The success of China’s cheap AI technology has clearly worried American AI companies. Industry leaders have openly called on the US government to take action on competition from the Chinese AI industry.
Washington has already imposed various restrictions on the sale of chips to China. Last year, it launched the Pax Silica initiative, aimed at securing supply chains for AI technology that exclude Beijing; 23 states have since joined, along with the European Union. There is now pressure from the US AI industry to ban or restrict Chinese AI models.
It increasingly appears that the two great powers are on the verge of an AI cold war, each building its own camp and bidding to set the rules of the AI era. The question now is: Having declared this war, can the US win it?
The limits of export controls
At present, the US and its allies hold the upper hand in several key areas of AI development. China’s frontier AI models remain heavily dependent on imports, chief among them chips, chip-designing software, and chipmaking equipment.
As a result, US export controls hinder China’s ability to keep pace with the latest AI technologies; they drive up costs and undermine investor confidence. They will, however, not hold China’s AI development back forever.
It is important to remember AI development is by no means a single contest over chips; human capital, financial capital, market size, regulatory agility, and industry adoption matter just as much. In response to US export controls, China is not only investing in domestic alternatives but also improving AI model architecture, optimising AI training and inference efficiency, and squeezing more out of the chips it already has.
While trying to catch up with development in chips, China is seeking various loopholes in US export controls to circumvent the restrictions. In June, the US government curbed the ability of Chinese companies to buy advanced US chipss will emerge
Whether the US can contain China’s advancement in AI also hinges on the willingness of its allies to cooperate. Advanced chips are not made in a single country. Instead, their supply chain spans multiple political entities, bringing together US design, Dutch and Japanese equipment, and Taiwanese manufacturing.
This dependence means that US export controls, if pushed too far, revised too often, or kept in place for too long, could backfire. US allies, unhappy that Washington keeps offloading steep commercial costs onto them for the sake of its own national security, may well develop policy fatigue and begin actively investing in alternative supply chains to bypass US technology and regulations altogether.
Christophe Fouquet, CEO of the Dutch chipmaking equipment giant ASML, has questioned restricting sales of the most advanced chipmaking equipment to China – under US pressure – would even work or would only push it to redouble its efforts to develop its own.
Doubts about the current US strategy have been voiced even within the US AI industry. US chip giant Nvidia, for example, is against the chip export restrictions, as China is a major market for the company. CEO Jensen Huang has called export controls a “failure” that has cost US companies billions.
Moreover, Chinese open- This year, nearly 200 US startups sent letters to the Trump administration arguing that banning access to advanced Chinese AI models – which tend to be open-ave the US market in the hands of a few AI giants. As with all monopolies, the result would be stifled innovation, a diminished US lead in AI, and higher prices for US consumers
Beijing’s rare earth card
In the AI cold war, China is not just playing catch-up. It is also wielding significant leverage over the US.
Rare earths, which are key to chipmaking, are probably Beijing’s strongest card. The country not only holds the world’s largest reserves of rare earths but dominates their mining and processing. It alone accounts for 91 percent of global rare earths separation and refining production.
Washington is well aware of this. Last year, in response to the renewed trade war, Beijing flexed its rare earths muscle by banning the export of heavy rare earths to the US, which dealt a heavy blow to US manufacturing. It forced Washington to call a quick truce – one that saw Beijing suspend the ban for a year.
Since then, the US has redoubled its efforts to wean itself off Chinese rare earths – to no avail. Not only are Chinese clay deposits – from which rare earth-bearing ores are extracted – more economical than those found in the US or its allies, but China has also mastered the ability to process those ores into rare earths, a capability long spurned by the West because of the pollution it generates.
Although the technology is hardly rocket science, the US and its allies lack the human capital and technology to catch up quickly. China has tens of thousands of experienced engineers working on rare earths processing; in the West, there are just a handful of such experts.
In light of this, should the US escalate its measures against China, it would itself be hit with embargoes and tariffs – measures that would also hinder its own AI development.
Worse still for Washington, rare earths are also used in weapons ranging from F-35 fighter jets to anti-ballistic missile defence systems. Given that fighting between the US and Iran could flare up again at any moment – and that the US army has nearly drained its interceptor stockpiles in the previous round – Washington cannot afford to enter the next round unprepared.
As much as the US would like to restrict China’s AI advance, doing so could be too costly for its own industry. Despite efforts to decouple, “Chimerica” – a term coined to describe the economic symbiosis between the two – is here to stay. The sooner Washington realises that the US needs China as much as China needs the US, the better – not just for both countries, but for the whole world, which could otherwise be caught in the crossfire when these two superpowers duel.
The views expressed in this article are the author’s own and do not necessarily reflect Al Jazeera’s editorial stance.
Source: www.aljazeera.com
