Jeffrey Kessler, the Deputy Undersecretary of Commerce for Industry and Security at the U.S. Department of Commerce, recently made a significant statement before Congress. He indicated that, according to U.S. assessments, Huawei’s production capacity for Ascend chips will be 200,000 units or less by 2025, with the majority expected to serve firms within China. This announcement was a stark reminder of the ongoing technological divide and the fierce competition in the artificial intelligence sector.
Furthermore, Kessler highlighted that China is investing heavily in enhancing its AI chip production capabilities, particularly in Graphics Processing Units (GPUs). He emphasized the importance of recognizing this escalating challenge and advised against complacency. His statements directly reflect the U.S. administration’s concerns about China’s rapid advancements in AI technology.
China’s Proximity in AI Technology
According to David Sacks, an AI and cryptocurrency expert and a former advisor to the Trump administration, China is only three to six months behind the U.S. in AI technology. Despite initial alarms from his statement, the White House clarified that he was specifically referring to AI models. Thus, Chinese chips are noted to still be about one to two years behind their American counterparts. This nuanced understanding highlights the critical evaluation of the competition between the two nations.
This week, Ren Zhengfei, the founder and CEO of Huawei, echoed these sentiments, stating that Huawei’s Ascend GPUs are still a generation behind their American equivalents. While this statement aligns with U.S. government assessments, it’s imperative to note that Huawei invests over $25 billion annually in AI hardware development, suggesting that they could soon catch up with industry leaders like NVIDIA and AMD.
The wafer performance of the integration technologies employed by SMIC for Huawei’s GPU production has ample room for improvement.
Nevertheless, Kessler’s prediction regarding Huawei’s AI chip production capacity rests upon a pivotal fact: the integration technologies utilized by SMIC (Semiconductor Manufacturing International Corp) for producing Huawei’s GPUs have significant room for improvement. Currently, SMIC can manufacture 6 nm integrated circuits and is transitioning to 5 nm semiconductors , yet they face limitations due to the deep ultraviolet lithography (DUV) equipment they possess.
It is commendable that the engineers at SMIC and Huawei have refined their processes to successfully produce chips at the 5, 6, and 7 nm levels using ASML’s DUV equipment. However, it is unlikely that they will surpass the 3 nm threshold with these machines. The technique they are employing, known as multiple patterning , comes with inherent limitations. This strategy generally entails transferring the pattern onto the wafer multiple times to enhance the lithographic resolution. Unfortunately, this approach often leads to increased costs for the chips and reduced production capacity.
This poses a significant challenge for Huawei, as not having access to the cutting-edge technology required to manufacture semiconductors on par with leading firms like Intel, TSMC , and Samsung greatly affects their competitive stance. To overcome these hurdles, Huawei is actively working on developing its own extreme ultraviolet (EUV) lithography equipment.
As global observers continue to monitor the semiconductor landscape , the difference in capabilities between the U.S. and China is increasingly significant, particularly in terms of AI advancements. Kessler’s remarks serve as a clear indication of the seriousness with which the U.S. administration regards China’s technological ascent. While challenges remain for Huawei and SMIC, their investment strategy and commitment to innovation could soon recalibrate the balance of power in the tech sector.