Nvidia expects $20 billion in sales from its Vera CPU this fiscal year, as it enters the general-purpose server market long dominated by Intel and AMD. AI search firm Perplexity has become one of the first major AI names to adopt the chip, planning to run agentic coding workloads on Vera.
Nvidia frames Vera as a "max single-threaded CPU at scale" rather than a parallel-processing monster. Source: tomshardware Vera is an 88-core Arm design with SMT (176 threads), 1.2 TB/s LPDDR5X memory bandwidth, and 3.4 TB/s core-to-core bandwidth—claimed to be 3x that of any other data center CPU. Source: tomshardware However, AMD contested Phoronix benchmarks, asserting a 3.3x performance advantage for its own hardware in a 100 kW rack. Source: tomshardware
Vera targets AI workloads where sequential reasoning limits parallelism, making single-thread speed critical. The CPU push is partly defensive as Nvidia's largest customers design their own AI chips, allowing Nvidia to sell more of every rack even where its accelerators face competition. Source: thenextweb
“Nvidia is reasserting Vera's advantage for AI work by describing it with a new product category: a 'max single-threaded CPU at scale' rather than a parallel-processing beast.”
“Perplexity Vice President for Computer Enterprise and Infrastructure Nate Kupp said ... 'Vera really stood out to us as just like a dead-on fit for a lot of the core workloads that we have.'”
“Nvidia has said it expects to generate $20 billion in sales from its 'Vera' CPU, a more generic computing chip than its AI-specific offerings, by the end of this fiscal year.”
“AMD had some things to say about the Phoronix test, firing back with its own metrics of a 3.3x performance gain over Vera for the projected output of a 100 kW rack of its hardware.”