A Chip on the Shoulder
China’s Technological Achilles’ Heel
Part 1 - to finding ways around the hurdles
By Darryl Lau
Date: 13th June 2024
Jensen (Renxun) Huang, Nvidia CEO, at SC18.
Courtesy of wikimedia (2018).
The recent visit by Nvidia CEO Huang Renxun to Taiwan and his subsequent remarks calling it “one of the most important countries in the world” sent ripples through the already turbulent relationship between China and Taiwan. While China views Taiwan as a breakaway province, Taiwan asserts its self-governance. This underlying tension is further amplified by the United States’ “strategic ambiguity” on the issue.
However, beyond the political firestorm, Huang’s comments highlight a lesser-discussed vulnerability for China: its chip industry. Unlike its increasingly assertive diplomatic approach on other fronts, China’s relative weakness in developing its own cutting-edge chips limits its ability to forcefully counter companies like Nvidia. This technological dependence exposes a critical Achilles’ heel, hindering China’s ambitions to become a global technological superpower.
The self-sufficiency of China Chip
China’s domestic chip self-sufficiency rate has been a topic of much debate in recent years. Before 2018, the self-sufficiency rate hovered around 5% without much contention. However, as the issue gained traction since 2020, so too did the discrepancies in reported figures.
Looking at specific numbers, IC Insights (acquired by TechInsight) estimated the 2020 chip market size in mainland China at $146 billion, with domestic production at $24.2 billion. This translates to a self-sufficiency rate of roughly 16.6%. The rate continued to rise slightly in 2021 and 2022, reaching 17.6% and 18.3% respectively.
For 2023, predictions diverge. International Business Strategies (IBS) puts the self-sufficiency rate at a more optimistic 25.61%, while TechInsights offers a more conservative estimate of 23.3%. TechInsights even suggests a national self-sufficiency rate of only 26.6% by 2027.
The difference in these figures boils down to how “domestic” is defined. TechInsights highlights the significant presence of foreign-invested chip manufacturers in China. In 2022, for instance, the total output value of China’s chip manufacturing industry was $30 billion, with domestic and foreign enterprises contributing roughly equally at $15.2 billion and $14.8 billion respectively. Based on this broader definition, the self-sufficiency rate sits at 18.3%. However, if we consider only chips manufactured by domestic companies, the rate plummets to just 9%. This trend holds true for 2023 as well – a broad definition yields a 23.3% self-sufficiency rate, while a narrower definition focused solely on domestic production lands at just 12%.
Taking these varying calculations into account, the internationally accepted self-sufficiency rate for China in 2023 falls between 23.3% and 25.61%. With this as a baseline, achieving the ambitious goal of 30%-35% self-sufficiency by 2024 isn’t entirely unrealistic, but it certainly won’t be easy.
Chinese Chip Production Trends. Courtesy of DIGITIMES Asia (2024).
China’s chip industry since 2020
Since 2020, China’s chip industry has witnessed a period of both significant strides and undeniable challenges. Fueled by government initiatives like the “Made in China 2025” plan, China has invested heavily in domestic chip production. This ambition, however, has been met with external pressure. The US imposed strict export restrictions on advanced chips to China in 2022, directly impacting major suppliers like Nvidia.
Nvidia’s attempt to navigate these restrictions by creating China-specific AI chips, like the A800 and H800, proved unsuccessful. These chips were not only technically inferior to their original offerings but also failed to compete with lower-priced alternatives from domestic players like Huawei’s Ascend series. This episode exemplifies the current state of China’s chip industry: rapid domestic development, spearheaded by companies like Semiconductor Manufacturing International Corporation (SMIC), struggling to match the performance and efficiency of established foreign counterparts.
However, China is not backing down. The government is actively pushing tech companies to prioritise domestically produced AI chips, through policies like preferential tax breaks and subsidies. This initiative aims to reduce reliance on foreign suppliers like Nvidia and AMD. While these domestic chips might not yet match the best in the world, they represent a crucial step towards technological self-sufficiency.
The recent move by US chipmakers to develop AI-specific chips, such as Intel’s Movidius and Google’s TPU, has further fuelled the narrative that circumventing Nvidia’s GPU chip technology is possible. This realisation has emboldened Chinese AI chip companies, many of whom lack prior GPU chip expertise, to pursue their independent designs.
This shift towards domestic AI chips is expected to have a significant impact on the Chinese AI chip market. Concerns over the performance of domestic chips compared to Nvidia’s offerings may diminish, leading to increased adoption of homegrown solutions. This, in turn, could erode Nvidia’s dominance in the market and provide an opportunity for Chinese AI chip companies to flourish. Furthermore, the widespread adoption of domestic AI chips could accelerate China’s progress in AI chip technology and potentially break Nvidia’s stranglehold on the CUDA ecosystem.
Semiconductor chips on display.
Courtesy of the New Centre (2022).
Finding Ways Around the Hurdles
The US has imposed restrictions on China’s access to advanced chips, chip manufacturing equipment, and high-end processors in an effort to hinder China’s progress in AI and supercomputing. These restrictions have forced Chinese AI chipmakers to seek alternative solutions, such as downgrading chip designs to meet US requirements.
Despite having 44 chipmakers, China’s production capabilities are limited. Only Semiconductor Manufacturing International Corporation (SMIC) can mass-produce advanced GPUs, but its capacity is currently reserved for Huawei. However, SMIC has reportedly agreed to allocate limited capacity to Chinese AI chip companies directly sanctioned by the US and banned from overseas production, including Cambricon, a government-backed company.
In an attempt to circumvent US export restrictions on AI chips and technology, two Chinese AI companies, MetaX and Enflame, submitted downgraded chip designs in late 2023. MetaX, founded by former AMD executives, developed a downgraded product called C280 after its most advanced GPU, C500, sold out in China. Both MetaX and Enflame, two tech giants, sell their chips to state-owned enterprises and collaborate with local governments.
The actions of MetaX and Enflame demonstrate the lengths to which Chinese AI chipmakers are going to navigate US restrictions and maintain access to TSMC’s production facilities. Their success in securing production will depend on their ability to balance compliance with US regulations while maintaining the performance and competitiveness of their chips.
Part two of the article will explore the PRC’s approach towards reaching chip self-sufficiency.
Chinese model promotes smartphone outside electronics center notorious for selling fake, gray market, and pirated electronics.
Courtesy of Stephen Shaver (2015)
Copyright © 2024 Sparklight Media