Nvidia Pitches Its Vera CPU to Chinese Clients, With Availability as Soon as August

The China Problem, Restated

Since CEO Jensen Huang acknowledged in October 2025 that Nvidia's market share in China had "effectively fallen to zero" on advanced AI chips, the company has been hunting for a legal workaround. US export controls blocked its flagship H100, then its H200. The result: one of the most profitable product lines in tech history largely locked out of the world's second-largest economy.

As of June 12, 2026, Nvidia appears to have found something to sell.

What Vera Actually Is

Vera is Nvidia's first standalone central processing unit built specifically for agentic AI — systems that perform tasks autonomously rather than just responding to prompts. Huang unveiled the Vera Rubin architecture at the Computex trade show in Taipei on June 1. According to Nvidia, Vera runs up to 1.8 times faster than comparable processors from Intel and AMD. The company says the chip is now in full production.

Crucially, Vera is a CPU, not a GPU. Export controls that killed Nvidia's China GPU sales do not, as of now, apply to general-purpose central processors.

The Sales Push

Three sources familiar with the matter told Business Times that Nvidia has told Chinese clients Vera units could be available as soon as August and that they can begin placing orders. The sources declined to be identified because the discussions are private. Nvidia declined to comment. Alibaba and ByteDance, which Nvidia publicly named in March as collaborators on Vera deployment, did not respond to requests for comment from Business Times.

One unnamed major Chinese cloud company is planning to order more than 300 servers, each containing two Vera CPUs, according to one of the sources. That company intends to deploy the systems for testing before committing to a full order.

Competitive Landscape

Vera puts Nvidia directly into CPU territory dominated by Intel and AMD. Both companies have been ramping server CPU supply for AI data centers, and both now have a motivated new competitor pitching Chinese cloud clients on delivery timelines measured in weeks, not quarters.

Huang said in March he expected Vera to become Nvidia's next multibillion-dollar business. That projection was made before it was clear the China push would happen this fast.

The Fair Counterargument

Skeptics have a real point worth stating clearly. A CPU is not a GPU. The reason Nvidia prints money on AI infrastructure is its graphics processors and the CUDA software ecosystem built around them — that is what trains large models and runs inference at scale. Vera handles the "agentic" coordination layer, not the heavy compute. Chinese cloud companies locked out of H100s and H200s are not getting a substitute. They are getting a different product that solves a different problem. Huawei's Ascend chips and domestic alternatives are already filling the GPU gap. Vera may find a market in China, but it does not restore what export controls took away.

What Isn't Answered Yet

The AASTOCKS report on the same story adds no material new sourcing beyond Business Times — it is essentially a headline relay with no additional attribution or data.

What neither source addresses: whether US export authorities have reviewed Vera's China eligibility, and whether the agentic AI use case could eventually trigger new controls. The Commerce Department has moved to restrict chips based on performance thresholds before. Nvidia is betting Vera's classification holds. If regulators decide agentic AI infrastructure deserves the same treatment as GPU compute, the August delivery window closes.

The more concrete near-term signal will be whether that 300-server test order from the unnamed Chinese cloud company converts to a full contract and whether Alibaba or ByteDance put a number on their Vera commitments before Nvidia's next earnings call.