Key Highlights
- Domestic demand for Huawei’s AI processors far exceeds current production capacity in China
- Ascend 950DT chip undergoes testing with Chinese AI model training expected to begin next year
- Next-generation Ascend 960DT processor timeline accelerated by nine months to early 2027
- Company introduces Peerium computing framework designed to interconnect up to 1 million chips
- Huawei reports capturing greater Chinese market share than Nvidia in AI chip sales
Huawei announced it is struggling to keep pace with surging domestic demand for its artificial intelligence processors, effectively ruling out significant international expansion. The revelation came from rotating chairman Eric Xu during Thursday’s Huawei Connect conference held in Shanghai.
According to Xu, while the company does serve select international clients in markets showing strong interest, shipment quantities remain minimal. Priority is squarely on satisfying China’s internal requirements.
Next-Generation Ascend 960DT Timeline Advanced
The Chinese technology giant has accelerated its Ascend 960DT chip launch to Q1 2027, representing a nine-month advancement from initial projections. The subsequent Ascend 960PR model will debut in Q3 2027, also arriving three months earlier than originally planned.
Huawei’s strategy includes annual releases of new Ascend processor generations, with the Ascend 970 and 980 series slated for 2028 and 2029 respectively.
The conference also featured the debut of the Atlas 960 SuperPoD computing cluster. This enhanced system surpasses the Atlas 950 SuperPod introduced mere months earlier, delivering superior performance in both AI training operations and inference tasks.
Xu reported positive testing outcomes for the Ascend 950DT processor. He anticipates Chinese artificial intelligence developers will commence model training on systems powered by this chip throughout the coming year. Huawei’s computing ecosystem currently supports over 40 trained AI models, while the Ascend chip software attracts more than 5,200 active developers monthly.
Strategic Push Toward Complete Semiconductor Independence
Xu spoke candidly regarding Huawei’s aggressive chip development strategy. American export regulations have curtailed China’s access to Nvidia’s cutting-edge processors since 2019, while Huawei itself has operated under US trade limitations since May of that same year.
“We cannot accept a destiny where we cannot control our fate being determined by others in terms of willingness to sell chips to China or not,” Xu stated.
To circumvent restrictions on individual processor capabilities, Huawei is pursuing strategies that aggregate vast numbers of chips. Contemporary large-scale AI model training typically requires clusters approaching 100,000 processors.
The company revealed Peerium, a novel computing architecture engineered to ultimately network up to 1 million processors functioning as a unified system. Complementing this, UnifiedBus technology interconnects processors, memory modules, storage components, and networking infrastructure within these massive installations.
According to Huawei, the Ascend 960 supernode configuration supports integration of up to 4,096 processors. These supernodes can subsequently be combined to form even more extensive computational clusters.
Over 1,000 systems utilizing the earlier Ascend 910C processors are already operational. Commercial deployment of Ascend 950-based systems has commenced.
Nvidia maintains a substantial competitive edge through its CUDA development platform, which remains the foundation for developers building and executing AI applications. Nvidia did not provide a response when contacted for comment.
These announcements from Huawei arrived less than fourteen days before the scheduled September 24 meeting between US President Donald Trump and Chinese President Xi Jinping in Washington.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
