AI in a great-power fight is a production problem, not a target problem

AI is being absorbed into the great-power playbook as a target problem: hit the data centers, you break the adversary's edge. RAND's wargame of a 2035 fight over Taiwan kept returning to a different answer. The decisive variable, its participants found, was not what gets struck but what gets fielded and lost, and how much of it a side can absorb over weeks of grinding combat.

The mechanism is closer to industrial depth than to cyber strike. Whoever can put more autonomous systems into a theater, replace them faster, and keep them operating through attrition holds the structural edge as a short war stretches into a long one. That is why the wargame's third observation points at protracted conflict: a one-shot raid on compute can be absorbed, while a slow bleed of autonomous mass cannot.

The authors are careful to call this a gameplay-generated hypothesis, not a forecast. The pattern, though, is portable: a capability whose value compounds with volume and endurance rewards the side whose production lines and allied networks can sustain the tempo, and penalizes any doctrine that treats AI as a single brittle node to be destroyed.