Rethinking the US-China AI Race: Why Diffusion Trumps Dominance

The United States’ current strategy for artificial general intelligence (AGI) is based on the premise of a decisive race against Beijing, prioritizing maximum acceleration and the denial of Chinese access to semiconductor chips and technology. However, this approach may be misdiagnosing the nature of the AI competition and could potentially degrade the long-term strategic position of the U.S. By focusing on frontier model benchmarks rather than the diffusion of technology across the economy.

This posture, as reflected in the White House AI Action Plan released in 2025, mirrors the logic of the 1940s atomic bomb race and the Cold War space race. Critics argue that treating AI as a binary capability—where one side wins by reaching a specific threshold first—ignores the fact that AI is a continuous, evolutionary technology. There is no single threshold that confers a permanent or decisive advantage.

The Diffusion Gap

While the U.S. Focuses on developing the most powerful frontier models, the economic advantage of AI is derived from the ability to diffuse these capabilities across various industries and scale them across the broader economy. In this area of deployment, China is likely outpacing the United States.

China’s strategy often prioritizes quantity, price, and speed to market over frontier quality. Evidence of this diffusion includes:

• ByteDance’s Doubao chatbot has exceeded 100 million daily active users.
• Alibaba’s Qwen models have surpassed 700 million global downloads, leading to more than 180,000 derivative models.
• Chinese open-source models are becoming the primary platform for sovereign AI efforts in the Global South and for startups globally, including those within the U.S.

These efforts are supported by structural advantages. China accounts for 54% of global industrial robot installations and produces approximately half of the world’s AI researchers. China leads in 66 of 74 critical technologies tracked by the Australian Strategic Policy Institute and adds more new electricity capacity annually than the rest of the world combined.

Commercial Ecosystems vs. State Monoliths

Washington often views the Chinese AI sector as a state-directed monolith, but it is actually a competitive commercial ecosystem with diverse business models. For example, Zhipu AI generates over 60% of its revenue from enterprise deployment services, while MiniMax earns roughly 70% from international API sales. Alibaba has used open-sourcing for Qwen to drive cloud adoption, and DeepSeek has open-sourced to attract research talent.

By framing this commercial ecosystem as a centrally planned threat, U.S. Policy responses have been either too narrow—focusing primarily on chip exports—or too blunt, restricting all Chinese AI. This focus on chips fails to address the deployment gap regarding how models are actually trained and used in practice.

Economic and Military Implications

The mismatch in strategies has created significant internal pressures in the U.S. Tech firms have committed more than $500 billion annually in AI capital expenditures for the period between 2025 and 2027. Simultaneously, U.S. Job openings have declined sharply, and World Bank data indicates that 60% of the U.S. Workforce is at risk of displacement due to AI without an adequate social safety net.

In the defense sector, the U.S. Military’s acquisition system is at a structural disadvantage. The process for vendor and model certification can take over a year. In contrast, the Chinese government reviews AI models before public release to streamline deployment. The decisive military advantage may not be the most capable model, but the ability to field AI-enabled systems across a force the fastest.

Proposed Strategic Reset

There is an opportunity for Presidents Trump and Xi to reset the terms of this competition during their meeting scheduled for May 2026. A more effective strategy would move away from the Manhattan Project model and toward Cold War-style arms control agreements that stabilize relations while allowing economic growth.

Recommended shifts in U.S. Policy include:

• Implementing allied industrial policy for AI diffusion with partners such as Japan, South Korea, the European Union, and Australia.
• Establishing targeted safety agreements with enforcement mechanisms to address cyber weapons and autonomous weapons.
• Creating domestic workforce legislation modeled on the post-WWII GI Bill to provide educational, housing, and living assistance for displaced workers.
• Building an international architecture with like-minded partners to nurture AI standards and provide guardrails for open-source civil applications.

By focusing on the safe diffusion of AI to remake the global economy rather than winning battles over chips and benchmarks, the U.S. Can ensure that AI strengthens its security and prosperity.