Nvidia releases new details of Vera CPU, AI server market opens new battlefield
[NVIDIA releases new details of Vera CPU, the AI server market opens a new battlefield] On Tuesday (July 21), NVIDIA announced more details of its data center CPU "Vera", including technical specifications, performance test results, chip architecture and other information to help potential customers comprehensively evaluate the product. Nvidia said that Vera chips were delivered to customers including OpenAI, Anthropic, and SpaceX in June this year.
On Tuesday (July 21), Nvidia announced more details about its data center CPU "Vera", including technical specifications, performance test results, chip architecture and other information to help potential customers fully evaluate the product. Nvidia said that Vera chips were delivered to customers including OpenAI, Anthropic, and SpaceX in June this year. NVIDIA has always been regarded as a bellwether in the information technology industry, but it is still in a challenger position in the CPU field and needs to compete with two mature competitors - Intel and AMD. These two companies have been deeply involved in the server CPU market for a long time and have established deep cooperative relationships with ultra-large-scale cloud computing manufacturers. However, some cloud service providers may choose to deploy NVIDIA CPUs in the future, which will become a new battlefield. Adding more CPU is also another manifestation of Nvidia’s vertical integration strategy. The company hopes to independently develop more chips and technologies needed within the system every year, and eventually sell Rack computing systems to customers instead of just selling individual chips. NVIDIA believes that this strategy can further release GPU performance and help its artificial intelligence (AI) system continue to become the platform of choice for leading AI laboratories in the face of competition from AMD and various customized AI chips. Nvidia said last week that Vera is the company's first server processor designed independently from the CPU core, rather than using off-the-shelf core designs provided by Arm. Therefore, R&D investment is higher, but the degree of autonomy and control is also stronger. The company stated that Vera is specifically optimized for performance bottlenecks during the operation of AI agents. Compared with CPUs using the x86 architecture (that is, the architecture used by Intel and AMD), the performance on AI agent tasks is improved by about 50%. Nvidia pointed out that in the past, cloud computing CPUs emphasized the number of cores, while the Olympus core used by Vera focused more on single-core performance. Hannah Coutand, head of product marketing for Vera at Nvidia, said that Vera focuses on optimizing single-core speed, memory bandwidth and latency performance, "allowing AI agents to return tasks to the GPU as quickly as possible, thereby maximizing the utilization of the GPU, the most expensive and most important asset in the AI factory." Nvidia said Vera can be sold as a standalone CPU or used with the company's GPUs. Product forms include: Vera CPU sold separately; A liquid-cooled cabinet system composed of 256 Vera chips; Server configuration equipped with two Vera CPUs; Vera Rubin system with NVIDIA GPU. Vera has higher power consumption, ranging from about 250 watts to 450 watts. At the same time, it supports up to 1.5TB of low-power memory (LPDDR), which is commonly used in laptops and smartphones. Before the rise of the AI wave, the CPU had always been the most important and expensive component in the server. When ChatGPT is released in 2022, the first-generation AI servers usually adopt the architecture of "one CPU with up to eight GPUs", marking the beginning of the shift in computing focus to NVIDIA GPUs. But with the rise of Agentic AI (intelligent AI) that can run autonomously and requires only minimal human intervention, the importance of the CPU has once again increased. The CPU is responsible for managing the operation of the agent and continuously providing data to it, so it is back in the spotlight. Capital markets have also noticed this trend. As of 2026, AMD and Intel have risen 128% and 149% respectively, becoming the two best-performing chip stocks this year, while Nvidia has only risen 8% during the same period. Ian Buck, vice president of Nvidia's hyperscale computing business, said last week that agents make CPUs "more indispensable," especially the speed at which the CPU can answer a single request, becoming increasingly important. Nvidia estimates that the entire server CPU market is expected to reach US$200 billion in the future. Bernstein estimated earlier this year that the size of the mature server CPU market in 2025 would be approximately US$37 billion.
Wolfe Research estimated in May this year that the average selling price of Vera chips would be about US$5,000, and Nvidia would ship about 1.3 million units this year. Nvidia declined to comment on product pricing. Gartner analyst Kevin Knox said that AMD is still the strongest competitor in the enterprise AI server CPU market. According to reports, AMD currently holds about 33% of the server CPU market share, with Intel accounting for 66.8%, but AMD is continuing to expand its share and maintain close cooperation with hyperscale cloud computing customers. "AMD has done a great job building an ecosystem around its chips," Knox said. Some analysts believe Nvidia has effectively created a new CPU category that AMD and Intel don't yet have a counterpart to. Karl Freund, founder of Cambrian AI Research, said that Vera will not be used for traditional server tasks such as website hosting, but will be specifically targeted at high-intensity AI computing. However, even Nvidia still faces challenges in getting large cloud service providers to fully adopt this CPU. NVIDIA's Coutand said Vera is still in the early stages of market introduction. Apart from Oracle, there are no other large cloud service providers in the list of partners. However, OpenAI plans to deploy Vera chips on a large scale starting this quarter. Freund said: "The launch of the CPU is essentially to free customers from dependence on Intel and AMD CPUs and at the same time capture this revenue. What they chose to create is a unique CPU that is not currently offered by other manufacturers in the market." Dongcai Illustration·Add some useful information
Wolfe Research estimated in May this year that the average selling price of Vera chips would be about US$5,000, and Nvidia would ship about 1.3 million units this year. Nvidia declined to comment on product pricing. Gartner analyst Kevin Knox said that AMD is still the strongest competitor in the enterprise AI server CPU market. According to reports, AMD currently holds about 33% of the server CPU market share, with Intel accounting for 66.8%, but AMD is continuing to expand its share and maintain close cooperation with hyperscale cloud computing customers. "AMD has done a great job building an ecosystem around its chips," Knox said. Some analysts believe Nvidia has effectively created a new CPU category that AMD and Intel don't yet have a counterpart to. Karl Freund, founder of Cambrian AI Research, said that Vera will not be used for traditional server tasks such as website hosting, but will be specifically targeted at high-intensity AI computing. However, even Nvidia still faces challenges in getting large cloud service providers to fully adopt this CPU. NVIDIA's Coutand said Vera is still in the early stages of market introduction. Apart from Oracle, there are no other large cloud service providers in the list of partners. However, OpenAI plans to deploy Vera chips on a large scale starting this quarter. Freund said: "The launch of the CPU is essentially to free customers from dependence on Intel and AMD CPUs and at the same time capture this revenue. What they chose to create is a unique CPU that is not currently offered by other manufacturers in the market." Dongcai Illustration·Add some useful information