July 8, 2026, (Inside AI) — Perplexity confirmed on Tuesday that it plans to integrate Nvidia's new Vera central processing units into its infrastructure. The AI startup is betting that the chip, designed for the relentless demands of AI agents, will accelerate its core workloads.
Nvidia expects to generate $20 billion in sales from Vera by the end of this fiscal year. The CPU is a more generic computing chip than its AI-specific offerings, marking a strategic push into a market long dominated by Intel and Advanced Micro Devices.
The move comes as AI companies like OpenAI and DeepSeek develop their own custom chips, threatening Nvidia's data center GPU stronghold. By targeting the CPU market, Nvidia aims to diversify revenue and embed itself deeper into the AI infrastructure stack.
Perplexity Vice President for Computer Enterprise and Infrastructure Nate Kupp said Vera excelled at AI agent coding tasks, performing about 1.5 times faster than traditional CPUs. The sustained, non-stop nature of agent workloads exposes the limitations of legacy chips designed for intermittent human use.
"Vera really stood out to us as just like a dead-on fit for a lot of the core workloads that we have," Kupp said in an interview.
Perplexity declined to disclose the volume of its planned purchase. Nvidia has previously revealed that OpenAI, Anthropic, and Oracle are also Vera customers, signaling broad industry appetite for a CPU built for the agentic era.
Why Traditional CPUs Stumble with AI Agents
AI agents operate without pause, executing complex, multi-step tasks continuously. This contrasts sharply with human-driven workloads, where idle time allows conventional processors to manage thermal and power envelopes effectively.
Intel and AMD have optimized their architectures for bursty, interactive use. Vera, however, is engineered for persistent throughput, a critical advantage when agents run inference, orchestration, and tool-calling loops around the clock.
Nvidia's entry challenges the x86 duopoly at a pivotal moment. The rise of agentic AI is reshaping data center requirements, and Vera's design reflects lessons from GPU-accelerated computing, emphasizing memory bandwidth and parallel task handling.
While Nvidia's $20 billion sales target is ambitious, it underscores the scale of the opportunity. The company is leveraging its CUDA ecosystem and AI partnerships to position Vera as the default CPU for next-generation AI workloads.
The Agentic Shift Redraws the Silicon Map
Perplexity's adoption of Vera highlights a broader industry inflection point. As AI agents become the primary consumers of compute, the infrastructure must evolve from serving human-paced interactions to sustaining machine-speed autonomy.
Nvidia's previous disclosures that OpenAI, Anthropic, and Oracle plan to use Vera suggest a coordinated shift among AI leaders. This could accelerate the erosion of Intel and AMD's server market share if Vera delivers on its performance promises.
However, Intel and AMD are not standing still. Both have announced upcoming architectures with enhanced AI acceleration and efficiency cores. The battle for the AI data center will hinge on total cost of ownership and software maturity.
Perplexity's early commitment gives Nvidia a crucial reference design win. As the startup scales its AI-powered search and enterprise tools, Vera's real-world performance will be closely watched by the industry.
Kupp's endorsement of Vera as a "dead-on fit" suggests that Nvidia has successfully tailored the chip for the unique demands of AI-native companies. The coming quarters will reveal whether this bet pays off in market share and revenue.
