The Chip Illusion and Competition with China – Part“The more the United States abuses its position as the dominant player… the more those on the receiving end will try to find workarounds or alternatives.”
Edward Fishman, America’s Economic Warfare Is Sowing Its Own Demise
Countries around the world have made AI a priority. This is because the use of data and data tools increases productivity and revenue and has important security applications. Underpinning this is a complex, interconnected, internet-based network of fiber optic cables (including undersea cables) satellites, mobile networks, and cloud computing services, to carry, store and process data for users. Data centers – large collections of computers and chips – are fundamental to the new economy, serving as the physical infrastructure that enables the storage, processing, and use of data by software applications and tools, the most prominent of which is AI. This is a different kind of industrial base and a new kind of industrialization, marked by rapid growth in scale and application.
Data infrastructure is at the forefront of a growing rivalry between Beijing and Washington. Competition over data centers is expected to be more like the earlier competition over 5G telecommunications, given the importance of data for economies and the central role of data centers, and extends into the data center supply chain, including artificial intelligence, but they are all part of the larger race to build the infrastructure of the emerging digital economy. The United States is the global leader in AI research, development, and investment while China has rapidly growing AI capabilities, with significant government support and investment.
China and the US are in a competition to see who will build the global infrastructure for a future digital economy driven by connectivity and software. The first competition was over telecom infrastructure with Huawei. Huawei successfully survived the US efforts to block it from global telecom markets and in fact, continues to do quite well. The current battle between the US and China is over artificial intelligence and data centers, since data centers will be the core infrastructure for the digital economy, and who builds them, who operates them and what technology they use will be a determining factor in the contest between US and China.
AI is best seen as a set of advanced software tools that will improve business operations and research. Companies that adopt AI tools for use in their businesses will have an competitive advantage. AI’s military applications will increased precision, speed and lethality. Having learned from earlier digital technology revolutions, investors are pouring money into the new tools and there is an unusual quantity of hype about AI. The US, eager to deny China these military advantages (and perhaps motivated by a degree of economic warfare), has sought to restrict China’s access to AI related hardware, principally chips. In this, it looks to exploit its undeniable technological lead over China and China’s’ reliance on advanced western technology, but this reliance has been steadily shrinking, and with it, the benefits of technology denial.
Unfortunately, new US regulations on AI “diffusion,” essentially gives China an advantage that it would not otherwise have. The new competition is over winning markets, and Chinese companies will take advantage of market opportunities that are foreclosed now to US companies. Their technology is not yet as good, but their prices are lower and they do not come with the onerous restrictions that the US imposes on its companies. More importantly, the regulation will slow US companies and make it hard for them to compete in global markets.
The previous Administration hoped that export controls would choke China in a tech competition and this was a mistake. This became apparent with the ill-advised effort to replace Huawei and the effort to restrict access to advanced chips, to derail China’s efforts to develop artificial intelligence (AI). The evidence for this failure is stark. In the case of Huawei, American restrictions were initially damaging to the company, but two years later, this was followed by Huawei having its best year on record in 2023 in terms of both revenue and market share. Claims from the White House that America’s chip export controls had forced Huawei to relay on older chips for its new model phones may have been true but went unnoticed by consumers or the market.
Similarly, the administration identified advanced chips as a chokepoint for AI, part of a larger misunderstanding of the role of chips (and technology generally) in creating strategic benefit in military and economic competition. Chips are not a chokepoint in this contest. The Chinese, denied access to advanced chips, created substitutes and work arounds most recently with DeepSeek. Why this was a surprise is puzzling as since the 1980s, every time that the US has attempted to use export controls to deny access to a technology the result has been to create foreign competitors and weaken the US technological lead.
In both cases, China was able to develops alternatives and work-arounds. These may have cost more to produce and may not be as good as American products, but the words of the former CEO of a major chip maker, they are “good enough.” When it comes to competing in strategic technologies, China is not subject to the same constraints as western companies. The price of a Chinese chip is higher, the performance somewhat less, and there is a greater error rate in production, but China can afford this and is willing to pay.
AI is not magic. It is best seen as a set of advanced software tools that will improve business operations and research. Having learned from earlier digital technology revolutions, investors are pouring money into the new tools. Companies that adopt AI tools for use in their businesses will have an competitive advantage. AI’s military applications will increased precision, speed and lethality. The US, eager to deny China these military advantages (and perhaps motivated by a degree of economic warfare), has sought to restrict China’s access to AI related hardware, principally chips. In this, it looks to exploit its undeniable technological lead over China and China’s’ reliance on advanced western technology, but this reliance has been steadily shrinking as China’s tech capabilities increase, and with it, the benefits of technology denial.
The simple explanation for this is that export controls were designed for the 20th century tech economy, which was both sharply bifurcated and also “defense-centric” (DOD played the leading role in funding and direction), and are increasingly ineffective for the more developed and interconnected world of the 21st. Seeking to block Chinese acquisition of advanced AI capabilities only incentivizes it to invest and develop, leading to what one leading AI researcher called “a highly impressive display of research and engineering under resource constraints.”
Chips are not a chokepoint, something that undercuts US export control efforts. AI depends as much on mathematics and software as on hardware. China is a leader in creating “open source” AI software. Heavy investment by the Chinese government have speeded up the development of domestic GPUs. In 2024, Nvidia, for example, began to list Huawei as a key competitor in the AI chip market, along with western suppliers. While Chinese businesses prefer Nvidia, domestically-made GPUs are adequate.
The regulations establish a tiered licensing system putting countries into three groups. Companies in a few countries, including Japan, South Korea, Australia, the European Union and Taiwan, are granted exceptions under an AI Authorization framework allowing access to advanced GPUs without individual licenses if they meet certain standards, such as verifying end-user compliance and avoiding unauthorized transfers. At the other end of the scale are the countries of concern, including China, Russia, North Korea, and Iran and 24 U.S. Arms Embargoed Countries. Exports to these countries are strictly prohibited, with Hong Kong and Macau subject to the same restrictions. These restrictions are unobjectionable
It is the middle tier that is ill-conceived and damaging. This tier includes about 150 countries, including Singapore, Vietnam and Saudi Arabia. Exporters must follow stringent guidelines, including verifying end users and preventing unauthorized transfers. This tier has a number of puzzling inclusion, such as Switzerland and Poland, economic and security partners where exports pose no risk. The new rules also impose limitations on data centers and cloud providers, with complicated an confusing quotas that allow companies to deploy no more than 25% of their AI computational capacity outside the U.S. and only a small fraction of their total AI capacity may be deployed in a single country outside this group. Companies headquartered in the U.S. are prohibited from transferring more than 50% of their total AI capacity outside the U.S. Approval is also needed for foreign-developed AI models that use U.S.-origin GPUs or servers.
The effect of this is to incentivize the Chinese to make their own. Some countries will be reluctant to use Chinese technology over security concerns, but others will not be inhibited. The rule puts the US on track to repeat the Huawei experience, where that company dominates markets in the global the global south to the detriment of US national security. Instead of blocking China, the US should seek to lead and gain market share in a global market.