Noah Smith

U.S.–China rivalry is not inevitable, but Washington no longer controls the relationship. Noah Smith argues the countries cooperated through most of the 20th century and remain culturally similar, yet similar powers “fight for niches.” His central historical call is that Xi Jinping defected first from engagement: America must stop imagining it “makes all the choices” when China is now “bigger than us in most of the important ways.”

The concrete contest is over trade routes and strategic optionality, not whether the renminbi becomes the reserve currency. Noah says China seeks greater control over the seas and digital infrastructure through which physical and services trade move. The desired end state is not Chinese-speaking Americans but an America pushed toward the geopolitical periphery: “They want you to be Argentina to them.”

Semiconductor controls are meant to preserve a military-technology edge, including by limiting China’s ability to mass-manufacture leading-edge chips for autonomous weapons. Nuclear weapons deter existential destruction, Noah argues, but cannot credibly answer smaller territorial moves; conventional power still determines whether the U.S. “gets a vote.” Advanced onboard compute could let precision drones defeat jamming and operate autonomously, partially offsetting China’s greater manufacturing capacity.

The effective choke points are chipmaking equipment, software, and personnel—not trying to keep finished chips out of China. Noah expects restrictions on chip sales to fail and calls them largely “cosmetic,” because slightly off-frontier Nvidia parts may suffice. He believes machinery and talent restrictions are already slowing domestic fabrication, citing China’s costly, low-volume 7-nanometer effort, while conceding nobody knows how much that will matter in an autonomous-drone war that does not yet exist.

DeepSeek does not persuade either speaker that China’s LLMs are the core security threat. Nathan places it roughly fourth or fifth globally; Noah sees LLMs as a “borderless, commoditized, universal thing” and says company- or country-level LLM supremacy probably will not work. Nathan reads an open model more as a “peace offering” than a threat. For investors, the distinction is between controls that may constrain leading-edge manufacturing and broadly diffusing model intelligence.

The sharpest unresolved risk is that containment could accelerate the very conflict deterrence is supposed to prevent. Nathan argues that denying China Taiwan-made chips lowers one cost of attacking Taiwan, while proposed Western AI alliances could create a “do something sooner rather than later” incentive. Noah’s answer is deliberately limited—“my blog is called Noahpinion for a reason”—because he understands precision weapons better than AGI-to-military-superiority scenarios.

Noah’s policy horizon is to buy roughly 10 years of stability, then repair the relationship. He calls the coming decade the peak-risk period, likens it to the 1950s, and expects bellicosity could diminish after Xi’s fourth term, when Xi will be about 80. The closing thesis is Cold War-style deterrence rather than appeasement or preventive war: “We’ll be friends with China again someday,” but unilateral weakening will not get there faster.

🔗 Original source & video: US-China Relations: History, Culture, and AI Competition, with Noah Smith, from Econ 102