Huawei is telling partners it expects revenue from its artificial intelligence processors to reach roughly $12 billion in 2026, a figure that would represent about a 60 percent jump from the prior year, according to Reuters and supply-chain sources cited by multiple outlets. The projection has not appeared in any public Huawei filing, and the company has not commented on it directly. But the number has circulated widely enough among analysts and procurement officials that it now functions as a benchmark for how fast China’s domestic chip ecosystem believes it can grow under the shelter of tightening U.S. export controls.
The backdrop is concrete: Washington has made it significantly harder for Nvidia and AMD to sell their most powerful processors into China, and Beijing is steering its largest technology companies toward homegrown alternatives. Huawei’s Ascend-series chips sit at the center of that shift.
Washington tightens the valve
In May 2025, the U.S. Bureau of Industry and Security formally revised its licensing framework for advanced semiconductors headed to China. Under the updated policy, chips such as Nvidia’s H200 and AMD’s MI325X now require individual license applications subject to case-by-case review. Each application must clear several conditions: independent third-party testing of the chips, verification of the buyer’s compliance history, and confirmation that adequate U.S. domestic supply exists before any units ship abroad.
The Associated Press confirmed that while H200 sales received a conditional greenlight, the approval path is narrow and slow. For Chinese cloud providers and AI labs operating on aggressive deployment schedules, the added weeks of review and compliance paperwork create a practical incentive to source locally.
Separately, the Commerce Department issued broader clarifications covering not just finished processors but also the lithography tools and components used to manufacture them. That tightens the entire supply chain, raising the bar for any Chinese entity trying to import high-end computing hardware or the equipment needed to build it domestically.
Huawei’s Ascend chips fill the gap
Huawei’s answer to the export squeeze is the Ascend 910 family, designed by its HiSilicon subsidiary and fabricated by Semiconductor Manufacturing International Corporation (SMIC). The Ascend 910B began shipping in volume in 2024, and an upgraded 910C variant has entered testing with select customers, according to reporting from Reuters.
By most independent estimates, the Ascend 910B trails Nvidia’s H100 in raw training throughput by roughly 20 to 30 percent, and the software ecosystem around Huawei’s CANN framework remains less mature than Nvidia’s CUDA platform, which has two decades of developer tooling behind it. But for Chinese buyers, availability now outweighs peak performance. Baidu, ByteDance, and China Telecom have all placed significant Ascend orders, Reuters and The Information have reported, driven partly by procurement mandates from Beijing that favor domestic silicon in government-linked projects.
That policy tailwind is substantial. China’s State Council and Ministry of Industry and Information Technology have issued guidance directing state-owned enterprises and public-sector cloud providers to prioritize domestically designed chips wherever feasible. For Huawei, those directives function as a guaranteed demand floor that no amount of Nvidia discounting can undercut.
Nvidia feels the drag
Nvidia has acknowledged the pressure. During its fiscal Q1 2026 earnings call in late May 2025, CEO Jensen Huang noted that China-related revenue remained “subject to various different regulations” and that the company was working through the new licensing requirements. Nvidia did not break out a precise China revenue figure, but analysts at Bernstein and Morgan Stanley estimated that the region accounted for roughly 10 to 12 percent of data-center revenue in the quarter, down from above 20 percent before the first round of export controls in October 2022.
The company has tried to thread the needle by developing different chip variants that fall just below restricted performance thresholds. Its H20 processor, designed specifically for the Chinese market, is technically eligible for export under current rules. But the H20 is far less powerful than the H200, and Chinese buyers have shown limited enthusiasm for a product they view as deliberately hobbled. Meanwhile, every quarter that Nvidia spends navigating compliance is a quarter that Huawei uses to deepen its foothold with the same customers.
The manufacturing question
Whether Huawei can actually deliver $12 billion worth of AI chips depends on a bottleneck it does not fully control: SMIC’s fabrication capacity. The Ascend 910B is produced on SMIC’s 7-nanometer-class process, which the foundry achieved without access to extreme ultraviolet (EUV) lithography equipment blocked by U.S. and Dutch export restrictions. Yields on that process have improved but remain below what leading-edge fabs like TSMC achieve, according to semiconductor analysts at TechInsights.
SMIC has been expanding capacity at its Shanghai and Beijing facilities, and Huawei is believed to have secured priority allocation. Still, scaling from roughly $7.5 billion in AI chip revenue (the implied 2025 base if the 60 percent growth figure is accurate) to $12 billion requires a meaningful jump in wafer output at acceptable yields. Any production hiccup, equipment maintenance delay, or further tightening of tool exports from the Netherlands or Japan could force Huawei to revise its targets downward.
There is also the question of advanced packaging. High-performance AI chips increasingly rely on sophisticated packaging techniques such as chip-on-wafer-on-substrate (CoWoS), an area where TSMC holds a commanding lead. SMIC and Chinese packaging houses are investing heavily, but capacity constraints in this segment could become a secondary chokepoint as Huawei ramps volume.
What the next two quarters will reveal
The $12 billion target is best understood as a signal of ambition rather than a guaranteed outcome. It reflects real demand from Chinese enterprises hedging against U.S. supply risk, real policy support from Beijing, and real progress in Huawei’s chip design capabilities. But it also depends on manufacturing execution that has not yet been proven at the required scale.
Several data points in the second half of 2026 will clarify the picture. Nvidia’s quarterly earnings will show whether its China data-center revenue stabilizes or continues to erode. SMIC’s capacity disclosures and capital-expenditure guidance will indicate whether fabrication can keep pace with Huawei’s order book. And large procurement announcements from Chinese cloud providers and state-backed AI initiatives will reveal whether Ascend chips are winning contracts on merit or primarily on regulatory default.
What is already clear is that the competitive landscape has structurally shifted. U.S. export controls have moved from a temporary disruption to a permanent feature of the semiconductor trade, and Huawei has positioned itself as the primary domestic beneficiary. The precise dollar figure matters less than the trajectory: China’s AI chip market is decoupling from American suppliers faster than most forecasts predicted even 18 months ago, and Huawei is the company absorbing the largest share of that redirected spending.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
