Nvidia is launching a computing platform for orbital data centers, aiming to meet AI demand and tackle challenges in space computing.
Nvidia has unveiled plans to develop a computing infrastructure specifically designed for orbital data centers, as revealed by CEO Jensen Huang during the GTC 2026 conference. He emphasized that as satellite constellations expand and humanity explores deeper into space, computing needs to occur where data is generated.
“Computing in space is the final frontier, and it’s already here,” Huang stated.
The company’s Vera Rubin Space-1 module, incorporating IGX Thor and Jetson Orin chips, will be utilized by various firms for their space missions. These chips are tailored for the constraints of space environments, focusing on minimal size, weight, and energy consumption.
Huang acknowledged ongoing collaboration with partners to address the technical issues involved in creating these computer systems, particularly the cooling challenges posed by the lack of convection in space. “Many engineers are working on this problem,” he noted.
Nvidia's initiative responds to the surging demand for AI, exacerbated by rising electricity costs. The concept of establishing data centers in space hinges on the potential for unlimited space and access to continuous solar energy, though the high costs associated with launches remain a significant hurdle.
Meanwhile, SpaceX has applied to the U.S. Federal Communications Commission to deploy a constellation of one million satellites for data processing, as part of a vision to create a low Earth orbit network linked by laser channels. This project is described as a pivotal step towards advancing civilization on the Kardashev scale.
In a similar vein, California's Aetherflux plans to launch solar mini-farms as low-orbit satellites to transmit solar energy back to Earth, while Google announced intentions to create a satellite system for solar energy harnessing aimed at powering data centers. Research forecasts suggest that by 2030, the cost of AI computations in orbit could become more economical than on Earth.
Projected Growth and Demand
Nvidia anticipates order volumes for its Blackwell and Vera Rubin chips to reach a staggering $1 trillion by 2027, up from an estimated $500 billion last year. Huang highlighted a growing demand for Nvidia’s technology from both startups and established corporations, stating, “Having more computing power allows for generating more tokens and increasing revenues.”
Autonomous Vehicle Partnerships
Nvidia is also bolstering its initiatives in autonomous vehicle technology, announcing new partnerships with major automakers including Hyundai, Nissan, and BYD. These collaborations will utilize the Drive Hyperion platform, aimed at enhancing driver assistance and level 4 autonomous driving capabilities.
“The ChatGPT moment for autonomous vehicles has already arrived,” Huang remarked.
Despite the absence of fully autonomous vehicles in the market, companies like Waymo are leading the way with level 4 taxi services.
The Drive Hyperion platform integrates training models and simulations within data centers and vehicle systems, currently serving clients such as Uber and Stellantis.
Additional Developments
At the GTC event, Nvidia introduced the Groq 3 Language Processing Unit, marking its first delivery from the recently acquired startup Groq. A Groq LPX server rack, designed to operate alongside the Vera Rubin system, was also announced, promising a 35-fold increase in computation efficiency per watt for the Rubin chips.
“We combined two processors with completely different characteristics,” Huang explained.
Further innovations include a prototype server architecture named Kyber, which features 144 vertically arranged GPUs for enhanced computing density, and a developer toolkit for creating new AI systems on Nvidia hardware. Huang has reiterated his belief that AI will not eliminate jobs, but rather generate new opportunities.
