Huawei does not plan to fully deploy its new artificial intelligence accelerators outside China. The main reason, as the company’s rotating chairman, Eric Xu, explained in September 2026, is that current capacity is not enough to meet even domestic demand. Ascend chips are processors designed to accelerate AI tasks, such as training models and generating responses.
The decision is not just commercial. It reflects a combination of local demand, difficulties in manufacturing advanced semiconductors, and tensions between China and the United States. It also shows that competing in AI does not depend solely on designing a fast chip: it must be produced in quantity, connected to other processors, and supported by compatible software.
Domestic demand exceeds supply
Huawei wants to reserve its newest chips for Chinese customers because it is already struggling to meet their needs. This allows the company to concentrate scarce production on businesses and data centers in its home market, where demand for AI computing capacity has grown and access to the most advanced foreign processors is subject to United States export restrictions.
The company has presented a strategy that combines Ascend chips with broader computing systems, such as its SuperPods: clusters of many chips working together. This approach matters because a large AI system needs to coordinate thousands of processors; system performance also depends on the connections between them, memory, and software, not just the capabilities of an individual chip.
Manufacturing advanced chips remains a bottleneck
Modern AI chips require sophisticated manufacturing processes, high-speed memory, and specialized equipment. U.S. restrictions limit Huawei’s and the Chinese industry’s access to several key technologies and components. This makes it harder to increase production and can affect both the available volume and the ability to maintain a regular pace of new products.
In May 2025, the United States Department of Commerce warned that using certain Ascend chips could involve violations of U.S. export control regulations. The warning adds uncertainty for potential international customers and suppliers: a cross-border sale or use may carry legal risks if technologies subject to these controls are involved.
Estimates of production also vary. In 2025, a senior U.S. Department of Commerce official told Congress that Washington estimated capacity of up to 200,000 Ascend chips that year, and that most or all would be delivered to Chinese companies. This is an estimate by the U.S. government, not a production figure publicly confirmed by Huawei; other analyses have suggested different volumes.
Competing also means building an ecosystem
For Huawei, prioritizing China may help establish a local alternative to foreign platforms. Companies that adapt their models and services to Ascend can help grow compatible software and tools. But switching platforms is not automatic: developers may need to modify code, adjust processes, and check that performance and costs fit their needs.
This is one of the challenges of competing with Nvidia, which has an established position in AI processors and the software used to develop with them. Huawei can offer an alternative that is particularly relevant to Chinese customers, but performance comparisons depend on the task, configuration, and complete system. Specifications announced by a company are not, by themselves, equivalent to an independent test under comparable conditions.
What this means for the global market
The geographic limitation currently reduces the international presence of Huawei’s new products, but it does not necessarily mean the company will never sell them abroad. Its public explanation is that it wants to meet Chinese demand first; production capacity and technological restrictions are factors that may constrain any future expansion.
For Chinese customers, greater local supply could give them an alternative amid export controls. For buyers in other countries, however, limited availability and regulatory uncertainty may make adoption more difficult. And for the sector as a whole, the case illustrates a key reality: in the AI race, designing competitive chips is not enough; they must also be manufactured at scale, their components secured, and an ecosystem built to make them useful.
Sources and references
- Associated Press: Huawei unveils new AI chip technologies in its race with Nvidia
- Bureau of Industry and Security: guidance on advanced computing chips and Huawei Ascend
- Reuters: Huawei says AI chip demand exceeds supply
- Georgetown CSET: analysis of Ascend chips and U.S. export controls
Image: Courtesy of the Naval Surface Warfare Center, Dahlgren, VA., 1988. | Public domain | Wikimedia Commons.
Educational notice
This content is for educational and informational purposes. It does not constitute financial advice or an investment recommendation.