AI-Driven Power Concentration: A Comparative Analysis

When technological advantage leads to power concentration

AI-driven power concentration resembles the U.S. nuclear monopoly in 1945, when the United States’ exclusive ability to build atomic weapons enabled it to exert asymmetric political and military pressure. In both cases, a breakthrough technology concentrates power in the hands of the first mover. However, unlike nuclear weapons, advanced AI such as AGI may create self-reinforcing advantages that make that leverage more durable and potentially winner-take-all.

When the United States dropped atomic bombs on Hiroshima and Nagasaki in 1945, it demonstrated an unprecedented concentration of technological power. Nuclear weapons created extreme asymmetry. Japan had no comparable capability and no possibility of retaliation. The Manhattan Project had required massive capital, industrial-scale infrastructure, concentrated scientific expertise, and strict secrecy. These barriers to entry produced a temporary technological monopoly.

Similarly, advanced AI development may concentrate power among a small number of actors. Training frontier AI systems requires massive computational resources, access to advanced chips, vast datasets, specialized talent, and enormous capital investment. In 1945, political leverage came from exclusive possession of a qualitatively new military capability: the ability to destroy entire cities instantaneously. Advanced AI could likewise enable qualitatively new capabilities in areas such as cyber operations, biotechnology, and economic or military planning.

To force Japan’s surrender, the United States showed a credible willingness to use that power, making the asymmetry undeniable. Current U.S. AI policy initiatives and export controls on advanced semiconductors reflect a similar effort to preserve technological asymmetry by restricting access to key inputs, especially advanced chip designs, manufacturing equipment, and the supply chains required for frontier training runs.

Geopolitical asymmetry and competition as a result

The concentration of nuclear power in U.S. hands did not only force Japan’s surrender. It also reshaped global politics. The imbalance alarmed the Soviet Union, which accelerated its own nuclear program to eliminate the asymmetry.

A similar dynamic is visible today in U.S.-China competition over AI, where concerns about asymmetric leverage are driving efforts to secure technological advantage. But unlike the nuclear arms race, where scientific parity was eventually achievable, an AI race could produce more durable dominance. If a sufficiently advanced system can accelerate its own research and development, the leading actor may not simply stay ahead but continuously widen the gap.

That changes the situation from a race to catch up into one where early advantages may compound rapidly. Even if such dominance later proves unstable or self-undermining, the asymmetry could become far more persistent than in the nuclear case. Nuclear weapons did not improve themselves. AI systems already do.

Why AI may be more dangerous than nuclear weapons

Unlike nuclear weapons, where catastrophic risk was largely tied to state deployment, AI systems may pose systemic risks even during development. Competitive pressure could incentivize cutting corners on safety, increasing the chance of unintended consequences.

Just as the Soviet Union’s 1949 test transformed nuclear competition into an arms race and helped define the Cold War, we may already be seeing a similar race-to-the-bottom dynamic in AI development.

AI could make temporary advantages permanent

The U.S. nuclear monopoly shows how technological breakthroughs can suddenly concentrate power and enable asymmetric pressure. But it also shows the limits of such dominance once knowledge diffuses and rivals mobilize.

AI development may begin in a similar way, with concentrated expertise, compute infrastructure, and early-mover advantages producing significant asymmetries. Yet if advanced AI systems enable recursive self-improvement, the dynamic shifts from temporary leverage to self-reinforcing structural dominance.

Whether that happens depends on both technical alignment and sustained political control. Only if increasingly capable systems remain stable under self-modification and responsive to their operators would technological leadership translate into geopolitical leverage.

In that case, technological leadership would not merely confer influence within the existing balance of power. It could reshape the balance itself by accelerating innovation, economic growth, and military capacity at a pace others cannot match.