FULL STORY

Nvidia's Vera CPU Launch and Perplexity Testing

Nvidia launched its agent-optimized Vera CPU, with Perplexity's subsequent tests confirming major speedups in agentic coding tasks.

2026-07-07 ~ 2026-07-13 · 2 episodes · 7 posts

Episode 1 · NVIDIA Unveils Vera CPU to Tackle Agentic AI Bottlenecks (2026-07-07, 5 posts)

NVIDIA has unveiled the new Vera CPU architecture, aiming to solve the systemic execution bottleneck in next-generation agentic AI. NVIDIA points out that agentic systems execute tasks sequentially, with each reasoning step, tool call, and code execution happening on the CPU one by one. When the CPU slows down under high load, the entire agent loop is delayed, leaving expensive GPUs idle. Therefore, maximizing single-thread CPU efficiency is the key to unlocking compute power for large-scale AI workloads.

Industry Response and Testing

Positioned as the "strongest single-thread" processor at scale, the Vera CPU has already garnered industry attention. According to @AravSrinivas, their team is collaborating with NVIDIA to run sandbox infrastructure on Vera CPUs to support agent execution environments. Additionally, @TheTuringPost noted that AI companies like Perplexity have already begun testing the Vera CPU. Alongside Vera, NVIDIA also introduced the next-generation Rosa CPU, paired with Rigel cores.

Clarifications and Scope

To address potential external confusion, @pstAsiatech specifically emphasized that Vera, as a CPU, and Rubin, as an AI accelerator, address entirely different control problems. Vera CPU's role lies in agent workflow orchestration, ETL, data movement, CPU-side inference, and feeding data to GPU clusters; it does not itself represent heavy HBM tensor compute capabilities.

Episode 2 · Perplexity Tests Nvidia Vera CPU (2026-07-13, 2 posts)

Perplexity plans to adopt Nvidia's new standalone Vera CPU, claiming it executes agentic coding tasks 1.5 times faster than traditional CPUs.