NVIDIA Says “Useful AI Has Arrived” as Vera Rubin Enters Full Production

Key Points

• Jensen Huang said “useful AI has arrived,” framing NVIDIA’s next growth phase around agentic AI, AI factories, enterprise software, PCs, robotics and physical AI.
• NVIDIA announced that its Vera Rubin platform is now in full production, with major customers like Microsoft, Dell and CoreWeave already operating engineering racks.
• The company also unveiled new tools and systems for enterprise and physical AI, including DSX for AI factories, Nemotron 3 Ultra, updated PC hardware, and robotics/autonomous vehicle platforms.

NVIDIA NASDAQ: NVDA Founder and CEO Jensen Huang used a keynote in Taiwan to outline the company’s next phase of growth around agentic AI, AI factories, new data center systems, enterprise software tools, personal computers and robotics.

Speaking at GTC Taiwan, Huang repeatedly credited Taiwan’s supply chain and manufacturing partners, saying NVIDIA’s ecosystem extends “all the way upstream” to suppliers in Taiwan and downstream to data centers and end users. He said the company’s business with its partners is growing rapidly as demand for AI compute rises.

Huang Says “Useful AI Has Arrived”

Huang framed the keynote around what he called the arrival of agentic AI, or AI systems that can reason, plan, use tools and complete tasks. He contrasted the new model with traditional software, saying agents combine large language models, orchestration software, memory systems, tools and runtimes.

He pointed to software development as an early example, citing GitHub activity and arguing that AI is increasing developer productivity rather than reducing demand for software engineers. “People talk about AI reducing jobs. Complete nonsense,” Huang said. “It’s causing more software engineers to be hired.”

Huang said the increased usefulness of AI has changed the economics of computing, describing tokens as “profitable units of revenues.” He said that shift is driving demand for AI factories, the large-scale data centers used to train and run AI systems.

“Compute is revenues. Compute is profit,” Huang said, arguing that performance per watt, reliability, speed of deployment and system longevity are becoming central economic measures for AI infrastructure operators.

Vera Rubin Enters Full Production

A central announcement was that NVIDIA’s Vera Rubin platform is now in full production. Huang described Vera Rubin as more than a GPU, calling it a multi-rack, pod-scale system built specifically for agentic AI workloads.

According to the presentation, Vera Rubin includes multiple connected rack-scale systems, including Rubin GPU systems, Vera CPU racks, BlueField-based storage and security systems, and NVIDIA networking. The company said the platform uses chips manufactured through TSMC processes, with HBM4 memory from Micron, SK hynix and Samsung.

Huang said the supply chain for Vera Rubin is twice as large as the one created for Grace Blackwell. He also said rack assembly has improved significantly, stating that what previously took two hours to assemble for one Grace Blackwell rack now takes five minutes.

NVIDIA said Microsoft has an operational Vera Rubin NVL72 engineering rack, while Dell and CoreWeave have also stood up Vera Rubin NVL72 engineering racks.

NVIDIA Positions Vera CPU for Agentic Workloads

Huang also highlighted the Vera CPU, which he described as a CPU “built for agents” rather than traditional human-driven computing. He said agentic workloads require low latency, high single-threaded performance, high bandwidth and strong energy efficiency because CPUs orchestrate tools, memory access and workflow around GPUs.

In the keynote, NVIDIA said Vera uses its Olympus core and a scalable coherency fabric. The company claimed Vera delivers 1.8 times the agentic sandbox performance of x86 CPUs and cited examples including SQL workloads and real-time stream processing for the New York Stock Exchange.

Huang said Vera will be a major growth driver for NVIDIA, adding that the company is building “millions and millions of Veras.” He said Taiwan’s ODMs and computer makers are part of the go-to-market effort.

AI Factories, Enterprise Agents and New PCs

NVIDIA also introduced DSX as a blueprint and operating framework for AI factories. The company described DSX Sim as an Omniverse-based digital twin tool to design and validate AI factories before construction, while DSX OS would provision, monitor and operate installed infrastructure. NVIDIA also discussed tools for power optimization, cooling and grid interaction.

For enterprise AI, Huang described the NVIDIA Agent Toolkit as a stack that includes models, harnesses, tools and runtimes. He said NVIDIA OpenShell is an open-source runtime designed to run agents securely in enterprises, with adoption expected from companies including Red Hat, Canonical and Microsoft.

NVIDIA also announced Nemotron 3 Ultra, which Huang described as an open model with training data and scripts made available. He said the model is based on a hybrid architecture combining state-space models with Mixture of Experts and is designed to be faster and cheaper to run than other open models, according to NVIDIA’s comparisons.

The company also highlighted a partnership with Cadence to build chip design agents. NVIDIA said the Cadence and NVIDIA design verification agent uses Codex, Cadence ChipStack, Nemotron and OpenShell to automate parts of RTL verification, test creation, regression testing and debugging. The company said verification cycles could move from weeks to hours.

Huang also said Microsoft and NVIDIA have been working to reinvent the PC for agentic computing. NVIDIA announced RTX Spark laptops and a broader line of Windows machines spanning desktop, laptop and workstation systems. The platform includes an N1X chip built with MediaTek, which Huang said supports NVIDIA’s software stack, CUDA applications, Windows applications and local or cloud-connected agents.

Physical AI, Autonomous Vehicles and Robotics

The keynote closed with a focus on physical AI. Huang said data is the hardest problem for robotics because robots need data from their own perspective, not only third-person video. NVIDIA announced Cosmos 3, which Huang called a frontier foundation model for physical AI. NVIDIA said Cosmos can function as a vision-language model, world model, simulator and foundation for action models.

NVIDIA also announced Llama Mio 2, described as an open model for self-driving cars, and demonstrated a reasoning autonomous vehicle system. Huang said automakers representing about 80% of the world’s cars have signed up for NVIDIA Hyperion, and that mobility services representing about 97% of the world’s mobility services are connecting with NVIDIA.

For humanoid robotics, NVIDIA announced Isaac GR00T reference design robots, a platform intended for higher education and university researchers. The system includes open models, simulation and training libraries, data generators and a robot computer running on Jetson Thor.

Huang summarized the event by saying the computing industry has changed in the last six months as agents became useful. He said the same agentic computing pattern will run across clouds, enterprises, PCs, robots, vehicles, satellites, base stations and factories.

About NVIDIA NASDAQ: NVDA

NVIDIA Corporation, founded in 1993 and headquartered in Santa Clara, California, is a global technology company that designs and develops graphics processing units (GPUs) and system-on-chip (SoC) technologies. Co-founded by Jensen Huang, who serves as president and chief executive officer, along with Chris Malachowsky and Curtis Priem, NVIDIA has grown from a graphics-focused chipmaker into a broad provider of accelerated computing hardware and software for multiple industries.

The company's product portfolio spans discrete GPUs for gaming and professional visualization (marketed under the GeForce and NVIDIA RTX lines), high-performance data center accelerators used for AI training and inference (including widely adopted platforms such as the A100 and H100 series), and Tegra SoCs for automotive and edge applications.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.