Everyone wants to win the AI race and the logic is hard to argue with. If increasingly capable AI is going to reshape economies, military power and scientific research, getting there first could mean getting the advantage first. Dario Amodei has argued exactly this in the context of US-China competition: falling behind could hand a rival a technological advantage that becomes difficult to recover from. So the obvious question is, why slow down?
The Prize Is Getting Strange
Here is where the logic starts to wobble. The same industry racing to build more capable systems is increasingly having to think about how those systems will be controlled. The 2026 International AI Safety Report notes major advances in AI reasoning and autonomous operation, while also finding that current risk-management techniques still have significant gaps. It reflects the possibility that AI could eventually operate beyond human control. With $341 billion invested in AI infrastructure in 2025, the stakes are rising, making falling behind seem more dangerous than slowing down.
Faster Because Everyone Else Is
Eric Schmidt and his co-authors have warned that treating advanced AI as a high-stakes, Manhattan Project-style competition could create incentives for rivals to accelerate and respond to one another, potentially undermining the stability the strategy is supposed to create.
And that’s where the race becomes self-reinforcing: if everyone thinks someone else might get there first, everyone has a reason to keep moving, without knowing where the finish line leads.
What Exactly Are We Winning?
The answer is not to pretend competition does not matter. The problem is assuming that first automatically means safest, strongest or most in control. A race makes sense when the finish line represents a stable prize. But with frontier AI, the finish line keeps moving because the technology itself changes the conditions of the race. Maybe the real mistake is not racing. Maybe it is confusing momentum with strategy. Because if the ultimate objective is to build something humans can control, then reaching the frontier first is only a victory if that control survives the finish line. Otherwise, everyone is sprinting toward a prize that disappears the moment someone reaches it.
