Nvidia Offers Vera CPUs to China as GPU Export Restrictions Persist
Nvidia is positioning its Arm-based Vera server CPUs for the Chinese market, with shipments potentially starting as early as August, even as high-end GPU sales remain effectively blocked.
Nvidia is moving to supply its Arm-based Vera server CPUs to Chinese clients, with orders currently being accepted for shipments as early as August. This strategy follows the continued freeze on high-end H200 AI GPU exports to the region. According to Tom's Hardware, citing Reuters, Nvidia is prioritizing Chinese customers for Vera allocation despite a global server CPU shortage. This timeline aligns with Nvidia's broader rollout strategy discussed at GTC Taipei during Computex, where the company noted that Vera systems would reach global customers through system builders and cloud partners starting this fall.
The Vera CPU, which originated as part of the Vera Rubin superchip, was launched as a standalone product at GTC San Jose this March. Nvidia's rack design for the processor supports 256 liquid-cooled units, capable of sustaining more than 22,500 concurrent CPU environments. At Computex, Nvidia confirmed the chip has entered full production, citing performance metrics that claim 1.8 times faster task completion compared to x86 processors specifically for agentic workloads. This focus on agentic AI is critical, as the shift toward inference and agentic execution increases the demand for host processors capable of handling complex tool calls and code execution.
Adoption of Vera in China faces a bifurcated regulatory environment. While U.S. export restrictions on general-purpose CPUs are significantly lighter than those on high-end AI accelerators, the primary barrier for Nvidia in China has shifted. Although the U.S. government previously licensed approximately 10 Chinese firms to purchase H200 GPUs, none have been delivered because Chinese officials withheld approval, aiming to prioritize domestic silicon development. Consequently, sources indicate that initial deployments of Vera CPUs will be restricted to the overseas data centers of Chinese cloud providers to mitigate the risk of domestic interference.
The market context for this rollout is defined by tight supply. With Intel quoting lead times of up to six months and AMD reporting that CPU demand is outpacing forecasts, Nvidia’s ability to supply Vera could be a significant differentiator. The company expects the Vera product line to generate $20 billion in revenue by the end of its fiscal year in late January. Currently, Chinese cloud companies are testing more than 300 Vera servers, with at least one major provider planning to place an order, signaling a continued effort to maintain a foothold in the Chinese infrastructure market despite geopolitical constraints.