Huawei expects its AI chip revenue to grow at least 60% in 2026 to approximately $12 billion, driven by Chinese companies shifting rapidly toward domestic processors as US export controls keep Nvidia's most capable chips out of the Chinese market, with the Ascend 950PR now in mass production and an upgraded 950DT version planned for the fourth quarter.
The Financial Times reported the figures, picked up by Reuters today, and the growth trajectory they describe is remarkable by any standard. Huawei's AI chip business generated $7.5 billion in revenue in 2025. Getting to $12 billion in a single year requires adding nearly $4.5 billion in new revenue from a product category that did not exist in its current form three years ago. The company is not doing this by winning new markets that Nvidia previously ignored. It is doing it by absorbing demand that would have gone to Nvidia if the export control environment had not closed that door. Understanding that distinction is essential to reading what this number actually means for the global AI infrastructure landscape.
US export restrictions on advanced AI chips to China have been progressively tightened since 2022, with each successive round closing loopholes that Nvidia had used to continue selling modified versions of its hardware to Chinese customers. The H20, which Nvidia had designed specifically to comply with earlier export thresholds, was effectively blocked from the Chinese market earlier this year. That decision accelerated a transition that Chinese technology companies had already begun but had not yet been forced to complete. The shift from imported AI compute to domestic alternatives is now less a strategic preference and more an operational requirement, and Huawei's Ascend series is the primary beneficiary.
Most of Huawei's 2026 AI chip orders are reportedly tied to the Ascend 950PR, which entered mass production in March. Mass production is a milestone that is easy to understate. Designing a competitive AI chip and demonstrating it in a lab environment is one challenge. Manufacturing it at the volume required to serve hyperscale cloud customers and enterprise AI deployments is an entirely different one, and it is the challenge that has historically separated credible chip companies from well-funded aspirants. The fact that the 950PR is in mass production and generating orders at a scale that supports a $12 billion revenue forecast means Huawei has cleared that manufacturing hurdle, at least at current demand levels.
The planned 950DT upgrade for the fourth quarter adds another dimension. In the AI chip market, a company that ships only one product generation per year loses ground quickly because the performance envelope of the workloads it needs to support is expanding continuously. An upgraded variant arriving within the same calendar year as the base product suggests Huawei's engineering and manufacturing cadence is moving faster than outside observers had assumed. It also gives Chinese cloud providers and enterprise customers a reason to plan longer-term procurement roadmaps around Ascend rather than treating domestic chips as a temporary workaround while they wait for geopolitical conditions to change.
The customers driving this demand are not small companies hedging their chip procurement. Alibaba Cloud, Baidu, ByteDance, and Tencent have all been reported as significant buyers of Huawei's Ascend hardware, and state-backed cloud and AI infrastructure initiatives have added further institutional demand. These are organizations with the engineering depth to integrate new chip architectures into their training and inference workflows, the scale to absorb large hardware commitments, and the political alignment with Beijing's broader technology self-sufficiency agenda to make domestic procurement a strategic priority rather than just a cost calculation.
The structural shift underneath the revenue number
The $12 billion forecast matters beyond Huawei's own financials because it is a proxy for the speed at which China's AI infrastructure stack is decoupling from Western technology. AI chips are the most visible layer of that stack, but the nationalization dynamic extends through the full architecture. Chinese AI model development has accelerated significantly over the past eighteen months, with Deepseek's performance attracting international attention and a growing list of domestic foundation models reaching competitive capability benchmarks. Chinese cloud platforms are expanding their AI services on domestic infrastructure. The software frameworks and toolchains that sit between hardware and models are being developed and optimized for Ascend and other domestic processors rather than for CUDA, Nvidia's proprietary software ecosystem that has been the de facto standard for AI development globally.
This matters for the global AI industry because it means the Chinese AI market is not simply adopting a version of the same technology stack that Western companies are building on. It is developing a parallel stack, with different hardware, different optimization frameworks, and different performance tradeoffs. Companies that want to compete in China's AI market will increasingly need to build for that stack specifically rather than assuming that CUDA-optimized software will port cleanly to a Huawei Ascend environment. That adds engineering cost and complexity to any strategy that involves both markets.
For Nvidia, the revenue loss from China is significant but the company has partially offset it through surging demand from US and international hyperscalers. The longer-term concern is that a Chinese AI infrastructure ecosystem that develops independently of Nvidia's hardware and software creates a structural competitor that does not depend on Nvidia for anything and has every incentive to continue improving its own stack. Huawei's 60% revenue growth in a single year is a data point about how quickly that alternative ecosystem is forming. The direction of travel is clear, and the pace is faster than most Western analysts had projected even twelve months ago.
Also read: China's four-month AI crackdown signals that compliance is now a core operating requirement for every platform in the market • Calligo Technologies is raising up to $15 million to prove that India can build the chips powering the next wave of AI infrastructure • Beacon Biosignals is turning sleep into a clinical data platform for the most underfunded frontier in medicine
