RAND Wargames Show How AI Could Reshape U.S.-China Conflict

RAND Wargames Show How AI Could Reshape U.S.-China Conflict

Researchers at RAND Corporation have completed ten iterations of an operational-level wargame examining how artificial intelligence and autonomous systems might transform military campaigns in a future U.S.-China conflict over Taiwan. The unclassified exercise, called Camp(ai)gn, involved more than 20 RAND researchers and affiliates simulating scenarios set in 2035.

Why it matters

As both the United States and China invest heavily in military AI, understanding how autonomous systems could alter conflict dynamics—including escalation patterns, force requirements, and allied roles—becomes essential for defense planning, industrial policy, and alliance strategy. These simulations suggest that AI-enabled warfare may not follow the patterns military planners currently anticipate.

Six Key Observations from the Simulations

The wargames produced several findings that challenge conventional assumptions about great-power conflict.

First, autonomous systems could enable new forms of limited warfare—graduated, low-intensity operations fought primarily with robotic systems rather than human forces. This represents a departure from traditional concepts of great-power war.

Second, these limited conflicts could produce escalation dynamics and end states that differ substantially from canonical visions of great-power wars. The presence of autonomous systems may create off-ramps or pressure points that don't exist in conventional conflicts.

Third, autonomous mass—the ability to field large numbers of unmanned systems—could become especially important during protracted conflicts. The simulations suggested this factor might favor China in certain scenarios.

Fourth, autonomous mass could expand commanders' decision space and reshape familiar operational trade-offs. However, it also increases the cognitive burden on commanders who must manage larger, more complex forces.

Fifth, AI and autonomous systems could introduce new roles for allies beyond traditional combat participation, including self-defense missions and coalition production of autonomous systems.

Sixth, attacking an adversary's data centers is unlikely to deliver decisive operational effects. The simulations suggest that AI-enabled forces may not have a single technological Achilles' heel that can be exploited for strategic advantage.

Research Implications

The RAND team emphasized that these observations represent gameplay-generated hypotheses rather than established conclusions. The findings point toward specific areas requiring further research, including force structure requirements, countermeasures against autonomous systems, allied integration strategies, and defense industrial capacity for producing autonomous systems at scale.

The research was independently initiated and conducted by the Center for the Geopolitics of Artificial General Intelligence within RAND Global and Emerging Risks, using operational income and donor support.

These findings were first reported by RAND Corporation in a new expert insights paper examining the Camp(ai)gn wargame series.