NVIDIA Vera CPU Demand Rises as Perplexity Backs Its AI Inference Performance
by Matt Lawrence · OnMSFTNVIDIA Vera CPU demand is rising as more AI firms look for faster processors built for inference and agentic AI workloads, with Perplexity now joining the list of companies backing NVIDIA’s latest data center CPU.
According to Reuters, Perplexity VP Nate Kupp said Vera was a strong fit for the company’s core workloads, as the chip delivered around 1.5 times faster performance than traditional CPUs in agentic AI coding tasks.
Why NVIDIA Vera Matters for AI Workloads
NVIDIA has positioned Vera as a CPU built for AI inference, where speed, predictable latency, and strong single-threaded performance matter more than simply adding more cores.
Agentic AI workloads run in repeated loops, where the CPU handles a task, sends results back, and then helps the model decide the next step. NVIDIA says conventional CPUs do not handle this pattern well because more cores do not always reduce the time each step takes.
Vera uses NVIDIA’s custom Olympus cores and brings 50% higher IPC than Grace, which means each core can handle more instructions while staying efficient under load.
Vera Focuses on Single-Threaded Speed
NVIDIA describes Vera as a “Max Single-Threaded” CPU at scale, with strong per-core performance, high memory bandwidth, and predictable latency across workloads.
The chip pairs its 88 cores with up to 1.2TB/s of LPDDR5X memory bandwidth while keeping memory power below 40W. NVIDIA also says Vera’s monolithic compute die provides 3.4TB/s of core-to-core bandwidth, which helps reduce bottlenecks during demanding AI tasks.
Perplexity’s early results show why AI companies are paying attention. NVIDIA also claims stronger results in concurrent sandboxes, SQL analytics, and real-time streaming workloads when compared with traditional x86 CPU offerings.
With Vera already reaching companies such as OpenAI, xAI, Oracle, Anthropic, and Perplexity, NVIDIA is now pushing deeper into the CPU market while preparing Rosa, its next-generation data center CPU with the newer Rigel core architecture.