Nvidia on Monday unveiled a new generation of computing platforms designed for orbital data centers, marking a significant step toward deploying artificial intelligence infrastructure in space as demand for computing power continues to surge.

The announcement came during the company’s GTC 2026 conference, where CEO Jensen Huang described space-based computing as the next frontier for AI systems that increasingly require massive processing capacity close to where data is generated.

“Space computing, the final frontier, has arrived,” Huang said during the event. “As we deploy satellite constellations and explore deeper into space, intelligence must live wherever data is generated.”

The move marks a major shift as the AI industry explores unconventional solutions to meet the soaring demand for computing resources driven by generative AI, robotics, and autonomous systems.

A New Computing Platform Built For Space

At the center of the announcement is Nvidia’s Vera Rubin Space-1 Module, a computing platform designed specifically for use in satellites and orbital infrastructure. The module integrates Nvidia’s IGX Thor and Jetson Orin processors and is engineered to operate in size-, weight-, and power-constrained environments, conditions that are critical for space missions where hardware must be compact, energy-efficient, and resilient.

According to the company, the platform will support space missions being developed by several industry partners, including Axiom Space, Starcloud, and Planet Labs. These systems are expected to enable satellites to process data directly in orbit, reducing the need to transmit massive volumes of raw data back to Earth before analysis.

Such an approach could transform how Earth observation, communications, and deep-space exploration missions operate. Traditionally, satellites have collected data and transmitted it back to ground-based data centers for processing.

Nvidia’s approach aims to bring AI computing closer to the data source, allowing satellites to analyze information in real time. This capability could allow satellites to filter and process imagery, track weather systems, monitor infrastructure, or detect anomalies without waiting for instructions from Earth.

For example, Earth-observation satellites could use onboard AI to identify natural disasters, track deforestation, or analyze agricultural activity instantly, dramatically reducing response times. The development also aligns with a broader trend in computing known as edge AI, where processing occurs near the point where data is generated rather than in distant centralized servers.

The Engineering Challenges

Although the technology is promising, significant engineering hurdles remain before orbital data centers become widespread. One of the most difficult challenges is cooling high-performance computing systems in the vacuum of space.

“In space, there’s no convection, there’s just radiation,” Huang said during his keynote address.

“And so we have to figure out how to cool these systems out in space, but we’ve got lots of great engineers working on it.”

Cooling is a major issue because traditional data centers rely on air or liquid circulation to remove heat generated by processors. In space, heat must instead be dissipated through radiation, requiring new thermal designs and materials.

The Search For Power Beyond Earth

The push toward space-based computing is partly driven by the rapidly escalating energy demands of artificial intelligence. The construction of massive AI data centers on Earth has already been linked to rising electricity consumption and strain on power grids in several regions.

By contrast, satellites and orbital facilities could potentially harness virtually unlimited solar energy without the land and infrastructure constraints faced by terrestrial data centers. Technology companies are increasingly studying whether space could provide a long-term solution to the energy demands of large-scale computing.

In November, Google announced Project Suncatcher, an initiative exploring the feasibility of deploying computing infrastructure powered by solar energy in orbit.

The concept is also gaining traction among companies involved in space launch and satellite infrastructure. Last month, Elon Musk’s AI startup xAI was acquired by SpaceX in a deal valued at $1.25 trillion, a move widely interpreted as part of a strategy to build AI-powered computing systems in orbit.

SpaceX is one of Nvidia’s largest customers for AI chips, supplying hardware used to train and operate advanced AI models. Earlier this year, SpaceX also asked the Federal Communications Commission for approval to launch as many as one million satellites intended to support orbital AI infrastructure.

With generative AI models becoming larger and more computationally demanding, traditional data centers on Earth may struggle to keep pace with the scale of future workloads. Against this backdrop, space has become an alternative. But Space-based computing remains an ambitious concept with significant technical and regulatory challenges.

Nvidia said the revenue opportunity for its artificial intelligence chips could reach at least $1 trillion by 2027, as the company pivots more aggressively toward the rapidly expanding market for real-time AI computing.

The forecast was outlined by CEO Jensen Huang during the company’s annual Nvidia GTC developer conference in San Jose, California, where the chipmaker introduced new processors and system designs aimed at accelerating AI responses for large-scale applications.

“The inference inflection has arrived,” Huang said during the keynote address delivered in a packed arena with more than 18,000 attendees, highlighting how the AI industry is transitioning from building models to deploying them widely.

The estimate represents a sharp increase from the roughly $500 billion AI infrastructure opportunity Nvidia previously projected for 2026, reinforcing investor expectations that global demand for AI computing capacity will continue expanding at a rapid pace.

Shares of Nvidia — currently the world’s most valuable publicly traded company with a market capitalization exceeding $4.3 trillion — briefly rose following the announcement before closing about 1.6% higher.

AI industry entering deployment phase

The company’s updated forecast indicates a broader shift taking place across the artificial intelligence industry. Over the past two years, the biggest technology companies have spent hundreds of billions of dollars acquiring computing hardware to train increasingly sophisticated AI models. Now the focus is rapidly moving toward inference — the process where trained models generate responses, make predictions, or execute tasks in real time for users.

This stage of AI computing is expected to be even larger in scale than the training phase because it involves serving millions or potentially billions of user queries daily across applications such as chatbots, search engines, productivity tools, and autonomous systems.

Companies including OpenAI and Anthropic are rapidly expanding AI services to support growing user bases, while technology firms such as Meta Platforms are integrating AI assistants into social media platforms used by billions of people. As a result, demand for hardware capable of delivering AI responses instantly — with minimal latency — is rising sharply.

While Nvidia dominates the market for chips used to train large language models, inference computing has become a more competitive arena. Large technology companies have begun designing their own specialized processors optimized for running AI models efficiently at scale.

Meta, for example, has invested heavily in developing custom AI chips for its internal infrastructure, while several cloud providers are also building proprietary silicon.

To maintain its leadership, Nvidia unveiled a new architecture designed specifically for high-performance inference workloads. The company’s upcoming Vera Rubin AI chip will perform a stage known as “prefill,” in which human prompts are converted into machine-readable tokens that AI systems can process.

The second stage of the process, called “decode,” will be accelerated by chips from startup Groq, which specializes in extremely fast AI inference.

Nvidia licensed Groq’s technology in a deal valued at $17 billion last year, denoting how critical inference performance has become for the next phase of AI deployment. The company aims to reduce the time it takes AI systems to generate responses — a key metric for applications ranging from digital assistants to automated customer service tools — by combining its own processors with Groq’s specialized hardware.

Even as competitors develop alternative AI chips, Nvidia continues to benefit from a powerful ecosystem built around its programming platform, CUDA. The CUDA software environment allows developers to design algorithms optimized for Nvidia hardware, creating a large installed base that reinforces the company’s dominance.

“The installed base is what attracts developers who then create the new algorithms that achieve the breakthrough technologies,” Huang said.

“We are in every cloud. We’re in every computer company. We serve just about every single industry.”

That ecosystem advantage has made it difficult for competing chip architectures to gain widespread adoption, even when they offer specialized performance improvements.

Next Generation Of AI Processors

Huang also introduced a new processor called the Feynman AI chip, named after the Nobel Prize-winning physicist Richard Feynman. The chip forms part of Nvidia’s long-term roadmap for advancing AI computing capabilities, with future processors expected to deliver significantly higher performance for both training and inference workloads.

The company is also investing heavily in technologies that enable faster communication between processors in large AI data centers. Analysts expect Nvidia to elaborate further on its recent $2 billion investments in optical networking companies Lumentum and Coherent Corp., which manufacture laser-based components used to transmit data between chips using beams of light.

Optical interconnects are increasingly essential as AI systems grow to include tens of thousands of processors operating simultaneously within massive data center clusters.

Beyond technology companies, national governments are also becoming major buyers of AI infrastructure. Countries, including Saudi Arabia, are investing heavily in sovereign AI systems designed to support national data processing, research, and digital services. These projects often rely on Nvidia’s processors, reinforcing the company’s role as a foundational supplier to the global AI ecosystem.

At the same time, AI hardware has become a key element of technological competition between the United States and China, with export controls limiting the sale of some advanced chips to Chinese companies. Despite those geopolitical tensions, Nvidia continues to release open-source AI software tools, positioning itself as a central platform provider for developers worldwide.

Nvidia’s $1 trillion revenue opportunity forecast highlights how the AI infrastructure boom is expected to extend well beyond the initial wave of model development.

As artificial intelligence becomes embedded across industries — from healthcare and finance to logistics and manufacturing — the amount of computing power required to support those systems will expand dramatically. Analysts say the next stage of growth will be driven by applications that require real-time AI responses, including autonomous vehicles, intelligent robotics, and advanced digital assistants.

If those technologies scale globally, demand for AI computing infrastructure could grow far beyond current projections. The transition from model training to real-time inference is being interpreted as the next major phase of growth, especially for Nvidia, which has already become the most valuable company in the world, largely due to the AI boom.