Nvidia Is Trying to Get Back Into China and Its New CPU Might Be the Way In
With GPU sales effectively frozen, Jensen Huang's company is pitching a brand new chip to Chinese clients. Here's what's actually at stake.
For a company that recently became the most valuable in the world, Nvidia has a China-shaped hole in its business that it very much wants to fill.
US export controls have effectively locked Nvidia's most powerful AI chips out of the Chinese market. Its chief executive Jensen Huang acknowledged as much in October, stating bluntly that Nvidia's market share in China had fallen to zero. The H200, its second-most powerful AI chip, has been stalled for months, with Washington licensing about ten Chinese firms to buy it, but not a single delivery has actually been completed. Chinese officials, in no hurry to hand wins to American semiconductor companies when they're simultaneously pushing domestic alternatives, have withheld the approvals needed to move the process forward.
So Nvidia is trying a different door.
Enter Vera, a CPU, Not a GPU
The product Nvidia is now pitching to Chinese clients is called Vera, and it's notable for what it isn't as much as what it is. Vera is a central processing unit (CPU), not a graphics processing unit. That distinction matters enormously in the current regulatory environment.
GPUs are the chips that made Nvidia what it is today, and they're the chips facing the tightest US export controls. CPUs occupy different regulatory territory, making sales to China potentially far less fraught than anything in Nvidia's GPU lineup. It's not a clear path, nothing involving US export policy and China ever is, but it's a meaningfully different conversation.
Vera is Nvidia's first standalone CPU built specifically for agentic AI, the systems designed to perform tasks autonomously rather than simply responding to prompts. It's now in full production, with Nvidia claiming it runs up to 1.8 times faster than comparable processors from competitors. Jensen Huang unveiled it in March and described it as the company's next multibillion-dollar business, with Nvidia expecting $20 billion in revenue from Vera sales by the end of this fiscal year.
What Chinese Clients Are Actually Saying
Three sources familiar with the discussions told Reuters that Nvidia has informed Chinese clients that Vera could be available as early as August, with orders already being taken. The response has been cautiously positive.
One major Chinese cloud company is reportedly planning to order more than 300 servers, each containing two Vera CPUs, though the plan is to deploy those systems for testing first before committing to larger official orders. That's a meaningful signal of interest, but it's also a very measured one. Deploying for testing is not the same as deploying at scale.
There's also the question of where these chips actually get used. Chinese clients are planning to initially deploy Vera chips in their overseas data centres for testing, rather than domestically, a cautious approach that limits exposure to potential complications on both the regulatory and the operational side.
The Practical Obstacles
Interest is not adoption, and several real obstacles stand between Nvidia's Vera pitch and meaningful Chinese revenue.
The first is cost. A single Vera processor runs well above $20,000 before bulk discounts, and a fully configured rack of 256 chips would cost around $10 million, depending on memory configuration. That's significant capital expenditure for chips being deployed initially for testing rather than production workloads.
The second is ecosystem compatibility. Chinese AI infrastructure has been built increasingly around domestic chips, as part of the push for self-reliance that Beijing has been pursuing actively. Migrating workloads built for one chip architecture to another is never simple, and Vera is built on Arm technology rather than the x86 architecture that Intel and AMD have used to dominate the server CPU market for decades. That means new software compatibility questions, integration challenges, and the organisational friction of changing established technical infrastructure.
Whether initial testing interest converts to large-scale adoption depends heavily on how those integration challenges are resolved in practice.
The Bigger Market Context
NVIDIA's China push with Vera is happening against a broader shift in the AI industry that makes CPUs more strategically important than they've historically been.
The AI race is pivoting from model training, where GPUs have dominated almost completely, to inference computing, the process of actually answering queries and running AI agents in production. At the inference stage, GPUs face more genuine competition from CPUs and custom chips. That shift has driven a global CPU shortage that neither Intel nor AMD has been able to fully address. Intel has reportedly told Chinese customers to expect server CPU delivery lead times of up to six months. AMD has flagged that global CPU demand is outpacing forecasts, with supply constraints expected to persist.
That supply tightness creates an opening for Nvidia in a market where it previously had no CPU presence at all. Vera puts Nvidia in direct competition with Intel and AMD in a segment that's suddenly more strategically important than it's been in years.
What This Means for Nvidia
The China angle is important for revenue recovery, but the Vera story is bigger than one market. NVIDIA is making a serious move into CPU territory that it has never occupied before, using an AI-native architecture to challenge two companies that have held that ground for decades.
If Vera establishes itself in China, in global hyperscale deployments, or both, it extends Nvidia's AI infrastructure footprint well beyond the GPU dominance on which it's built its recent valuation. Jensen Huang called it the next multibillion-dollar business for a reason. The $20 billion revenue target for this fiscal year alone suggests the company's ambitions for the product are substantial.
For Chinese clients, the calculation is different. Domestic alternatives exist, and Beijing actively wants them to succeed. But in a global CPU shortage where even domestic champions are supply-constrained, a high-performance chip available in August with a familiar hyperscaler ecosystem behind it is worth at least testing.
The orders being placed now are small and cautious. The question 2026 will start to answer is whether caution gives way to commitment and whether Nvidia has found the route back into a market it can't afford to ignore.
About the Creator
Mark Lim
Hi I am mark an automotive student and a car, tech and food enthusiast ! Im gonna try and post daily & hope you enjoy what I write and do share my page with people you know. I would gladly appreciate it! Cheers
Enjoyed the story? Support the Creator.
Subscribe for free to receive all their stories in your feed.
Comments
There are no comments for this story
Be the first to respond and start the conversation.