Huawei Technologies is projecting a massive surge in its artificial intelligence chip business. The company expects revenue from its Ascend processors to jump by at least 60% to reach $12 billion in 2026. The aggressive forecast, reported by the Financial Times and Reuters, underscores how the Chinese tech giant is capitalizing on US export controls that have effectively locked Nvidia out of the world’s second-largest economy.
The projected growth highlights Huawei’s rapid emergence as the dominant force in China’s domestic AI hardware market. As Chinese tech companies race to build and deploy advanced AI models, they are increasingly turning to Huawei’s Ascend chips as the primary alternative to Nvidia’s industry-leading GPUs, which are restricted from export to China under US regulations.
Capitalizing on US Export Controls
The surge in Huawei’s AI chip revenue is a direct consequence of the escalating US-China tech war. Washington’s efforts to choke off China’s access to advanced semiconductors have inadvertently created a massive, captive market for domestic alternatives. Huawei, with its deep pockets, extensive R&D capabilities, and strong government backing, has been the primary beneficiary of this dynamic.
According to recent statements by Nvidia CEO Jensen Huang, the US chipmaker’s market share in China has effectively dropped to zero. This vacuum has allowed Huawei to rapidly scale its Ascend business, securing bulk orders from major Chinese tech players, including Alibaba, ByteDance, and Tencent, who are desperate for computing power to train their large language models.
The $12 billion revenue projection for 2026 represents a significant acceleration from previous years. It suggests that Huawei has not only improved the performance of its Ascend chips but has also managed to overcome the manufacturing bottlenecks that previously constrained its ability to meet surging demand. Earlier this year, we reported that Huawei’s Ascend 950 production was scaling to 750,000 units in 2026, a figure that now appears consistent with the projected revenue trajectory.
The Ascend Ecosystem
Huawei’s success is not just about hardware; it’s also about building a viable software ecosystem. The company has invested heavily in its CANN (Compute Architecture for Neural Networks) software stack, aiming to create a domestic alternative to Nvidia’s ubiquitous CUDA platform.
While developers widely acknowledge that CANN still lags behind CUDA in terms of maturity and ease of use, the gap is narrowing. Major Chinese AI labs, such as DeepSeek, have actively collaborated with Huawei to optimize their models for the Ascend architecture, demonstrating that frontier-level AI performance can be achieved on domestic hardware.
The recent release of DeepSeek V4, which was optimized to run on Huawei chips, was a watershed moment for China’s AI industry. It proved that Chinese companies could build world-class models without relying on US hardware, further validating Huawei’s position as the cornerstone of China’s AI infrastructure. As we reported, Chinese chipmakers rushed to embrace DeepSeek V4, with Huawei leading the charge by achieving day-one compatibility and claiming 2.87x the inference performance of Nvidia’s H20 on its Ascend 950PR.
(Related: DeepSeek V4 to Run on Huawei Chips as Chinese Tech Giants Place Bulk Orders)
The Performance Gap and How It Is Narrowing
A key question for any assessment of Huawei’s AI chip business is the performance gap between its Ascend chips and Nvidia’s latest offerings. While Huawei claims that the Ascend 950PR delivers 2.87x the inference performance of Nvidia’s H20 on DeepSeek V4, it is important to note that the H20 is itself a downgraded chip designed specifically for the Chinese market to comply with US export rules.
Compared to Nvidia’s flagship H100 or H200 chips, which are not available in China, the performance gap is likely still significant. However, for the specific workloads Chinese AI companies need to run, particularly inference on models like DeepSeek V4, the Ascend chips appear increasingly competitive.
The key metric for AI companies is not raw compute performance but cost per inference, how much it costs to process a given amount of AI computation. Huawei’s claim that deploying DeepSeek V4 on Ascend chips reduces costs to approximately one-tenth of comparable GPT-based services is a striking assertion that, if accurate, would make the Ascend platform highly attractive for cost-sensitive applications.
Challenges and Limitations
Despite the impressive revenue projections, Huawei still faces significant challenges. The company remains cut off from the most advanced global semiconductor manufacturing technologies, such as ASML’s extreme ultraviolet (EUV) lithography machines. This means Huawei must rely on older, less efficient manufacturing processes, which limits the ultimate performance and energy efficiency of its chips compared to Nvidia’s latest offerings.
Furthermore, the US government continues to tighten export controls, targeting the loopholes and alternative supply chains that Chinese companies have used to acquire advanced chipmaking equipment. The recently proposed MATCH Act in the US Senate aims to further restrict the flow of semiconductor manufacturing tools to China, which could impact Huawei’s ability to scale production of its next-generation chips.
The Broader Implications
Huawei’s projected $12 billion AI chip revenue is more than just a corporate milestone; it is a barometer of China’s progress toward technological self-sufficiency. The figures suggest that US export controls, rather than crippling China’s AI ambitions, have accelerated the development of a parallel, domestic AI ecosystem.
As Huawei continues to improve its hardware and software offerings, it is not only securing its position in the domestic market but also potentially positioning itself as an alternative supplier for countries in the Global South that may be wary of relying solely on US technology. The battle for AI supremacy is increasingly fought not just in software labs but also in the silicon foundries and data centers that power the AI revolution.