The paper is titled "2028: Two scenarios for global AI leadership" and it is exactly what it says. Anthropic's research team lays out two possible worlds. In the first, the US tightens chip export controls, blocks China's access to American frontier models, and the democratic coalition holds a 12-to-24-month compute lead by 2028. In the second, policymakers hold back, China closes the gap, and the norms governing AI globally are written by authoritarian states.

The paper places semiconductors at the center. Not capabilities. Not safety frameworks. Chips. "There is a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI." That sentence is the thesis, and the rest of the paper is the argument for it. The paper published on anthropic.com while Trump was in Beijing for the first time in nearly a decade. The timing was not accidental.