Nvidia is offering its new Vera central processing units for AI data centers to Chinese clients as early as August. The company is pivoting to this product to revive revenue in a market where its graphics processor shipments have stalled for months due to strict U.S. export controls.
Bypassing export restrictions
Nvidia's market share in China effectively fell to zero, Chief Executive Jensen Huang said in October. Beijing's push for self-reliance and U.S. restrictions halted shipments of the H200 graphics processor. Selling CPUs in China presents fewer regulatory hurdles. Washington has licensed about 10 Chinese firms to buy the H200, but no deliveries have been approved by Chinese officials.
The Vera chip is Nvidia's first standalone CPU built for agentic AI, handling the background computing these systems require. It is based on Arm technology, placing Nvidia in direct competition with Intel and AMD, which dominate the market with x86 architecture. Nvidia claims Vera runs up to 1.8 times faster than comparable rival processors.
Early adoption and testing
The sales process for complex AI hardware like the Vera CPU involves managing B2B technology sales and product demonstrations for enterprise data centers, which the AI Learning Path for Technical Sales Representatives addresses. A single Vera processor will cost "well north" of $20,000 before bulk discounts, according to SemiAnalysis. A fully configured rack of 256 chips runs to around $10 million, depending on memory configuration.
Nvidia expects $20 billion in revenue from Vera chip sales by the end of this fiscal year in January. One major Chinese cloud company plans to order more than 300 servers, each containing two Vera CPUs, for initial testing. Chinese clients intend to deploy these chips only in their overseas data centers first.
Widespread adoption faces hurdles related to software ecosystems and compatibility. Migrating workloads originally built around domestic AI chips remains a constraint for some buyers. Alibaba and ByteDance were collaborating with Nvidia to deploy Vera, though they did not respond to requests for comment on the actual ordering process.
Global CPU supply constraints
The push for Vera coincides with a global shift from AI model training to inference computing, where graphics processors face greater competition from custom chips and CPUs. This transition has created a CPU shortage. Intel recently notified Chinese customers of server CPU delivery lead times of up to six months. AMD also flagged that the global CPU market is "tight," with demand outpacing forecasts.
Nvidia's effort to capture new revenue and rebuild market share highlights the pipeline management and revenue growth strategies covered in the AI Learning Path for Sales Managers. Executives are actively restructuring their go-to-market approaches to address these new hardware demands.
Why this matters for sales professionals
Sales teams must adapt to a market where AI infrastructure buyers are shifting from graphics processors to central processing units for inference workloads. Understanding the pricing models, deployment constraints, and regulatory hurdles of enterprise hardware will be critical for closing high-value data center contracts.
Your membership also unlocks:
