There Can Be No Winners in a US-China AI Arms Race

By: Alvin Wang Graylin & Paul Triolo (MIT Technology Review)

In this piece for the MIT Technology Review, authors Alvin Wang Graylin and Paul Triolo examine the ongoing rivalry between American and Chinese AI technologies.

During the initial phase of this geopolitical competition, US policymakers focused primarily on securing an economic advantage, framing the issue as a contest that had to be “won.” More recently, major AI research organizations like OpenAI and Anthropic have reinforced this perspective, seemingly in an effort to align with the incoming Trump administration by emphasizing the urgency of “outpacing China.” The assumption that the US could maintain dominance largely rested on its early lead in cutting-edge GPU compute power and the scaling efficiencies that have historically driven AI advancements.

However, it is now becoming evident that simply having access to vast amounts of high-performance computing is not the long-term strategic advantage many once believed it to be. In reality, the technological gap between top-tier AI models in the US and China has effectively closed. Moreover, Chinese AI models may hold a distinct edge in at least one crucial area: they are capable of producing comparable results while utilizing significantly fewer compute resources than their Western counterparts.

The framing of AI competition has increasingly shifted toward a national security lens, portraying the landscape as a zero-sum conflict. This view is further shaped by the belief that a military confrontation between the US and China—centered on Taiwan—is inevitable. In response, the US has pursued “chokepoint” strategies to restrict China’s access to vital technologies such as advanced semiconductors. Yet, rather than stalling China’s progress, these measures have fueled the country’s drive for technological self-reliance and accelerated homegrown innovation—ultimately undermining the effectiveness of US policies…

CONTINUE READING…