How War Slows Down AI Progress

Artificial intelligence is often framed as a race of algorithms and breakthroughs. In reality, its progress depends on something far more fragile: a small set of critical resources—advanced chips, stable infrastructure, and highly skilled talent.

When conflicts arise, these foundations become targets. Modern warfare doesn’t need to attack “AI” directly to slow it down. It only needs to disrupt what AI depends on.

The Real Bottlenecks Behind AI

AI development relies on three key pillars:

• Compute: advanced chips and data centers
• Infrastructure: electricity, networks, and physical facilities
• Talent: researchers, engineers, and institutions

If any of these are weakened, progress slows. If several are hit at once, progress can stall entirely.

How Conflict Disrupts AI

Across recent and historical conflicts, a consistent pattern emerges. The most effective disruptions target bottlenecks—not AI itself.

1. Attacking Infrastructure

Strikes on power grids, telecom networks, or data centers can immediately halt AI training and deployment. While systems can sometimes recover quickly, repeated disruptions create instability that pushes companies and researchers to relocate.

2. Cutting Off Chips

Advanced AI systems depend on specialized hardware. Export controls and supply chain disruptions can block access to these chips, limiting how powerful systems can become. Unlike infrastructure, this is not easily fixed—replacement can take years.

3. Driving Away Talent

The most lasting damage comes from people leaving. War forces researchers to emigrate, switch careers, or stop working entirely. Rebuilding that expertise takes far longer than rebuilding buildings.

4. Cyber and Information Warfare

Cyberattacks can disrupt communications and destroy data, while misinformation campaigns degrade the quality of information systems rely on. These effects are harder to measure but can quietly erode progress.

What Matters Most: Talent and Compute

Not all disruptions are equal. Infrastructure can often be repaired. Cyberattacks are usually temporary. But two factors create lasting setbacks:

• Loss of talent can take decades to recover
• Restricted compute can cap how far AI systems can scale

When both are affected, progress slows dramatically.

A New Kind of Technological Competition

Modern geopolitical competition increasingly focuses on controlling access to advanced technology. Export controls on AI chips, for example, are explicitly designed to limit rivals’ ability to build cutting-edge systems.

This marks a shift: instead of competing only through innovation, states are also competing by constraining each other’s capacity to innovate.

Why This Matters

AI is deeply embedded in civilian systems—healthcare, infrastructure, communication, and research. That makes its underlying resources both strategic and vulnerable.

The result is a difficult balance. Actions that slow an adversary’s technological progress can also disrupt global research and civilian life.

The Bottom Line

AI progress doesn’t just depend on better models—it depends on stable systems.

War reveals how fragile those systems are. When compute, infrastructure, and talent are disrupted, progress doesn’t just pause—it can move backwards.

Understanding this is key to the future of AI. Not just how to advance it, but how to protect it.