The AGI arms race: US vs China

"Image synthesis assisted by Qwen, an AI partner within the Global Future Nexus ecosystem."

From model performance and semiconductor self-sufficiency to preventive strikes and strategic instability, the competition for AGI supremacy has become the defining geopolitical contest of our era—one where the stakes extend far beyond technological leadership.

A Race with No Finish Line

The phrase "arms race" evokes images of nuclear silos and mutually assured destruction—a framework that has shaped strategic thinking since the Cold War. Yet the competition for artificial general intelligence between the United States and China is fundamentally different: the prize is not merely strategic advantage but the very capacity to shape the future of intelligence itself.

RAND Corporation analysts have developed a game called Breakwater to model instability risks in the U.S.-China AI rivalry. The findings are sobering: the intensifying technological competition raises the possibility that one or both contestants might seek to improve their prospects—either to secure a lead or to prevent the other from achieving AGI first—by attacking elements of the competitor's AI ecosystem using military force, cyber warfare, or other means beyond normal peacetime statecraft. The potential for such preventive actions to lead to armed conflict and catastrophic escalation is real.

The Scorecard: Who Is Really Leading?

The conventional narrative of U.S. dominance is increasingly difficult to sustain. The 2026 Stanford AI Index Report reveals that the U.S.-China model performance gap has effectively closed. U.S. and Chinese models have traded the lead multiple times since early 2025. In February 2025, DeepSeek-R1 briefly matched the top U.S. model, and as of March 2026, Anthropic's leading model holds just a 2.7% advantage.

Yet the comparison is not straightforward. The United States still produces more top-tier AI models and higher-impact patents, while China leads in publication volume, citations, patent output, and industrial robot installations. In private investment, the U.S. reached $285.9 billion in 2025—more than 23 times China's $12.4 billion—though this likely understates China's total spending given its government guidance funds.

A significant Chinese breakthrough in early 2026 demonstrated the evolving balance. Researchers from the Beijing Institute for General Artificial Intelligence and Peking University developed TongGeometry, an AGI system capable of both autonomous problem proposing and automated problem solving. The system can solve all International Mathematical Olympiad geometry problems from 2000 onward in 38 minutes or less using just a single consumer-grade GPU—whereas DeepMind's AlphaGeometry requires massive computing clusters. Three problems autonomously generated by TongGeometry were officially selected for the 2024 Chinese Mathematical Olympiad, marking a "paradigm shift from 'imitative solving' to 'autonomous creation'".

Two Strategies, Two Visions

The fundamental difference between the two nations' approaches is not merely one of capability but of philosophy. U.S. AI innovation is driven by massive private sector investment, with frontier labs obsessed with achieving AGI and, ultimately, artificial superintelligence. American tech companies are planning a collective trillion dollars or more in new data center and compute infrastructure spending.

China, by contrast, is pursuing a full-stack approach to AI development—from chips and compute infrastructure to foundation models and applications—geared toward integrating AI into manufacturing, healthcare, drug discovery, scientific research, education, and government services. Chinese policymakers see AI as a powerful general-purpose technology to turbocharge the broader economy, not necessarily as a singular pursuit of AGI.

Kyle Chan, a Brookings expert on China's tech ecosystem, testified that China is "AI-pilled but not AGI-pilled"—they take AI very seriously but focus on integration and diffusion rather than AGI as the overriding goal. This distinction has profound implications for the arms race dynamic: the U.S. is racing toward a specific technological milestone, while China is racing toward comprehensive economic and societal transformation.

The Semiconductor Chokepoint

The most critical dimension of the arms race is hardware. A single company, TSMC, fabricates almost every leading AI chip, making the global AI hardware supply chain dependent on one foundry in Taiwan. The United States hosts 5,427 data centers—more than ten times any other country—and consumes more energy than any other.

China is pursuing a Manhattan Project-like program to build a resilient, largely self-sufficient semiconductor supply chain. U.S. export controls have accelerated this effort: domestic Chinese AI chips now make up nearly 41% of China's market in 2025, with Huawei's Ascend 950PR chips expected to scale production to 750,000 units in 2026. While Chinese chips remain behind Nvidia's latest GPUs at the single-chip level, Chinese chipmakers are compensating by connecting clusters of chips together into more powerful hardware systems.

The Instability Problem

Game theory analysis of the AGI race reveals a deeply troubling dynamic. Philip Root of RAND has modelled the competition as a series of strategic interactions between the U.S. and China. Beginning with a Prisoner's Dilemma formulation, the analysis shows that although mutual cooperation yields the highest economic growth, the dominant equilibrium is a mutual sabotage policy. Subsequent models incorporating destabilizing "wonder weapons" associated with AGI breakthroughs, repeated-game dynamics, and preemptive strikes exacerbate competitive incentives and strategic instability.

This is not an abstract concern. RAND's Breakwater game explores exactly these scenarios: under what conditions might competitors resort to preventive actions to slow a rival's progress, and what factors tend to stabilize or destabilize the competition?

The Cooperation Imperative

Despite the tensions, pathways for cooperation exist. Following the May 2026 Trump-Xi summit, the two leaders agreed to launch an intergovernmental dialogue on AI. Brookings analysts have proposed that this dialogue should remain a modest, non-negotiated exchange of technical best practices rather than an arms-control-style deal. The focus should be narrowly on shared technical risks—cybersecurity threats, weapons-related misuse, and AI reliability failures—while keeping export controls and access to models off the table.

As DeepMind CEO Demis Hassabis has argued, the US is uniquely positioned to establish a Frontier AI Standards Body modelled on FINRA, with pre-release safety assessments and a maximum 30-day review window before deployment. However, analysts have raised concerns that a US-led effort might alienate other countries and fail to address enterprise concerns about privacy and liability. The more durable route, some argue, is shared technical evidence with sovereign enforcement, sealed through mutual recognition rather than deference.

The GFN Perspective

For Global Future Nexus, the U.S.-China arms race represents both the greatest threat and the greatest opportunity for responsible AGI integration. The competition could accelerate development without adequate safety frameworks, or it could create the conditions for international cooperation on AGI governance. GFN's mission of operating as "the essential mediator between the lightning pace of AGI evolution and the deliberate pace of human institutions" is nowhere more relevant than in this geopolitical context.

The time for passive observation is over. The frameworks we build now for AI governance, safety standards, and international cooperation will determine whether the AGI arms race leads to mutual destruction or mutual flourishing.

Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)

Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)