Nvidia Releases "Space Computing Module," "Space Version" Vera Rubin to Follow

Nvidia is extending its AI computing footprint into Earth’s orbit.

At the overnight GTC annual developer conference, Nvidia announced the launch of a dedicated computing module for space scenarios, disclosed plans for a space version based on the Vera Rubin architecture, and released the enterprise AI agent platform NemoClaw, showcasing its expansion ambitions in AI infrastructure.

Nvidia CEO Jensen Huang announced in his keynote that six partners—Aetherflux, Axiom Space, Kepler Communications, Planet, Sophia Space, and Starcloud—will deploy Nvidia hardware in orbit. The company stated that the new module targets applications such as orbital data centers, advanced geospatial intelligence processing, and autonomous space operations, offering up to 25 times the AI inference power compared to the previously launched H100 GPU.

The Vera Rubin space module will be officially launched “at a later date,” with IGX Thor and Jetson Orin products currently available. Meanwhile, Nvidia also released the enterprise AI agent platform NemoClaw, designed to provide secure and private on-premises AI autonomous agent deployment for enterprises, currently in early Alpha testing.

Space Computing Deployment: From H100 to Vera Rubin

Nvidia’s move into space computing dates back to November last year. According to Data Center Dynamics, Starcloud launched a test satellite equipped with an Nvidia H100 GPU into orbit, marking Nvidia’s first GPU in space.

The space computing module system released covers three levels: Vera Rubin space module for high-intensity workloads like orbital data centers, IGX Thor based on the Blackwell architecture for edge scenarios, and Jetson Orin for real-time processing of visual, navigation, and sensor data. IGX Thor and Jetson Orin are now available, while the specific release date for Vera Rubin remains undisclosed.

On the partnership front, Aetherflux plans to use Vera Rubin for solar-powered high-performance AI inference in orbit; Sophia Space will incorporate Jetson Orin into its modular platform for satellite operators; Kepler Communications also selected Jetson series products.

Huang Huang announced, “As we deploy satellite constellations and explore deeper space, intelligence must exist wherever data is generated. AI processing capabilities span space and ground systems, transforming orbital data centers into exploration tools and spacecraft into autonomous navigation systems.”

Notably, Nvidia earlier this month posted a job opening for a space data center systems architect, describing it as “an opportunity to join an AI systems leader at the dawn of a new industry.”

Competitive Landscape: Google, Musk Enter the Arena

The space computing race is attracting major players. According to Data Center Dynamics, Google plans to launch multiple batches of TPUs into space, having completed tests simulating low Earth orbit radiation environments with particle accelerators, and has partnered with Planet for small-scale deployments, with long-term goals of delivering gigawatt-level computing power into space.

SpaceX CEO Elon Musk is also seeking to deploy a mega constellation of hundreds of thousands of satellites for orbital AI data centers. Reports suggest these satellites will use Tesla’s self-developed chips, though the timeline remains unclear.

However, the commercial prospects of space data centers face skepticism. OpenAI CEO Sam Altman, short-seller Jim Chanos, AWS CEO Matt Garman, and Gartner analysts have all expressed critical views of this concept.

NemoClaw: Providing a Secure Foundation for Enterprise AI Agents

In the enterprise AI agent space, Nvidia also launched the NemoClaw platform. According to TechCrunch, NemoClaw is built on the open-source local AI autonomous agent project OpenClaw, developed in collaboration with OpenClaw founder Peter Steinberger.

NemoClaw’s core positioning is to add enterprise-grade security and privacy mechanisms on top of OpenClaw, allowing companies to access the platform with a single command and uniformly control agent behaviors and data handling. The platform supports integration with any programmable or open-source AI models, including Nvidia’s own NemoTron open-source models, and is deeply integrated with Nvidia’s AI agent software suite NeMo. Notably, the platform remains hardware-agnostic and does not require running on Nvidia GPUs.

Huang Huang compared the emergence of OpenClaw to historic moments like Linux, Kubernetes, and HTML, stating: “Every company today needs to develop an OpenClaw strategy—a unified AI agent system strategy.”

Currently, NemoClaw is in early Alpha, with Nvidia’s official developer documentation acknowledging “rough edges” and indicating ongoing efforts toward production-ready sandbox orchestration capabilities. Competition in the enterprise AI agent platform space is intensifying—OpenAI launched its enterprise platform OpenAI Frontier in February, and Gartner released a report last December listing AI agent governance platforms as critical infrastructure for enterprise AI adoption.

Risk Warning and Disclaimers

Market risks are present; invest cautiously. This article does not constitute personal investment advice and does not consider individual users’ specific investment goals, financial situations, or needs. Users should evaluate whether any opinions, views, or conclusions herein are suitable for their circumstances. Investment is at your own risk.