China vs. USA – Chip Debate

AI chip imports into China more unlikely than ever.

• The controversy surrounding Jensen Huang has reignited the debate over whether the West should allow China access to its best technology, but I suspect that the pro-import side of the argument has taken a fatal blow regardless of the merit of its case.
• Jensen Huang has been one of the most outspoken advocates for allowing imports into China, and given how strong demand for AI in China is, it is not hard to understand why.
• More than 80% of China’s population has a positive view of how AI will impact their lives (way above Western countries), and China has plenty of electricity to power any data centres that it builds.
• As Nvidia arguably has the best systems for training and running AI algorithms in the world, it is easy to see that its addressable market is being hampered by not being able to sell into the Chinese market.
• However, it also clear that its revenue growth is not, as it has been able to switch all of the capacity it had earmarked for China to other customers, and Nvidia remains sold out 12 months in advance.
• At some point, the lack of revenue in China will start to impact Nvidia’s financial performance, but it is hard to see that much before 2029.
• The argument for allowing China access to advanced silicon has some merit in that China would become dependent on foreign technology, giving the West a lever to use should the geopolitical need arise.
• However, this is precisely what China has stated that it does not want, and the Chinese regulator remains the main reason why the H200 chip is not selling in China, not the Department of Commerce.
• Nvidia has been lobbying the US Government to allow it to sell advanced chips on the premise of the above argument but also to maximise its long-term financial performance.
• This is Jensen and the board’s fiduciary duty to Nvidia shareholders, but what has come to light recently has most likely holed this side of the argument below the waterline regardless of its merit.
• The allegation from the pro-restrictions side will always be that China has influenced Jensen into taking this side of the debate and pushing it in Washington, thereby rendering it worthless.
• I have always taken the view that China is seeking complete independence from Western technology, which is why it has banned the import of all Nvidia data centre chips, but recent events imply that reality may be a bit more nuanced.
• Anyone who has been in a trade negotiation with China will tell you that the first thing China asks for is advanced Nvidia chips rather than the watered-down versions.
• Take this in conjunction with what could be interpreted as an elaborate scheme to have the restrictions removed, and one would conclude that China wants the advanced chips but not the second-rate ones the US will allow it to have.
• This makes sense, as it would allow China to be much more competitive in AI than it already is and buy it time until it can bridge the gap in silicon chip performance and remove Nvidia and everyone else from its systems.
• If it turns out that Ms Huang was an agent of the Chinese state, then it can be easily argued that her role was to influence Jensen to make the argument in Washington to relax the export restrictions, which is precisely what he has done.
• Unfortunately, it no longer matters why Jensen lobbied to end the restrictions, as his credibility has been shattered when it comes to this issue and his side of the debate has taken a mortal wound.
• Hence, I think that the outcome of this affair will be that restrictions on imports to China will be tightened further and that its disadvantage in AI silicon will persist.
• This will mean that China becomes even more dependent on importing compute from data centres built in neighbouring countries with Western chips.
• While this solution works, it is vulnerable to US policy in that the US could mandate that US chips are not allowed to run workloads destined to be used in China.
• This would cause considerable disruption to the Chinese AI industry and greatly slow its development and use of AI on the Chinese mainland.
• So far, the USA has chosen not to take this route, but the possibility remains for as long as the dependency continues to exist.
• This is why I believe that the US and the West still have a large advantage over China, which is not going to change any time soon.