Winning the AI Race: A Three-Pillar Strategy for US Leadership

The high-stakes world of technology competition often feels like a chess match played on a global scale.

Each move, each innovation, each strategic alliance, carries immense weight.

For decades, my grandfather, a quiet man who spent his life in engineering, would pore over schematics, understanding the intricate balance of power that lay in scientific advancement.

He’d talk about the delicate dance between sharing knowledge and protecting it, a conversation that feels more urgent than ever as we confront the era of artificial intelligence.

Today, the United States finds itself at a pivotal juncture in the AI Race, where leadership is not a given, but a prize to be strategically won and fiercely defended.

This isn’t just about silicon and software; it’s about national interest, economic prosperity, and the very fabric of 21st-century power.

In short: To win the AI race, the US needs a coordinated technology strategy: promote its tech stack, proactively counter China’s circumvention of export controls, and cooperate with allies to build a resilient technology order, as outlined by the Center for a New American Security.

Why This Matters Now: Beyond the Headlines

Artificial intelligence is no longer a futuristic concept; it is central to 21st-century power, making US leadership in frontier AI development critical for national interests (Center for a New American Security CNAS, 2025).

Sustaining this leadership demands a strong position in the AI Tech Stack, which encompasses everything from technical talent and training data to algorithmic innovation and critical hardware components like advanced semiconductor chips and the specialized tools to manufacture them.

These chips form the very computing foundation for advanced AI systems.

Fortunately, the United States currently holds a significant comparative advantage in this AI dominance race.

For instance, the US maintains a lead in total compute capacity worldwide, owning roughly ten times more advanced AI chips than China for research and development, training, and sustaining AI deployment across society (CNAS, 2025).

Preserving this production gap should be the linchpin of American technology strategy, ensuring that competitors like China cannot easily replicate US capabilities despite their heavy investments.

The Core Challenge: Fragmentation and Leakage in the AI Ecosystem

While the US possesses a critical lead, its position in the AI Race is neither self-sustaining nor guaranteed.

The core problem lies in the fragmented nature of current policy approaches, particularly concerning export controls on AI components.

Export controls, now central tools in great-power technology competition, aim to slow an entire technological ecosystem, restricting not just AI chips but also high-bandwidth memory (HBM), semiconductor manufacturing equipment (SME), and the technical know-how to produce them (CNAS, 2025).

However, these controls have largely been implemented in isolation, with unclear objectives and haphazard enforcement (CNAS, 2025).

This reactive approach creates a game of regulatory whack-a-mole that China has proven adept at circumventing.

New front companies and routing mechanisms emerge faster than regulators can respond, allowing significant leakage through direct sales, third-party transfers, and inadequate verification (CNAS, 2025).

This asymmetry favors the evader over the enforcer, systematically undermining US efforts to preserve its strategic advantage.

The counterintuitive insight here is that sometimes, commercial interests can inadvertently fuel an adversary’s capabilities.

Direct leakage of controlled chips and SME, alongside domestic market sales from companies like Nvidia directly into China, effectively supports competitor scaling and funds their efforts to eliminate dependency altogether (CNAS, 2025).

Policymakers face a critical choice: either treat export controls as isolated bargaining chips in trade agreements or integrate them into a comprehensive strategy for long-term Technology Competition (CNAS, 2025).

The past, where a Trump Administration suspended Entity List affiliates for a year in exchange for China’s suspension of rare earth controls, showed that such controls can be viewed as tradeable, risking long-term technological competition for near-term trade agreements (CNAS, 2025).

What the Research Really Says: A Three-Pillar Strategy for Victory

A recent playbook from the Center for a New American Security (CNAS, 2025) outlines a comprehensive strategy built on three primary elements: Compete, Counter, and Cooperate.

This framework demonstrates how export controls, rather than being reactive mechanisms, can transform into proactive strategic instruments that reinforce US AI leadership.

Maintaining US Leadership in Frontier AI Requires Controlling Access to Advanced Semiconductor Chips

The so-what: US strategic advantage hinges on preventing competitors from accessing critical chip technology.

The implication: US strategy should focus on preserving its technological lead and making it prohibitively expensive for competitors to replicate US capabilities, ensuring the compute capacity gap remains wide (CNAS, 2025).

Current US Export Controls Are Reactive and Fragmented, Leading to Circumvention and Undermining Strategic Goals

The so-what: Isolated controls are ineffective against sophisticated evasion.

The implication: Policymakers must integrate export controls into a coordinated technology strategy across government, industry, and allies to move from reactive responses to proactive enforcement, establishing a clear Digital Geopolitics strategy (CNAS, 2025).

Direct Leakage, Indirect Diffusion, Domestic Market Sales to China, and Treating Controls as Bargaining Chips Undermine US AI Advantage

The so-what: Multiple pathways actively erode US technological superiority.

The implication: A fundamental reframe is needed to treat export controls as a comprehensive technology competition strategy, integrating denial measures with positive inducements, multilateral coordination, domestic capacity building, and industry partnership (CNAS, 2025).

US Industry Can Compete in the Chinese Market While Maintaining Control by Shifting from Outright Chip Sales to Leasing Compute Through Cloud Service Providers

The so-what: This model offers a way to balance commercial interests with national security.

The implication: Formalizing this existing practice as the primary mechanism for providing compute to China would allow the US to retain ownership and control, preventing unauthorized military or surveillance applications while still serving legitimate commercial demand (CNAS, 2025).

Your Playbook for Winning the AI Race

Christian Chung, a former US intelligence officer and a key voice in this research, emphasizes that coordinated implementation of the Compete-Counter-Cooperate framework creates the foundation for sustained US technological leadership (CNAS, 2025).

Here’s how to translate that into action:

• Prioritize Promoting the US Tech Stack (Compete).

  Invest in domestic industrial capacity to expand an affirmative vision for US AI around the world (CNAS, 2025).

  This includes efforts to establish federal preemption of AI governance and standardization through the Commerce Department’s Center for AI Standards and Innovation.

• Shift to Compute Leasing for China.

  Formalize the practice of US industry leasing compute through cloud service providers to Chinese entities, rather than selling physical chips outright (CNAS, 2025).

  This allows the US to retain ownership and control while serving legitimate commercial demand, and cut off access in contingencies.

  The Remote Access Security Act, which passed the House Foreign Affairs Committee in April 2025, provides a framework for this.

• Implement Proactive Export Control Enforcement (Counter).

  Move beyond reactive policy development.

  This requires maintaining current controls on critical chokepoints, expanding selectively to cover loopholes (like the deep ultraviolet immersion, DUVi, loophole), deepening allied coordination, and enabling continuous monitoring of diversions and capabilities (CNAS, 2025).

• Strengthen Intelligence Integration at BIS.

  Embed intelligence collectors and analysts at the Bureau of Industry and Security (BIS) to run targeting cells on military-civil fusion end-users, front companies, and logistics hubs (CNAS, 2025).

  This transforms end-user validation into an ongoing intelligence-targeting mission.

• Forge a Resilient Allied Technology Order (Cooperate).

  Build a CoCOM 2.0 coalition for AI and semiconductors, combining existing alliances (US-Netherlands-Japan, Fab 4) and expanding to include the UK, EU, Canada, Australia, Singapore, and Malaysia (CNAS, 2025).

  This intelligence-fed coalition harmonizes export control packages and enforcement, eliminating arbitrage opportunities.

• Cultivate Industry Partnership.

  Treat industry as a partner in enforcement.

  Companies possess critical technical knowledge and market intelligence that governments often lack.

  Lightweight technical standards, like chip location verification technology, can reduce compliance costs while improving assurance (CNAS, 2025).

The Shadows Behind the Shine: Risks, Trade-offs, and Ethics

Pursuing a robust AI Race strategy, while vital for US Leadership, is fraught with risks and ethical dilemmas.

The very strength of export controls can provoke retaliatory measures or accelerate indigenous development in competitor nations, potentially creating unintended consequences.

There’s a constant trade-off between maximizing commercial opportunities in a vast market like China and the imperative of National Security.

Allowing domestic market sales, for instance, generates revenue for US firms but can also indirectly fund competitor scaling and efforts to eliminate dependency (CNAS, 2025).

Ethically, the dual-use nature of AI capabilities—its potential for both groundbreaking innovation and military or surveillance applications—demands a vigilant approach.

The transfer of permanent capabilities to Chinese data centers, beyond US control, raises concerns about human rights and global stability (CNAS, 2025).

The integrity of multilateral coordination can also be strained if allies perceive US actions as driven by narrow economic interests rather than shared security threats.

Mitigation involves unwavering transparency with allies about shared threats, ensuring that export control decisions are framed purely through a national security lens, not as bargaining chips for broader trade deals (CNAS, 2025).

Continuous monitoring, intelligence sharing, and embedding ethical considerations into policy development are essential to balance strategic advantage with global responsibility.

The AI Strategist’s Toolkit: Tools, Metrics, and Cadence

Tools of the Trade

Leverage the Bureau of Industry and Security (BIS) for core export control tools, including the Entity List and the Foreign Direct Product Rule.

Integrate intelligence community insights to proactively identify circumvention networks.

Utilize cloud service management platforms for leasing compute to maintain ownership and control over advanced AI chips.

Employ chip location verification technology and other lightweight technical standards to enhance compliance assurance (CNAS, 2025).

Key Metrics to Monitor

Focus on metrics that gauge both the effectiveness of denial measures and the growth of US technological advantage:

• Compute Capacity Gap: Monitor the relative difference in advanced AI chips held by the US (roughly ten times more than China, CNAS, 2025) and its allies versus competitors.
• Leakage and Diversion Rates: Track instances of banned GPUs entering black markets or restricted entities accessing chips via intermediaries (e.g., iFlytek accessing H100-equivalent chips via AWS, CNAS, 2025).
• Indigenous Production Acceleration: Assess the pace and quality of competitor efforts to produce advanced chips and SME domestically.
• Allied Coordination Score: Measure the degree of harmonization in export control packages and the effectiveness of intelligence-sharing architectures within the CoCOM 2.0 coalition (CNAS, 2025).
• US AI Tech Stack Growth: Track investment in domestic industrial capacity and the success of AI governance and standardization initiatives.

Review Cadence

Establish a dynamic and adaptive review cadence.

Implement continuous monitoring of diversions and capabilities through integrated BIS-intelligence targeting cells.

Hold monthly or quarterly multilateral coordination meetings with allies to harmonize export control packages and adapt to evolving threats.

Conduct annual strategic reviews to assess the overall effectiveness of the Compete, Counter, Cooperate framework against long-term national security objectives and the global AI landscape.

Glossary

• AI Tech Stack: The collective capabilities and infrastructure needed to develop advanced AI systems, including technical talent, training data, algorithmic innovation, and key hardware components like semiconductor chips.
• Agentic AI: AI systems capable of perceiving their environment, making decisions, and taking actions to achieve specific goals, often interacting with other AI agents or humans.
• HBM (High-Bandwidth Memory): A high-performance RAM interface for 3D-stacked synchronous dynamic random-access memory (SDRAM) that is often used with GPUs and AI accelerators.
• SME (Semiconductor Manufacturing Equipment): The specialized tools and machinery required to design and manufacture advanced semiconductor chips.
• BIS (Bureau of Industry and Security): A bureau of the US Department of Commerce responsible for implementing and enforcing export controls.
• CoCOM (Coordinating Committee for Multilateral Export Controls): A Cold War-era multilateral regime that successfully slowed Soviet technological access through coordinated export controls.

FAQ

• Q: Why are export controls critical in the AI race?

  A: Export controls on AI components, especially advanced semiconductors, are crucial because they aim to slow an entire technological ecosystem, maintaining US advantages in frontier AI and making it prohibitively expensive for competitors like China to replicate US capabilities (Center for a New American Security CNAS, 2025).

• Q: What are the main challenges undermining current US export controls?

  A: Current controls are undermined by direct leakage (smuggling, non-compliance), indirect diffusion to military uses, domestic market sales to China, and the risk of treating controls as bargaining chips in broader trade agreements (Center for a New American Security CNAS, 2025).

• Q: What is the ‘Compete, Counter, Cooperate’ framework?

  A: It’s a three-pillar strategy for winning the AI race: ‘Compete’ to promote the US tech stack and secure market advantage, ‘Counter’ to aggressively thwart China’s circumvention efforts, and ‘Cooperate’ with allies to build a resilient technology order (Center for a New American Security CNAS, 2025).

• Q: How can the US compete in the Chinese AI market while maintaining control?

  A: The US can shift from selling physical chips outright to China to leasing compute through cloud service providers.

  This allows the US to retain ownership and control while serving legitimate commercial demand, and cut off access in contingencies (Center for a New American Security CNAS, 2025).

Conclusion

The chessboard of global AI competition is complex, but the path forward for the United States is clear: strategic action is paramount.

Without Compete, the US cannot maintain advantages worth protecting.

Without Counter, those advantages diffuse to adversaries.

Without Cooperate, the United States faces a larger competitor alone (CNAS, 2025).

Just as my grandfather meticulously planned each engineering design, so too must the nation approach this AI Race with precision, foresight, and a unified front.

The question is not whether to use export controls, but whether to wield them strategically to secure sustained US technological leadership and a resilient global technology order.

The future of AI will be written not just in code, but in the deliberate choices made today.

References

• Center for a New American Security (CNAS).

  (2025).

  A Playbook for Winning the AI Race: Compete, Counter, Cooperate.