Perplexity Adopts Nvidia Vera CPUs for Agent Workloads

The AI startup Perplexity officially announced its plans to incorporate Nvidia’s new central processing units (CPUs), internally branded as "Vera," into its computing infrastructure. This decision aims to bolster the company's ability to manage complex, autonomous AI agent tasks. By shifting toward specialized processing power, Perplexity seeks to optimize the core workloads that drive its search and coding capabilities, setting a new benchmark for its server operations.

Nvidia, a dominant leader in the AI-specific chip market, designed the Vera CPU to address a growing demand for generic, high-performance computing power that differs from its traditional AI accelerators. With sales projections hitting $20 billion by the end of the fiscal year, this move signifies Nvidia's tactical push to challenge the entrenched market dominance of Intel and AMD. These chips are specifically optimized for the unique demands of AI agents, which run continuously and require significant computational efficiency compared to standard user interactions.

Perplexity’s leadership confirmed that the Vera chips perform coding tasks approximately 1.5 times faster than traditional CPU models. For the Indian market, this hardware transition highlights a broader industry shift toward AI-native infrastructure, as local tech hubs increasingly adopt specialized compute resources to scale generative AI services. As OpenAI, Anthropic, and Oracle also incorporate these chips, the race for infrastructure supremacy continues to reshape the landscape of high-performance computing in the ongoing AI arms race.