China's weapons in the AI race with the US — Huawei chip clusters, cheap energy

Chinese advancements in artificial intelligence are making waves despite the global dominance of U.S. firms like Nvidia. While it's acknowledged that Chinese semiconductors currently do not match Nvidia's prowess, the country is nevertheless pushing forward with the development of sophisticated AI models powered by locally produced chips. The driving force behind this progress is a combination of cheap energy and robust chip clusters, particularly from Huawei, a leading technology player in China. According to Wendy Chang, a senior analyst at the Mercator Institute for China Studies, China views AI as pivotal for both national and economic security, prompting a strong push for self-sufficiency in this vital sector. Despite facing significant challenges from U.S. export restrictions, which have limited access to top-tier technology, Chinese companies such as Alibaba and DeepSeek have successfully developed competitive AI models. Nvidia's GPUs are regarded as the benchmark for AI applications, but U.S. regulations have hindered Nvidia from exporting its most powerful chips to China. Recently, the company received approval to sell a downgraded version of its technology in the Chinese market, yet Beijing is encouraging domestic firms to rely on homegrown alternatives. Huawei's Ascend series of chips are at the forefront of this initiative, offering a unique advantage through the ability to interconnect multiple chips into high-performance clusters. One exemplary product is the Huawei CloudMatrix 384, which utilizes 384 Ascend 910C chips to achieve performance levels comparable to Nvidia's advanced systems, despite utilizing a greater number of less powerful chips. This approach, while effective, comes with a trade-off in terms of energy consumption. However, China's access to inexpensive energy resources, bolstered by substantial investments in green energy and nuclear power, allows it to mitigate some of these inefficiencies. Cities like Shanghai and Shenzhen are actively supporting AI infrastructure development through subsidies aimed at lowering operational costs for companies that utilize local chip technology. Yet, the sustainability of this strategy raises questions, especially considering the semiconductor supply chain and production capabilities. Huawei's chips are produced by Semiconductor Manufacturing International Co. (SMIC), which faces technological challenges due to restrictions on acquiring advanced manufacturing tools. As the competition in AI intensifies, experts warn that China must enhance its semiconductor production capacity to keep pace with leading firms like Nvidia and TSMC. As Hanna Dohmen from Georgetown's Center for Security and Emerging Technologies notes, the journey to bolster domestic chip manufacturing is ongoing, but significant hurdles remain due to the current technological limitations imposed by international export controls.