Nvidia Vera CPUs Dodge Export Controls While Chinese Buyers Rush Orders

Published by James Harris on June 12, 2026June 12, 2026

Nvidia is quietly taking orders from Chinese customers for its Vera server CPU, a chip category that currently sits outside the most restrictive U.S. export control rules, and expects $20 billion in Vera revenue by January 2027. This is not a return to the Chinese AI accelerator market. It is an attempt to enter a different market entirely before that door closes too.

The distinction matters because Vera is a CPU, not a GPU, and the regulatory line between the two has historically lagged the technology. The same pattern played out with Nvidia’s A100 and H100 chips: the Commerce Department tightened rules only after those chips had already become central to Chinese AI infrastructure. Nvidia is now moving into server processors precisely because the controls have not caught up, and the timing is deliberate. Intel flagged six-month lead times on server CPUs for Chinese customers in February; AMD called the global CPU market “tight” last month. Nvidia is walking into a supply gap with a chip that, on paper, is 1.8 times faster than comparable x86 processors on AI agent workloads.

At least one Chinese cloud provider is planning to deploy Vera chips in data centers outside China, which is the clearest signal that domestic political risk is already being priced into the purchasing decision.

The geopolitical ceiling here is real and probably closer than Nvidia’s revenue projections assume. If U.S. regulators conclude that Vera meaningfully accelerates AI inference capacity in China, an updated entity list or new licensing tier could arrive faster than the $20 billion forecast can materialize. Chinese buyers appear to understand this: treating Nvidia hardware as a bridge technology while domestic alternatives from Huawei and others mature is a rational procurement strategy, not loyalty. Nvidia hiring Bruce Andrews, a former Commerce Department official, to run government affairs suggests the company is already lobbying to keep CPUs off the restricted list, which implies internal acknowledgment that the risk is live.

Nvidia’s Q1 fiscal 2027 results showed $81.6 billion in total revenue with data center contributing $75.2 billion, so Vera is a growth bet on top of an already dominant position, not a lifeline. Whether Alibaba and ByteDance, both named as collaborators on Vera deployments, convert early testing into large-scale orders will be the first real read on whether the CPU exemption holds long enough to matter.