Nvidia deep dives Vera CPU for AI data centers — SPEC CPU 2026 benchmarks revealed, Olympus architecture specifics, and more

NVIDIA has unveiled its first custom-designed data center CPU, Vera, aiming to capture a growing share of the AI infrastructure market. While its predecessor, Grace, has seen success—particularly at M... Read original →

NVIDIA’s Vera CPU marks a strategic pivot from accelerator vendor to full-stack AI infrastructure architect. By ditching Arm’s reference cores for a proprietary microarchitecture, Vera outperforms AMD’s Epyc 9755 in SPEC CPU 2026 integer workloads with fewer threads—signaling deep optimization for low-latency, concurrent AI inference. This forces re-engineering across compilers, schedulers, and OS kernels. Geopolitically, reliance on TSMC (Taiwan, China) for 3nm EUV fabrication injects supply chain fragility into cost models. AMD will likely accelerate Turin-ROCm integration, while Intel may counter with Granite Rapids tightly coupled to Gaudi accelerators. Within 18 months, the server CPU market will bifurcate: x86 retains general-purpose dominance, while custom Arm-like cores target AI workloads—enabling NVIDIA to capture premium value in AI data centers and redefine chip economics.