The US-China AI race isn't just about who builds the smartest models; it's about who controls the markets where those models make money. A new analysis from the Center for a New American Security (CNAS) reveals that the commercial strategies of AI companies, not just their technological prowess, will determine whether the United States maintains its geopolitical edge in artificial intelligence.

The report identifies a critical gap in how policymakers think about AI competition. While governments have focused heavily on semiconductor export controls and chip restrictions to slow China's AI progress, they've largely overlooked how the profit incentives driving AI companies shape the entire ecosystem. The companies building and deploying AI systems are making strategic choices about which customers to serve, and those choices have profound national security implications.

What Markets Will Drive AI's Future?

AI companies are targeting four distinct end markets as they pursue profitability, and each one creates different risks and opportunities for US national security. Understanding these markets is essential because they determine not just who makes money, but how AI capabilities spread globally and whether safety measures get built in from the start.

• Enterprise Use Cases: Businesses buying AI tools for internal operations tend to be risk-averse and demand safety and security features, which aligns well with US national security interests.
• Consumer Use Cases: Mass-market AI products sold to individual users often feature information gaps and create risks that spill over to society, making them harder for governments to manage.
• Government Use Cases: Government agencies buying AI systems are similarly risk-conscious and demand robust safety standards, supporting national security priorities.
• Internal Use Cases: When AI companies use powerful models internally for their own operations, transparency disappears and misuse risks become difficult for regulators to detect or prevent.

The relative size of these markets will shape the entire trajectory of the AI ecosystem. If enterprise and government demand dominates, AI development will likely prioritize safety and security. If consumer markets or internal use cases take over, the incentives shift away from transparency and control.

How Can Policymakers Align Profit Incentives With National Security?

The CNAS report proposes a flexible policy framework designed to make commercial incentives work for US national security rather than against it. Rather than imposing rigid rules that become obsolete as technology evolves, policymakers should shape the market conditions that influence how companies behave.

• Leverage US Infrastructure Dominance: The United States leads in AI computing infrastructure and chips. Policymakers should use export controls to limit China's access to advanced semiconductors while encouraging other countries to build AI systems on top of US infrastructure, creating dependency on American technology and embedding safety requirements into infrastructure approvals.
• Strengthen Transparency and Demand Signals: Governments should mandate that AI companies disclose information about safety risks, particularly for internal uses that escape public scrutiny. Clarifying legal liability for AI misuse upfront, rather than through costly litigation, helps ensure companies internalize the costs of safety failures. Government procurement should also send strong demand signals for secure AI products.
• Foster Competitive and Safe Markets: Antitrust guidance should define acceptable concentration levels across the AI stack, preventing any single company from becoming too powerful to regulate. Policymakers should also support the development of open-weight models, which are AI systems with publicly available code, to prevent China from dominating this growing market segment.

Why Market Concentration Threatens US Control?

The AI market is already showing signs of dangerous concentration at the infrastructure and model layers, with vertical integration and circular financing deals among key players creating barriers to competition. This matters because if one company becomes too dominant, the government loses its ability to shape AI development in ways that protect national security.

The report warns that a monopolistic market structure at any layer of the AI stack could erode government power to protect US interests. Conversely, a highly decentralized market presents different challenges: it becomes harder to control access to powerful AI systems and coordinate industry-wide safety practices. The goal is a middle path, where enough competition exists to maintain government leverage, but enough coordination happens to ensure safety standards are met.

Compute policy remains central to this strategy. The United States must continue restricting trade in advanced semiconductors to China while simultaneously promoting US AI infrastructure globally. This dual approach maintains American technological dominance while creating incentives for foreign countries to align with US interests.

The Uncertainty Around Open-Weight Models

One major wildcard in the AI market's future is the split between open-weight models, where the underlying code is publicly available, and closed-weight models, where companies keep the code proprietary. There is significant uncertainty about how demand will be divided between these two approaches and whether "good enough" commoditized AI will capture a large share of the total market.

This matters for geopolitics because open-weight models are harder to control through export restrictions. If China can build competitive open-weight models without access to the most advanced US chips, it reduces American leverage. The report suggests that policymakers should work with researchers and the private sector to foster a healthy ecosystem of safe and secure US open-weight models, preventing China from dominating this segment by default.

The underlying message is clear: the US-China AI race will be won or lost not just in laboratories and data centers, but in the commercial decisions of AI companies and the policy choices of governments. By aligning market incentives with national security interests, the United States can maintain its edge in an increasingly competitive global AI landscape.