera Rubin for Space Data Centers

The core of NVIDIA’s new offerings is a computing acceleration platform designed specifically for space environments, including the Space-1 Vera Rubin module, as well as edge AI systems like IGX Thor and Jetson Orin. These products are optimized for size, weight, and power (SWaP) constraints, aiming to provide near ground data center-level AI processing in orbit. Unlike traditional architectures that rely mainly on ground-based CPU batch processing, this generation enables real-time inference and data processing directly on satellites or space vehicles.

The Vera Rubin module features a new generation GPU that, in space inference tasks, can deliver up to 25 times the performance of previous NVIDIA H100 GPUs, supporting large language models and foundational models to run directly in orbit. Satellites are no longer just data collectors but intelligent nodes capable of real-time decision-making.

IGX Thor emphasizes industrial-grade stability and security, supporting real-time AI processing, autonomous operation, and secure boot; Jetson Orin offers a high-performance, low-power compact module for real-time analysis of vision, navigation, and sensor data.

Unlimited Solar Resources: Pros and Cons of NVIDIA’s Space Data Centers

This “from ground to space” AI architecture is attracting many commercial space companies, including Axiom Space, Kepler Communications, Planet Labs, and Starcloud. These companies are attempting to embed AI capabilities into satellite networks and space infrastructure, enabling data to be analyzed and utilized immediately upon generation, rather than being transmitted back to Earth for processing.

NVIDIA CEO Jensen Huang stated at GTC 2026 that as satellite constellations and deep space exploration missions increase, intelligence must exist where data is generated. By pushing AI into space, satellites will gain real-time sensing, decision-making, and autonomous operation capabilities. Orbiting data centers will evolve from simple storage nodes to engines of scientific discovery and real-time insights.

The narrative of space data centers is closely tied to the energy bottlenecks faced by ground AI infrastructure. As data center power consumption surges, the industry is increasingly looking toward the nearly limitless solar energy in space, aiming to establish computing capabilities directly in orbit to reduce costs associated with data transmission and energy limitations.

However, this vision still faces engineering and practical challenges, including heat dissipation in space (lack of convection), rocket launch costs, and orbital resource competition. Nonetheless, from Google’s space computing initiatives to SpaceX’s large-scale satellite deployments, the integration of AI and space infrastructure has become an industry consensus.

This article, NVIDIA GTC 2026 | NVIDIA Launches Space-1 Vera Rubin into Space to Build a True “Cloud Computing” Platform, originally appeared on Chain News ABMedia.