A group of Chinese graphics processing unit (GPU) developers are taking advantage of Nvidia’s absence from China’s biggest artificial intelligence (AI) show to promote their wares, even as they face problems related to manufacturing and bottlenecks in the software ecosystem.
Iluvatar Corex, Moore Threads, Tencent Holdings-backed Enflame Technology, Sophgo and Huawei Technologies’ Ascend were among the headliners at the World Artificial Intelligence Congress in Shanghai last week. The California-based company that designed the chip against which all other AI GPUs are based was not physically present at the conference. But Nvidia’s name was frequently mentioned in the Chinese chipmaker’s booths, comparing its wares to those of the U.S. tech giant.
“China’s computing clusters are changing from being dominated by foreign GPUs to a mix of Chinese and foreign GPUs,” Li Xingyu, Enflame’s chief ecosystem officer, said in a session on Friday. “There is a demand shortage problem, and a lot of China’s computing capacity is still untapped.”
Have questions about the biggest topics and trends from around the world? Find the answers on SCMP Knowledge, our new platform that provides curated content with explainers, FAQs, analyses and infographics from our award-winning team.
“Further lowering the barriers to accessing domestic computing power will be key to expanding its applications in China,” he added.
Mooresled’s MCCX D800 AI server was on display at the World Artificial Intelligence Congress in Shanghai on July 5, 2024. Photo: Che Pan alt=Mooresled’s MCCX D800 AI server was on display at the World Artificial Intelligence Congress in Shanghai on July 5, 2024. Photo: Che Pan>
AI solutions company Infinigence, another startup backed by Chinese social media and video game giant Tencent, is working with Shanghai-based chip design firm Enframe to provide computing resources that use a variety of chips from Nvidia and other vendors, including Chinese GPU makers. “Enterprises don’t need to worry about which GPU to use,” the company said.
Nvidia is unable to export cutting-edge semiconductors to China due to U.S. export restrictions. Washington has also added major Chinese chip developers to a trade blacklist, making it harder for the company to find foundries to manufacture its designs. Huawei, which claims its Ascend 910B is equivalent to Nvidia’s A100, is among the companies facing production obstacles after it was blacklisted in 2019 and sanctions have been tightened since then.
The area around Huawei’s booth was one of the busiest areas at the show. According to a slide presentation at the show, Huawei’s Ascend chips are configured for open source large-scale language models, including Meta’s Llama2. The chips are also used to train half of China’s 70+ large-scale models, Huawei claimed in a presentation at the Zhongguancun Forum in May.
Enflame showed off its Cloudblazer T20 and T21 AI training chips at the show. Company employees said its biggest advantage over larger rivals like Huawei, MooreThreads and Viren Technology is that it’s not on the U.S. trade blacklist. That means it has access to global foundries like TSMC, which makes about 90% of the world’s most advanced chips, even as broader sanctions limit the computing power of its chips.
Nvidia chips customized for Chinese customers to circumvent U.S. restrictions also remain popular.
Nvidia is expected to ship more than 1 million H20 GPUs in China this year, generating $12 billion in revenue, according to San Francisco-based semiconductor research firm Semianalysis Inc. Dhilan Patel, chief analyst at Semianalysis Inc., said Nvidia’s H20 still offers better performance than Huawei’s 910B.
“China is a big market with strong demand for AI, but the biggest challenge is that domestic technology lags behind global peers and is not sufficient to meet domestic demand,” said Robert Chen, head of Taiwan research and Asia-Pacific hardware at Bank of America Global Research.
“China’s AI GPUs are not as powerful as the rest of the world.” [alternatives] Performance wise, and… [the Chinese supply chain] Overall we are still far behind,” he added.
Mainland China remained Nvidia’s third-largest market for the fiscal year that ended Jan. 28, with sales in China increasing 78 percent year-over-year to $10.3 billion.
Additional reporting by Iris Deng.
This article originally appeared in the South China Morning Post (SCMP), the most authoritative news source on China and Asia for more than a century. For more SCMP articles, visit the SCMP app or follow SCMP on Facebook. twitter P a g e Copyright © 2024 South China Morning Post Publishers Ltd. All rights reserved.
Copyright (c) 2024. South China Morning Post Publishers Ltd. All rights reserved.
