By Zak Killian
Publication Date: 2026-05-23 14:00:00
A new report from GF Securities (via SeekingAlpha) claims NVIDIA is gearing up to use its June 1st Computex keynote to pitch its upcoming Arm-based Vera CPU as an x86 killer. The financial analysts claim NVIDIA will boast that Vera delivers up to “1.5x faster speeds, 2x the performance, and 4x the density per rack,” as compared to traditional x86 alternatives from AMD and Intel.
Before you assume NVIDIA has invalidated x86, it’s worth looking at the fine print. Vera isn’t even trying to fight the same war as AMD’s upcoming Zen 6 “Venice” or Intel’s Diamond Rapids. Instead, NVIDIA’s 1.5x performance claim is almost certainly hyper-focused on agentic AI inference. These are workloads where massive LLMs are actively executing loops and managing multi-tenant data pipelines. The Vera CPU architecture is heavily optimized for massive memory bandwidth and tight coupling with NVIDIA’s own Rubin GPUs via ultra-fast NVLink fabric.
When it comes to raw data center compute, AMD’s Zen 6 Venice is scaling up to a massive 256 cores per socket. By comparison, Vera packs a more modest 88 custom Arm cores. For heavy, generalized enterprise workloads, 88 Arm cores aren’t going to magically dust 256 cutting-edge Zen 6 cores, but NVIDIA’s platform gives it specific advantages in AI workloads. In short, they’re simply built for different jobs. GF Securities even notes that the rapidly expanding…
