NVIDIA used its annual GTC conference to deliver the most comprehensive statement of strategic intent in the company's history, unveiling a seven-chip platform called Vera Rubin, claiming $1 trillion in demand for its infrastructure through 2027, and announcing partnerships that cement its position at every layer of the AI stack.
The Vera Rubin platform pairs new Rubin GPUs with Vera CPUs and the Groq 3 LPX inference accelerator, which NVIDIA says delivers up to 35x higher inference throughput per megawatt compared to prior generations. The system is designed to function as a single AI supercomputer across five rack configurations, addressing the scale requirements of what NVIDIA calls AI factories: data centers optimized for continuous AI inference rather than batch training. Dynamo 1.0, entering production simultaneously, serves as the software layer for these factories, boosting Blackwell GPU inference performance by up to 7x.
Jensen Huang, NVIDIA's CEO, told the conference that computing demand for its GPUs is "off the charts," citing a 40 million times increase in available compute over the past ten years and projecting at least $1 trillion in revenue from 2025 through 2027. A year ago, that figure was $500 billion.
The hyperscale commitments announced at GTC underpin that projection. AWS committed to deploying more than one million NVIDIA GPUs alongside Groq LPUs. Microsoft Azure, Google Cloud, and Oracle are also onboard for Vera Rubin deployments. The deals reflect an ongoing shift in cloud infrastructure: hyperscalers are no longer simply buying GPUs for training workloads but building out permanent inference capacity to support continuous agentic AI operations.
At the agent layer, NVIDIA announced NemoClaw, a one-command install stack for the OpenClaw agent platform. NemoClaw bundles NVIDIA Nemotron models with the new OpenShell runtime, adding privacy and security controls to autonomous agent deployments. Huang called OpenClaw "the operating system for personal AI" and "the fastest-growing open source project in history." The launch extends NVIDIA's infrastructure reach into the agent layer itself, giving the company a foothold in how agents are deployed and secured, not only the hardware they run on.
NVIDIA also launched a Nemotron Coalition with Mistral, Cursor, and Perplexity to advance open frontier model development -- a positioning move that addresses dependence on model providers who could theoretically steer customers toward competing hardware.
The agentic AI framing was consistent throughout. NVIDIA announced robotaxi deployments with Uber across 28 markets by 2028, expanded its industrial robotics ecosystem with FANUC, ABB, and Figure, and Huang called autonomous vehicles "the first multitrillion-dollar robotics industry."
What GTC 2026 clarifies is the shape of NVIDIA's competitive moat: the company is no longer competing on chips alone but on a vertically integrated stack running from silicon to software to agent deployment. The partners announced this week represent a dependency structure that will be difficult for any rival to replicate at comparable speed.
Sources: NVIDIA Newsroom, NVIDIA NemoClaw announcement, AWS
--
By the Control Plane Editorial Team