Winning the AI Race is More Complicated Than You Think
Being first to the frontier is not enough for victory
by Colin H. Kahl and Jim Mitre
Washington is abuzz about the "AI race" between the United States and China. The narrative is often simplified to a sprint for the ultimate prize: artificial general intelligence (AGI), a hypothetical AI that could outthink the smartest humans. In a recent piece in Foreign Affairs, "The Real AI Race," we argue this view is dangerously incomplete.
The United States is not in a single race but in at least five distinct AI races with China:
The race to innovate
The race to integrate AI into national security organizations
The race to integrate AI into the economy, while managing disruptions
The race to shape the global AI tech ecosystem
The race to the bottom
While the United States might have a narrow lead in the sprint for AGI, its position in the other four races is far more precarious. Losing these could mean ceding global leadership, economic prosperity, and military security to Beijing.
Here’s a breakdown of the five races and what the U.S. government (USG) needs to do to win them.
1. The Race to Innovate
This is the race for AGI. U.S. companies like OpenAI, Google DeepMind, and Anthropic currently have an edge, thanks in part to controls on advanced semiconductor exports to China.
The Problem: This lead is fragile. China is catching up fast through indigenous innovation and intellectual property theft. Its top models are only months behind their U.S. counterparts. Furthermore, the United States’ open system is vulnerable, and a shift away from federal research investment could stall progress. As frontier models become more accessible, the first-mover advantage may not prove decisive. Victory will therefore depend on who can actually use the technology most effectively.
What the USG Should Do:
Keep a close watch: The government must stay ahead of AI developments in both Silicon Valley and China to avoid strategic surprise.
Fuel the engine: Ensure AI labs have what they need—massive computing power, high-quality data, top talent, and sustainable energy (like nuclear power).
Protect the advantage: Enforce strict export controls and bolster security at U.S. research labs and data centers to prevent espionage.
2. The Race for National Security Integration
The goal here is to translate AI breakthroughs into a decisive military and intelligence advantage. This means using AI for everything from processing intelligence and accelerating decision-making to enabling sophisticated autonomous systems. The goal is not to replace humans with machines—but to effectively team humans with machines to achieve a national security edge.
The Problem: Despite nascent public-private partnerships, the United States continues to struggle to bridge the gap between Silicon Valley and the Pentagon. National security agencies don't have early access to the latest AI models, procurement is slow, and the government's operational culture is resistant to change. China, with its state-mandated civil-military fusion, has a structural advantage in rapidly turning commercial tech into military capability.
What the USG Should Do:
Avoid strategic surprise: Maintain close communication with domestic industry leaders, evaluate frontier AI models for misuse risks as well as national security applications, and closely monitor imminent technological breakthroughs abroad.
Build a better partnership: Develop a scalable public-private model where the government and AI companies can co-develop applications for national security, like advanced cybersecurity and biodefense tools.
3. The Race for Economic Integration
This race is about embedding AI across the entire national economy—from manufacturing and logistics to healthcare and finance—to supercharge productivity.
The Problem: U.S. companies excel in software, but they have fallen behind China in physical industries like manufacturing. China leads the world in installing industrial robots and is pioneering "dark factories" that run 24/7 without human workers. Beijing also sees AI-driven automation as a solution to its aging workforce, while in the United States, the prospect of AI causing massive unemployment in the near future is a potential source of major instability.
What the USG Should Do:
Launch an "Industrial AI" initiative: Use tax credits and grants to push AI and robotics into manufacturing, logistics, and energy sectors to close the gap with China.
Support the workforce: Massively invest in retraining and upskilling programs for workers displaced by automation. Update labor laws and strengthen unemployment benefits to manage the economic transition.
4. The Race to Build the Global Tech Stack
This is a competition to control the underlying global infrastructure of AI: the data, the chips, the data centers, and the technical standards. As White House AI Czar David Sacks put it, "If 80 percent of the world uses the American tech stack, that’s winning. If 80 percent uses Chinese tech, that’s losing."
The Problem: The U.S. strategy of letting market forces dictate investment means it focuses on wealthy partner countries, leaving many developing nations in Africa, Asia, and Latin America underserved. China is stepping into this vacuum, offering "good enough" AI systems and subsidized infrastructure. This risks Beijing exporting its model of digital authoritarianism, complete with surveillance, censorship, and propaganda.
What the USG Should Do:
Compete for the Global South: Use public-private partnerships and government-backed financial tools to incentivize buildouts of digital infrastructure and offer access to U.S. cloud computing in key emerging markets.
5. The Race to the Bottom
The final, crucial race is one to avoid—the race to the bottom. The rapid, competitive development of AI increases the likelihood of disaster, either through misuse by rogue states or non-state actors, unintentional military escalation, or a loss of control over a superintelligent system.
The Problem: In the rush to compete, both sides are tempted to cut corners on safety. Experts at leading AI labs warn that their models will likely cross dangerous capability thresholds—enabling malicious actors to design pathogens or launch devastating cyberattacks—this year.
What the USG Should Do:
Prepare for the worst: Run crisis simulations for scenarios involving catastrophic AI misuse or rogue AI to improve readiness.
Pursue narrow cooperation with China: Just as the United States and the Soviet Union created guardrails around the nuclear competition, Washington and Beijing must find a path to collaborate on AI safety even as they engage in an intense AI competition. This could start with sharing information on AI incidents and best practices for safety and control.
Our paper is a reminder that winning the intense AI competition requires more than just innovation. It demands a holistic strategy that adopts AI technology across the national security enterprise, integrates it throughout the economy and society, and responsibly diffuses it around the world—all while managing immense risks. Without it, the United States could end up winning the race to the frontier only to lose the broader contest to shape the global AI future.