AMD's EPYC processors are now capable of handling every stage of agentic AI workflows, according to Quantum Zeitgeist. The report, published on 19 September 2026, describes the chips as covering the full span of these workloads rather than only parts of them.
The claim arrives alongside AMD's first official benchmark figures for its EPYC "Venice" CPUs, which Tom's Hardware reports are aimed squarely at Nvidia. According to Tom's Hardware, AMD says its 256-core Venice chip is more than twice as fast as Nvidia's Vera, while the 96-core model is 20 percent faster on a per-core basis.
Agentic AI refers to systems that carry out multi-step tasks with limited human involvement, a workload that places sustained demands on processors across planning, execution and supporting functions. AMD's positioning suggests it wants EPYC to serve that entire chain instead of a single portion of it.
The comparison with Nvidia is notable because Nvidia has been the dominant name in AI hardware. AMD's benchmarks, as reported by Tom's Hardware, frame Venice as a direct competitor at both the high core-count and per-core ends of the market.
Both the workflow claim and the benchmark numbers come from the companies and outlets named above; independent verification was not included in the available material.
What to watch next: whether third-party testing confirms AMD's Venice performance figures against Nvidia's Vera, and how Nvidia responds.