NVIDIA’s GTC Reveals Vera CPU To Expand AI Knowledge And Capabilities
During NVIDIA’s GTC (Graphics Processing Unit Technology Conference) the new NVIDIA Vera CPU, built on the success of Grace, was also announced. Vera is designed for AI servers to support their autonomous system, which is built to be capable of perception and reason, acting to achieve goals with limited human interaction. This is set to be groundbreaking in the computing industry.
NVIDIA’s GTC Breakdown
At the annual NVIDIA GTC, CEO Jensen Huang made an announcement that could change the way advancements are made and the way artificial intelligence is used. The entire presentation touched on many facets of what NVIDIA has been doing, from touching on robotics, graphics, medicine, stepping up the computing game, and how efficiently cloud services could work. Stepping beyond generative computing and into AI learning to read, code, and implement.
During their reveal, they also announced the Vera CPU, an efficient Arm-based processor similar to those in high-end smartphones. It is meant to support agentic AI, or AI that takes action. Unlike typical AI that relies on a question, agentic AI browses the web or uses software to complete a collection of tasks on its own; it isn’t just answering the user, it works for them. It will also be used in support of physical AI, or digital clones of the real world, as NVIDIA integrates Vera with Omniverse, allowing the AI to learn in a digital copy of a location before being put in a real robot.
NVIDIA Shifting The Workload
Vera is being dubbed a new class of CPU that processes, trains, and generates data more efficiently and responsively for large-scale services. Whether assisting with coding or being used at a consumer level, the new Vera CPU is expected to have the highest single-thread performance. NVIDIA already has cloud providers working alongside them to deploy, along with consulting and technology firms. With over 8 different collaborators backing the adoption of Vera as a new standard for AI workloads, it will democratize access to AI and create a surge of innovation in the field.
“Vera is arriving at a turning point for AI. As intelligence becomes agentic — capable of reasoning and acting — the importance of the systems orchestrating that work is elevated. The CPU is no longer simply supporting the model; it’s driving it. With breakthrough performance and energy efficiency, Vera unlocks AI systems that think faster and scale further.”
- Jensen Huang, founder and CEO of NVIDIA
Why Vera CPU Matters
NVIDIA is creating a new Vera CPU rack, a high-powered server central hub for AI, that will house 256 CPUs with liquid-cooling. Each will be capable of efficiently running close to 100 projects at the same time without bogging down or lagging. With so many CPUs processing so many tasks, one cabinet will be capable of handling over 22,500 independent AI tasks at once. That news means they will be able to compute, execute, and innovate far beyond what any team can do in real time while having access to vast amounts of data and endpoints to make sure it works.
This rack will be built with their base blueprint, their MGX Architecture, already being used by 80 other companies around the world. Alongside their MGX Architecture, they’re creating the NVIDIA Vera Rubin NVL72 platform where they are paired with NVIDIA GPUs through their NVLink, which is a high-speed GPU-to-GPU interconnect technology, enabling faster communication and multiple GPUs to function as one unified accelerator. This will allow system control and coordinated data movement from the accelerated workloads.
This will bring forth the most advanced AI computing the world has ever seen, not just in what the system outputs but in what it will be capable of learning and implementing.
