Nvidia has announced the development of a computing platform for orbital data centers. This was stated by CEO Jensen Huang at the GTC 2026 conference.

“Computing in space is the final frontier, and it’s already here. As satellite constellations are deployed and we venture deeper into the galaxy, intelligence must be where the data is generated,” the entrepreneur noted.

Nvidia's press release mentions that several companies will utilize the Vera Rubin Space-1 module, which includes IGX Thor and Jetson Orin, for space missions. The chips are specifically designed for environments where size, weight, and power consumption are severely limited.

Huang emphasized that the company is collaborating with partners on a new computer for orbital data centers, but there are technical challenges at this stage.

“There is no convection in space—only radiation. So we need to figure out how to cool such systems. Many engineers are working on this problem,” he said.

The construction of data centers to meet the growing demand for AI is linked to rising electricity prices. One proposed solution is to place computing power in space, where there is unlimited space and constant solar energy. However, the high cost of launches remains a significant barrier.

In February, SpaceX filed a request with the U.S. Federal Communications Commission to launch a group of 1 million satellites for data processing centers into orbit.

The project aims to create a network of data centers in low Earth orbit connected by laser channels. The document uses bold phrases like “the first step towards a second-level civilization on the Kardashev scale.”

In 2026, California startup Aetherflux plans to launch solar mini-farms in the form of low-orbit satellites to transmit energy from space to Earth using lasers. The technology will be deployed using SpaceX rockets.

In November 2025, Google announced its intention to create a satellite system in low Earth orbit to harness solar energy and power data centers. In the same month, experts from research group 33FG predicted that by 2030, AI computations in orbit would be cheaper than on Earth.

Trillion

Huang stated that the expected order volume for Blackwell and Vera Rubin chips is projected to reach $1 trillion by 2027.

Last year, the company estimated potential revenue from these two generations of semiconductors at $500 billion. However, following the release of financial results last month, CFO Colette Kress noted that growth for 2026 could exceed previous estimates.

According to Huang, demand for Nvidia's solutions is increasing from both startups and large corporations.

“Having more computing power allows for generating more tokens and increasing revenues,” he said.

Autonomous Vehicles

Nvidia is expanding partnerships in the development of autonomous vehicles. The company announced new agreements with Hyundai Motor, Nissan Motor, Isuzu, BYD, and Geely.

These agreements pertain to the Drive Hyperion platform designed for vehicles. The system aids in developing and integrating driver assistance and level 4 autonomous driving tools.

“We have been working on self-driving cars for a long time. The ChatGPT moment for autonomous vehicles has already arrived,” Huang stated.

Currently, there are no vehicles on the market that can operate completely without human oversight. However, some companies like Waymo are already offering level 4 taxi services.

Most current autopilots operate at level 2, requiring the driver to maintain constant control.

Drive Hyperion includes training models in data centers, large-scale simulations, and computing systems within the vehicle. Current clients of the platform include Aurora Innovation, Nuro, Sony Group, Uber, Stellantis, and Lucid Group.

Other Releases

At GTC 2026, Huang introduced the Groq 3 Language Processing Unit (LPU)—the first chip from the startup Groq, which Nvidia acquired in December 2025 for $20 billion. Deliveries are expected in the third quarter.

The entrepreneur also announced the Groq 3 LPX server rack, which will consist of 256 LPUs. It is designed to work alongside the Vera Rubin system, with deliveries expected later in 2026. According to Huang, the rack can increase token computation efficiency per watt for Rubin by 35 times.

“We combined two processors with completely different characteristics: one for high throughput and the other for low latency. This does not negate the fact that we need a lot of memory. So we will just add a large number of Groq chips, which will expand its available volume,” Huang noted.

Additionally, a prototype called Kyber was demonstrated—a next-generation server architecture consisting of 144 GPUs arranged vertically to enhance computing density and reduce costs.

Kyber will be part of the Vera Rubin Ultra system, with deliveries expected in 2027.

Nvidia's CEO also unveiled a developer toolkit that allows for the creation and testing of new AI systems on the company's hardware. He showcased the NemoClaw stack, created specifically for OpenClaw.

Recall that in March, Huang dismissed the notion of AI as a "job killer" and emphasized it as a primary generator of new jobs.