At the World Artificial Intelligence Conference in Shanghai last week, a consortium of Chinese GPU developers seized the spotlight in the absence of Nvidia, the California-based tech giant synonymous with AI chip supremacy. Companies like Iluvatar Corex, Moore Threads, Enflame Technology backed by Tencent Holdings, Sophgo, and Huawei Technologies’ Ascend drew significant attention, positioning themselves as formidable contenders in the global AI chip market despite facing manufacturing and software ecosystem challenges.
Enflame Technology’s Chief Ecosystem Officer, Li Xingyu, emphasized the evolving landscape of China’s computing clusters, shifting from reliance on foreign GPUs to a blend of domestic and international technologies. “There remains a significant underutilization of computational power in China, stemming from a lack of demand,” Xingyu noted during a session on Friday, advocating for broader accessibility to homegrown computational resources.
Collaborating with Shanghai-based Enflame, Infinigence, a startup supported by Tencent, showcased its versatile compute resources utilizing GPUs from Nvidia and other global vendors alongside Chinese alternatives. “Our solutions eliminate concerns over GPU compatibility,” the company affirmed.
Nvidia, restricted from exporting its most advanced chips to China under US regulations, has prompted major Chinese chip developers like Huawei to navigate significant manufacturing hurdles. Huawei’s Ascend 910B, touted to rival Nvidia’s A100, faces constraints exacerbated by trade blacklistings since 2019, limiting access to crucial foundries.
Despite these challenges, Huawei’s booth remained a focal point, showcasing chips integral to open-source large language models and training a substantial portion of China’s extensive AI models.
Enflame also showcased its Cloudblazer T20 and T21 AI-training chips, underscoring its advantage of operating outside US trade restrictions, enabling access to global foundries pivotal for advanced chip manufacturing.
Meanwhile, Nvidia continues to cater to Chinese clients with tailored solutions that circumvent export barriers, maintaining strong market demand. According to SemiAnalysis, Nvidia anticipates delivering over 1 million H20 GPUs in China this year, driving an estimated $12 billion in sales.
However, challenges persist. Robert Cheng, head of Taiwan research at BofA Global Research, highlighted that while China boasts a robust AI market, its technological capabilities lag behind global peers, constraining domestic supply chain efficacy.
Despite these disparities, mainland China remains a critical market for Nvidia, with sales up 78% year-on-year to $10.3 billion in the last fiscal year, underscoring the region’s strategic importance in the global AI landscape.
Related topics:
Intelligent Building Management System Software
Intelligent Data Extraction: Revolutionizing Data Processing in the AI Era
Condition Monitoring and Control for Intelligent Manufacturing