Prioritizing Infrastructure Foundation Over AI Model Supremacy

Global leadership in AI is often framed as a race for the most powerful model, but this view ignores how technology is actually adopted. True dominance is not about cutting-edge innovation; it is about building the foundation that the world relies on. While the U.S. focuses on model supremacy, China is using a "good enough, low cost" strategy similar to its approach to 5G, embedding its technology into global infrastructure. For policymakers and industry leaders, the advantage lies in moving from a purely competitive mindset to one that prioritizes accessibility and democratic alignment. Failing to do this risks a future where U.S. models are technically superior but globally irrelevant, as the world standardizes on more affordable alternatives.

The "Choke Point" Strategy

The competition between the U.S. and China has moved beyond espionage. Ann Neuberger, a former U.S. deputy national security advisor, points to a shift toward "digital choke points." China’s cyber program is not just gathering intelligence; it is pre-positioning within critical infrastructure like power, water, and pipelines across the U.S. and its allies.

This creates a systemic vulnerability. Because much of this infrastructure is aging and privately owned, it lacks the unified security standards needed to repel state-level actors. The consequence is clear: in a crisis, these digital footholds give China the ability to disrupt the physical operations of the U.S. and its partners.

> "We have a concern that China will use its prepositioning in digital infrastructure to disrupt the operations of those infrastructure in a period of crisis or conflict."
>
> -- Ann Neuberger

The Defensive Asymmetry of AI

Conventional wisdom suggests that AI is a tool for attackers, but Neuberger argues the opposite: AI is a force multiplier for defenders. In a traditional network, the attacker only needs to find one entry point, while the defender must secure every door and window.

AI changes this by automating the analysis of massive data streams. This allows defenders to identify strange behavior in real time and patch legacy code, work that was previously impossible to scale through human effort alone. The strategic advantage is not in building the most powerful model, but in deploying AI that can manage the complexity of existing, vulnerable systems.

The Trap of "Cutting Edge" Leadership

The most important insight from the U.S.-China dynamic is the difference in how they define leadership. The U.S. prioritizes having the best models, while China prioritizes market penetration through low-cost, "good enough" technology.

> "I think the big area of debate has been in the area of whether leadership means having the best models or whether leadership means the world running on a technology as its foundation."
>
> -- Ann Neuberger

This strategy has already worked in the 5G space. By ignoring the need for the "best" tech and focusing on cost-efficiency, China ensures its systems become the default foundation for global infrastructure. If the U.S. continues to ignore this, it will find itself in a position where its superior models are used by a shrinking minority, while the rest of the world builds on Chinese-backed open-source models.

Key Action Items

• Prioritize Infrastructure Security: Over the next 12 to 18 months, shift focus from AI model development to securing the aging, critical water and power systems. This is an immediate, unglamorous necessity.
• Adopt "Good Enough" Accessibility: To compete with Chinese open-source dominance, U.S. companies must lower the barrier to entry for their models. Recognizing that cost is the primary driver for global adoption is a requirement for maintaining influence.
• Establish Bilateral Hotlines: In the immediate term, the U.S. should pursue dedicated hotlines for AI-enabled incidents to prevent miscalculation, similar to past collaborations on fentanyl precursor controls.
• Focus on Dual-Use Risk Mitigation: Work toward narrow, high-impact agreements with China, specifically preventing AI models from assisting in the generation of chemical or biological weapons. This is a rare area where shared risk creates a foundation for cooperation.
• Shift from "Model-First" to "Foundation-First": For policymakers, the metric of success should not be the capability of the latest model, but the number of countries and companies that adopt democratic, U.S.-aligned technological foundations. This is a long-term investment that requires patience, not just hype.