Verifying The Stack: How to Promote Global Adoption of US AI While Preventing Adversary Exploitation
The United States is navigating a decisive point at the intersection of AI technology, infrastructure, security, and policy.
AI is the pivotal technology of our era, and a global race is underway for technological dominance, market share, and the ability to shape the rules and standards that will govern AI worldwide for a generation. Whoever wins that race will command compounding advantage in economic productivity, social organization, and strategic capability. The race is fundamentally for full-stack AI deployment: chips, servers, data centers, storage, power, cloud, models, and the trusted relationships that sustain them. It is, in other words, a competition to determine whose infrastructure becomes the foundation of the emerging AI era. The stakes of that competition are difficult to overstate.3
Over $1 trillion in new AI infrastructure demand is projected through 2027.4 Large-scale AI programs across the Middle East and Indo-Pacific are translating that demand into hyperscale data center builds, national champion platforms, and multi-billion-dollar investment frameworks anchored to the American technology stack.
The policy debate in Washington has organized itself around two policy objectives that may appear at odds: “Running Faster”: accelerating American AI deployment to win the global technology race, and “Capability Denial”: preventing adversaries from exploiting US technology to accelerate their own programs or achieve decisive strategic capabilities. This paper argues that these objectives, while understandable given the real tensions involved, are complementary. Both are correct in their diagnosis, and they are not inconsistent. The United States must both run faster and deny exploitation, simultaneously, by design.
This paper is written in the context of the June 3, 2026, convening, “Verifying the Stack,” which brought together pragmatists from policy, technology, and integration to address the WHY, WHAT, and HOW of this challenge. It is intended as supporting architecture for that conversation, not a single silver bullet, but a modular foundation upon which multiple complementary answers can be built.
References
3. This paper frames the AI build out as a competition. The authors acknowledge that other framings are possible and, in some respects, more complete. The global AI future could be conceptualized as a field that multiple nations grow together, producing mutual benefit through interoperable systems and shared standards. There are scenarios in which the optimal outcome is not zero-sum dominance but collaborative leadership. This paper is aware of and respects those alternative framings. It adopts the race framing because the decisions being made now about whose infrastructure, standards, and governance architecture become foundational are not abstract. They carry real consequences, commercial and technological, but also moral and security-related. The values embedded in the infrastructure that powers the AI era will shape the future lived experience of humanity. In the authors’ view, those stakes justify the urgency and directness of the competitive framing, while remaining open to a future in which that race produces conditions for broader collaboration.
4. “Nvidia CEO Jensen Huang: $1 trillion in chip sales coming,” Axios, March 16, 2026.
