Huawei Announces Three-Year Plans to Overtake Nvidia in AI Chips

Huawei has announced major plans for its AI chips, demonstrating the central role the technology plays in both the company and China’s push for tech self-sufficiency.

On Thursday, rotating chair Eric Xu introduced the next generation of Huawei’s Ascend AI chips. The roadmap begins with the Ascend 950PR, set to launch in the first quarter of 2026. Later that year, Huawei will release the Ascend 950DT. The line will continue with the Ascend 960 in late 2027 and the Ascend 970 in late 2028. All of these processors will use Huawei’s own high-bandwidth memory.

Huawei Announces Three-Year Plans to Overtake Nvidia in AI Chips

The company also showcased its updated SuperPoD cluster designs. These clusters link thousands of Ascend chips together, acting like one massive computer. Xu described them as “a single logical machine” that can learn, think, and reason collectively. He added that Huawei’s Atlas 950 SuperPoD would remain the most powerful in the world for years and would outperform rivals in every key area.

This clustering strategy is Huawei’s way of competing with Nvidia, which leads the global AI chip market. U.S. sanctions have blocked Huawei from using the latest manufacturing methods from companies like TSMC. Instead, Huawei is focusing on packaging techniques and large-scale chip clusters to narrow the performance gap.

Earlier this year, Huawei founder Ren Zhengfei admitted the company’s chips were still one generation behind U.S. processors. But he emphasized that smart packaging could help bridge that gap.

AI chip race in China

Huawei is not alone in this effort. Other Chinese companies like Cambricon, Baidu, and Alibaba are also designing their own AI processors. Smaller startups are focusing on domestic production, while big tech firms are creating chips in-house to support their AI platforms.

This optimism is lifting Chinese tech stocks. Shares in Semiconductor Manufacturing International Corporation (SMIC), which makes Huawei’s Ascend chips, have risen about 35% in the past month. Cambricon and Baidu stocks have jumped nearly 50% over the same period.

Much of this momentum started after Hangzhou-based startup DeepSeek released an open-source AI model earlier this year. Xu praised DeepSeek’s approach of training models with less computing power but stressed that advanced AI, including artificial general intelligence (AGI), will still require huge amounts of processing.

Geopolitical backdrop

The importance of domestic AI chips has grown since 2022, when the U.S. banned Nvidia and AMD from selling their most advanced processors to China. Both companies created weaker versions of their chips for the Chinese market, which were still attractive compared to local alternatives.

However, attitudes are changing. Beijing has reportedly told tech companies to stop using Nvidia’s RTX Pro 6000D, a chip designed specifically for China. Earlier, officials also raised concerns over Nvidia’s H20 processor. This has given local chips like Huawei’s Ascend line an even greater role in China’s AI future.

Huawei’s roadmap signals not just the company’s ambitions, but also China’s determination to secure its own technological foundation in the AI era.

See Also: What the TikTok Algorithm Trump Deal Means for App’s Future in U.S.