The AI search domain leader Perplexity adopts the Vera CPU! Under the wave of intelligent agents, NVIDIA Corporation (NVDA.US) aims at the $200 billion general computing battlefield.

Focus on AI startups driven by AI large models search engine - AI search leader platform Perplexity AI confirmed on Tuesday local time that it plans to use the data center CPU, Vera CPU, exclusively built by NVIDIA Corporation (NVDA.US) on a large scale. Currently, this global company with the highest market value and a super giant in AI chips is working hard to expand its market share and challenge Intel Corporation (INTC.US) and AMD (AMD.US), the two super giants with profound roots in the data center CPU field based on the x86 architecture. NVIDIA Corporation is actively upgrading itself from being the "super king of AI GPU domain" to being the supplier of AI era computing infrastructure as the "AI data center CPU+GPU+high-performance network infrastructure+AI computing system at the rack level". This move comes as a result of the increasing demand for AI computing power and the AI wave brought about by Agentic AI. With the explosive launch of AI agents such as Claude Cowork from Anthropic and OpenClaw in 2026, the wave of AI agents is rapidly sweeping the world. The bottleneck in AI computing architecture is shifting from GPUs, which are focused on matrix multiplication throughput, to data center CPUs focused on control flow, task scheduling, and memory/IO coordination. High-performance CPUs for ultra-large-scale AI data centers are facing real challenges in supply shortage. As the wave of AI agents spreads globally, the main investment focus in AI computing power is shifting from the "competition around GPU computing power" to the "full-stack computing power system driven by AI agents". The next round of super profits will no longer be limited to the strongest players in the AI GPU/AI ASIC domain but will systematically expand to data center high-performance CPUs, DRAM/NAND/HBM storage, AI PCBs, liquid cooling systems, data center optical interconnection systems, ABF substrates/glass substrates, MLCCs, electronic fabrics, and widespread wafer fabrication in the "AI factory" full-stack AI computing infrastructure layer. In this narrative shift, data center CPUs, optical interconnection, and memory chips may be the biggest power winners. From GPU dominator to data center CPU challenger, the wave of AI agents reshapes the demand for data center CPUs NVIDIA Corporation's management recently stated that they expect their "Vera" CPU product to generate approximately $20 billion in sales by the end of this fiscal year. Compared to their AI-specific GPU products, this is a more general-purpose computing chip. As companies like OpenAI, Anthropic, and DeepSeek focus on developing their own AI training/inference acceleration chips, Vera CPU chips are an important part of NVIDIA Corporation's efforts to diversify its sales. NVIDIA Corporation is entering a fiercely competitive field in the data center CPU market, long dominated by Intel Corporation and AMD, who supply high-performance CPUs for various devices from laptops to network servers and infrastructure. Many of the x86 architecture CPU chips were designed before the rise of AI "agents," which can execute extremely complex agent-based workflows upon receiving instructions from human users. AI agents capable of autonomously performing various tasks are likely to be the ultimate trend in AI applications over the next decade. The emergence of AI agents signals a shift in artificial intelligence from being an information assistance tool to becoming a highly intelligent productivity tool. This is also why the valuation of Anthropic has surpassed 1 trillion dollars and even exceeded OpenAI. Unlike human CPU users who take breaks between tasks, AI agents never rest. Nate Kupp, VP of Computing Enterprise and Infrastructure at Perplexity, stated in an interview that NVIDIA Corporation's exclusively designed CPU can perform AI agent encoding tasks 1.5 times faster than traditional CPUs. Kupp said, "Vera really impresses us because it precisely fits many of our core AI inference workloads." Perplexity, founded in 2022, has gained global developer and researcher recognition for its real-time, efficient, precise AI outputs and inference results, similar to a search engine user interface. The company's AI applications can conduct in-depth research and can access AI large models other than its own, including those from OpenAI and Anthropic PBC. Perplexity refused to disclose how many NVIDIA Corporation CPUs they plan to purchase. However, NVIDIA Corporation has revealed that companies like OpenAI, Anthropic, and Oracle Corporation all plan to purchase its Vera CPU on a large scale. The official list of Vera CPU buyers from NVIDIA Corporation includes companies such as Anthropic, OpenAI, SpaceXAI, CoreWeave, Oracle Cloud Infrastructure, Lambda, Nebius, and Nscale. AI chip giant NVIDIA Corporation moves into the data center CPU market! Intel Corporation and AMD face a new battlefront in the data center Compared to traditional x86 server CPUs from Intel Corporation and AMD, the core advantage of NVIDIA Corporation's Vera CPU is not simply "stronger general-purpose CPU performance," but rather the redesign of CPU's role in AI agent workloads in the AI factory. Traditional x86 server CPUs have long served general tasks such as databases, virtualization, web services, and enterprise applications. Meanwhile, Vera is designed for sandboxed code execution, tool calls, retrieval, data processing, task scheduling, and GPU orchestration in the AI agent loop. NVIDIA Corporation states that Vera can complete tasks 1.8 times faster than x86 CPUs across various AI agent workloads. Perplexity also claims that Vera performs AI agent encoding tasks about 1.5 times faster than traditional CPUs. This illustrates that Vera is truly entering into the intelligent agent closed loop of "GPU generation of next steps, CPU execution of actions, and feedback of results," rather than simply replacing enterprise servers in the market. In the intelligence agent loop, a large amount of workload is not only spent on generating tokens on GPUs, but also on CPU-dominant tasks such as Python interpretation, web scraping, database retrieval, RAG index access, lexical processing, task queue scheduling, RPC/IPC communication, KV state updates, and more. This implies that what determines user experience is increasingly not just the peak computing power of a single GPU, but whether the CPU has sufficient core counts, thread concurrency, cache levels, memory bandwidth, PCIe/CXL/interconnect scheduling capacity to support high-frequency tool calls and high-density task switches. If CPU cores, memory subsystem, or I/O scheduling are insufficient, even with ample nominal GPU computing power, utilization can collapse due to data preparation, task coordination, and system waiting. Therefore, there is no doubt that the bottleneck of AI computing architecture is shifting from GPUs, which focus on matrix multiplication throughput, to data center CPUs focusing on control flow, task scheduling, and memory/IO coordination. This change is rooted in the fundamental shift in workload paradigm. CPUs are no longer just general computing chips but control plane processors, system orchestration engines, and resource scheduling centers of the intelligent agent era. The "underestimated CPU becoming the new bottleneck in AI" is not an emotional assessment but a necessary outcome as AI workloads evolve from "inference computing problems" to "complex system engineering problems." From a hardware platform perspective, Vera's advantages lie in high bandwidth, low latency, GPU collaboration, and energy efficiency density. NVIDIA Corporation discloses that Vera Rubin NVL72 integrates 72 Rubin GPUs, 36 Vera CPUs, ConnectX-9 network cards, and BlueField-4 DPUs into a rack-level AI supercomputing system. A single Vera Rubin Superchip contains 88 proprietary Olympus Arm-compatible CPU cores, 1.5TB of LPDDR5X CPU memory, and provides 1.8TB/s bandwidth through NVLink-C2C. Compared to traditional server architectures where x86 CPUs cooperate with GPUs through PCIe, Vera's value lies in packaging CPU, GPU, DPU, network, and memory into a unified AI factory system, reducing data transfer bottlenecks and increasing token output per megawatt. NVIDIA Corporation states that Vera Rubin NVL72 can reduce the cost of highly interactive deep inference AI agents per million tokens by one-tenth compared to GB200 NVL72 and achieve up to 10 times higher token output per megawatt. Huang Renxun recently stated in an interview that Vera enables NVIDIA Corporation to enter a new CPU service market worth around $2 trillion. Among Wall Street analysts, the most optimistic target price for NVIDIA Corporation is as high as $500, set by the senior analyst Tristan Gerra from Baird. The rating remains "outperform the market." If the stock reaches $500, the corresponding market value will be about $12.2 trillion (the current market value is approximately $4.8 trillion), indicating a potential upside of about 154% from the current stock price. Baird's core bullish logic is not just based on the booming demand for GPU, but on the belief that NVIDIA Corporation is expanding from being an "AI GPU supplier" to a full-stack AI computing infrastructure platform at the AI factory level, with emphasis on the Vera CPU opening up a new $2 trillion CPU growth opportunity.