Unlock Editor’s Digest Lock for Free
The founder of Huawei said the US exaggerated the capabilities of Chinese chipmakers and defeated his company’s technology, including export control debates, during trade talks between Beijing and Washington.
In a rare interview with China’s state-run people Daily on Tuesday, Len Zhengfei said Huawei’s Ascend chip, the main rival of China’s Nvidia products, is “still lagging behind the US for a generation.” He added, “The United States exaggerates Huaway’s capabilities — we’re not that strong yet.”
Ren’s comments have stated that Nvidia CEO Jensen Huang has issued a warning in recent weeks about Huawei’s progress on artificial intelligence chips, and that Washington’s restrictions on US chip makers are being sold to China, causing a “terrifying” competitor that threatens to undermine American domination of AI technology.
The US and China began new trade talks in London on Monday. This included discussions on Washington’s export control over key technologies.
In its first talk in Geneva, the US did not discuss export controls. However, the use of restrictions on several important rare earths and minerals used in Beijing’s recent automobile manufacturing (which could close factory lines, US, Europe and Japan), has brought this issue to the trade debate.
Huawei is benefiting from Washington’s ban on shipping Nvidia chips to China as Chinese technology giants accelerated their purchases of ascend chips and prepared to adopt Huawei’s technology.
Still, the majority of Chinese AI companies, including DeepSeek, are using NVIDIA chips to train large-scale language models that power AI tools. Domestic alternatives are increasingly being used for the less complicated task of inviting models to generate responses with tools such as chatbots.
Analysts and Huawei researchers have previously complained about technical flaws that cite the difficulty of using company chips to train LLMS, working with each other and distributing computing workloads throughout.
Ren suggested on Tuesday that the company has made progress in solving these issues, saying Huawei can “compensate” for performance degradation through cluster computing.
“Using clustering and stacking, computing results are comparable to the best in the world,” he said.
Huawei’s new CloudMatrix 384 AI server is at the heart of a strategy that competes with Nvidia by enhancing bandwidth capacity and assembles many chips to process more data. The server uses Huawei’s optical technology to connect a 384 AI processor.
Several Huawei customers have already tested CloudMatrix servers and are working with the company’s engineers to solve heat-related issues generated by the numerous chips that run at once, said those involved in the test.
Recommended
Huawei’s biggest challenge was creating an “ecosystem” of developers that use the platform to build AI models, the person added. One of the advantages of Nvidia is the software platform Cuda. Developers say it’s easy to use.
Ren said Huawei invests RMB180 billion ($25 billion) per year in research and development, and although RMB60bn is not intended to develop products, it is taking part in basic research aimed at making groundbreaking discoveries.
Ren said China has distinct advantages in building technical capabilities.
“AI relies on abundant power and sophisticated network infrastructure,” he said. “China’s power generation and grid systems are world class. Our communications infrastructure is the most advanced in the world.”
Additional Reports by Demetri Sevastopulo in Washington
