Perplexity has announced it will run its AI agent workloads on Nvidia's new Vera CPU, an Arm-based general-purpose server chip that is Nvidia's bid to expand beyond GPU accelerators. Perplexity's VP of enterprise infrastructure reported Vera ran agentic coding tasks roughly 1.5x faster than traditional CPUs, citing strong single-threaded performance as a key fit for inference pipelines. Nvidia expects around $20bn in Vera CPU revenue this fiscal year, targeting a $200bn general-purpose computing market currently dominated by Intel and AMD. The move is partly defensive — several major Nvidia customers are building their own AI chips — and Vera lets Nvidia sell more of every rack regardless. Perplexity, handling over 400 million search queries monthly, sees faster CPUs as a way to reduce cost-per-query and fund agentic features without ballooning infrastructure spend.
