Nvidia says it’s ‘absurd’ to assume it predicted the agentic moment during Vera’s development – ‘product intuition’ helped drive CPU design, superior memory bandwidth will play critical role in future AI agents, CPU boss claims
Reported by and
AMD has criticized Nvidia, and to a lesser extent Arm, for claims about creating a CPU designed for agentic AI.
Key points
- Vera development began four years ago, before Nvidia says it saw the current agentic moment
- Finder said the team had noticed early signs such as tool calling
- Nvidia designed Vera to work alongside its accelerators, not specifically for today’s agent workloads
- Finder says Vera’s memory bandwidth could become more important as agent systems evolve
Sources
- Tom's HardwareRead ↗Nvidia says it’s ‘absurd’ to assume it predicted the agentic moment during Vera’s development – ‘product intuition’ helped drive CPU design, superior memory bandwidth will play critical role in future AI agents, CPU boss claims
- news.lavx.huRead ↗Nvidia CPU Chief Rejects Agentic-CPU Label for Vera
Details
In an interview with Tom’s Hardware at SC26, Nvidia data center CPU business head Ian Finder rejected the idea that the company anticipated the current agentic AI wave when Vera development began. He called that assumption “absurd,” while saying Nvidia had seen early signs of workloads involving tool calling and understood what it needed from a head node.
Finder described Vera as an extension of Nvidia’s CPU-and-accelerator co-design approach, which began with its Grace data center CPU. He said the design focused on keeping accelerators supplied with work, and cited memory bandwidth, loaded memory latency, single-thread performance and concurrency as considerations.
Finder argued that agent workloads require CPUs as well as GPUs: CPUs handle tasks such as starting agent environments and compiling or testing code, while GPUs run models. He said Vera’s memory bandwidth could matter more as agent systems develop, but Tom’s Hardware’s reported benchmarks show Vera does not lead in every workload and AMD’s EPYC Venice has regained top positions in several benchmarks.
Written from coverage as of
Images
- Tom's Hardware
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh