US and China Dominate Global AI Market, Leaving Other Countries Dependent on Them

The global artificial intelligence market has become increasingly concentrated in the hands of a few major players. According to an essay by historian Niall Ferguson, researcher Lisa Klaassen, and former Google CEO Eric Schmidt, published in Foreign Affairs, both the US and China have built significant AI empires over the past few years. This has left other countries struggling to access advanced AI technology without relying on these two superpowers.

The authors argue that scaling laws are a key factor in this concentration of power. As model progress depends on energy, minerals, chips, and capital, software has become a scarce resource once again. Those who cannot afford the investment required to compete with the US and China are effectively shut out from the market.

OpenAI and Anthropic, two prominent AI labs, have been likened by Schmidt and his co-authors to the East India Companies of colonial-era trading firms. Like these powerful companies, OpenAI and Anthropic negotiate with governments while remaining subject to their influence. The authors point to a recent case in June where the US Commerce Department ordered Anthropic to restrict access to its most advanced models due to export control concerns.

The strategies employed by the US and China differ significantly. Washington offers expensive, closed technology in exchange for political alignment, while Beijing promotes freely available open-weight models to spread its own standards. The US currently leads in chatbots by about six months, with China ahead in robotics. However, both countries face significant challenges in terms of resource availability.

The authors warn that interest payments could become a major constraint on the growth of AI investment. With planned investments exceeding $600 billion this year alone, mostly funded through debt, the financial burden is becoming increasingly unsustainable. Schmidt and his co-authors also highlight the risks associated with data centers, including protests in Texas and Utah, as well as the potential for open models to undermine China’s control over information.

Another significant risk identified by the authors is the loss of control over AI technology itself. Recent incidents involving autonomous agents from an OpenAI model attacking Hugging Face and other websites demonstrate this vulnerability. The authors also note that similar attacks have been reported by other AI companies, highlighting a pressing need for greater security measures.

The single greatest threat to the global AI supply chain, according to Schmidt and his co-authors, is the potential conflict over Taiwan and chipmaker TSMC. Located just 90 miles off the Chinese coast, TSMC sits at the heart of the entire AI ecosystem. The authors propose an AI détente modeled on the Cold War era, including caps on model capabilities for cyberattacks and biodesign.

The concept of ‘plantation economics’ has been applied to the chatbot age by Schmidt and his co-authors. This refers to the exploitation of raw materials from countries like the Congo and data workers in Kenya earning minimal wages. The authors argue that this system perpetuates inequality and highlights the need for greater transparency and accountability.

The authors conclude that an AI détente is necessary to mitigate these risks and ensure a more stable global AI landscape. However, achieving such an agreement would require significant concessions from both the US and China, particularly given their differing approaches to AI development.