Nvidia GTC 2026 Highlights: Huang’s Keynote, Vera Rubin and Space Plans

In Focus

• Nvidia projects at least $1 trillion in GPU order revenue from 2025 through 2027
• Vera Rubin, the next-generation AI platform, ships in 2026; Feynman follows in 2028
• OpenClaw and NemoClaw bring agentic AI systems to Nvidia’s enterprise stack
• AWS commits to deploying more than 1 million Nvidia GPUs across global cloud regions this year

Nvidia’s annual GPU Technology Conference (GTC) happening in San Jose is a four-day event focused on innovative AI chips, agentic software platforms and space computing. According to NVIDIA’s official website, CEO Jensen Huang addressed a capacity crowd at the SAP Center at the AI Conference, outlining a trillion-dollar vision.

Huang said computing demand has increased “by 1 million times over the last few years,” citing $150 billion in venture investment flowing into AI-native startups.

Trillion-Dollar Forecast

Jensen Huang opened his keynote with a bold financial outlook, projecting at least $1 trillion in GPU order revenue between 2025 and 2027. He attributed the surge to an explosion in AI-native startups, with $150 billion in venture investment flowing into the sector over the past year alone.

Huang said computing demand has increased “by 1 million times over the last few years,” describing NVIDIA as “the inference king” due to its extreme codesign approach. On the cloud side, AWS committed to deploying more than 1 million Nvidia GPUs across global regions this year. Microsoft Azure became the first hyperscale cloud to power up the Vera Rubin NVL72 systems.

New Hardware and a Generational Roadmap

GTC 2026 saw the following key platform announcements:

• Vera Rubin (2026): A full-stack platform with seven chips, five rack-scale systems, and one supercomputer for agentic AI, including the new Vera CPU and BlueField-4 STX storage architecture.
• Feynman (2028): The next architecture, featuring the Rosa CPU, LP40 LPU, BlueField-5, and dual copper and co-packaged optical networking.
• DLSS 5: Uses neural networks to generate entire game frames rather than upscale them, turning standard gaming PCs into local AI inference engines.
• RTX PRO 4500 Blackwell Server Edition: Delivers 100x performance over CPU-only servers for vision AI in a compact 165-watt, single-slot form factor.

Agentic AI Goes Enterprise

Making agentic AI systems Nvidia-ready for enterprise was a defining theme at GTC 2026. Major updates included:

• OpenClaw: Huang called it “the most popular open source project in the history of humanity” as it crossed 100,000 GitHub stars and drew over 2 million visitors in its first week.
• NemoClaw: A new enterprise software stack combining policy enforcement, network guardrails, and privacy routing for secure, always-on AI agent deployment.
• Nemotron Coalition: Six open frontier model families launched, covering language, robotics, autonomous driving, biology, chemistry, and climate.
• Autonomous Vehicles: BYD, Hyundai, Nissan, and Geely joined the robotaxi-ready platform, alongside an Uber deal to deploy Nvidia Drive AV across 28 cities on four continents by 2028.
• IGX Thor (GA): Nvidia’s industrial edge AI platform went generally available, with adoption from Caterpillar, Johnson and Johnson, CERN, and Planet Labs.

Physical AI Enters Production

At GTC 2026, Nvidia moved physical AI from concept to deployment across multiple industries. On the automotive front, BYD, Hyundai, Nissan, and Geely joined the robotaxi-ready platform, alongside an Uber deal to deploy Nvidia Drive AV across 28 cities on four continents by 2028.

Huang was direct about the milestone: “The ChatGPT moment of self-driving cars has arrived.” In industrial settings, IGX Thor, Nvidia’s edge AI platform, went generally available, with adoption from Caterpillar, Johnson and Johnson, CERN, and Planet Labs. The company also confirmed that NVIDIA Space-1 Vera Rubin is being designed to run AI data centers in orbit.

Nvidia Sets Its Sights on Space

One of the more unexpected Nvidia GTC 2026 announcements was the company’s push into space computing. Huang confirmed that future systems like NVIDIA Space-1 Vera Rubin are being designed to bring AI data centers into orbit, extending accelerated computing beyond Earth for the first time.

The Vera Rubin architecture itself is named after the astronomer whose work revealed dark matter, and the Rosa CPU in the Feynman generation honors Rosalind Franklin, whose X-ray crystallography revealed the structure of DNA. For Nvidia, the naming is intentional; at every scale, science is the destination.

A Shift Toward AI Infrastructure

The Nvidia GTC 2026 highlights signal a decisive shift in the AI industry from model training to large-scale inference and autonomous agent deployment. With a projected $1 trillion order horizon, a multi-year hardware roadmap, and enterprise-grade agentic software now available, Nvidia is shaping infrastructure decisions across cloud, automotive, healthcare, and industrial sectors for years ahead.

The announcements at this Nvidia AI conference 2026 are likely to accelerate enterprise adoption timelines and intensify competition among chipmakers, cloud providers, and AI software vendors.