FEBRUARY 2025
The AI Diusion Framework
Securing U.S. AI Leadership While
Preempting Strategic Drift
By Barath Harithas
Introduction
In one of its nal acts, the Biden administration introduced what may be the most ambitious exercise
of technological statecraft in modern history—the Framework for Articial Intelligence Diusion
(AI Diusion Rule). What was once a simple line drawn around China and a small handful of countries
has turned into a sweeping, interventionist framework that dictates not just who can import advanced
semiconductors and how many but where they can be deployed, under what conditions they must be
secured, and even how AI capabilities can be shared.
Unpacking the U.S. Calculus
In some ways, the framework is a carefully modulated equation, attempting to balance multiple, often
competing variables in a single formula. On the one hand, the United States needs to prevent the
diversion of advanced graphics processing units (GPUs) to China, increasingly through third countries,
to ensure that China never reaches escape velocity in the AI race. DeepSeek may have tilted the
board, but the equation holds—China still faces stubborn compute bottlenecks in inference scaling
and deployment.
On the other hand, the exponential growth in compute demand makes it neither practical nor
expedient to limit AI infrastructure to U.S. soil. American and allied rms must expand overseas,
but every new deployment introduces an additional vector for diversion. This tension is further
complicated by market dynamics. U.S. and allied hyperscalers and neocloud providers operate on
economic logic, not national security priorities. They will naturally gravitate toward locations where
power is cheap, infrastructure hurdles are minimal, and nancial incentives sweeten the deal. But
for the United States, these same AI-friendly jurisdictions may not always be reliably aligned with its
priorities, nor can they be easily insulated from Beijing’s reach.
