If AI leadership depends mainly on building the strongest general-purpose model, advantage will accrue to actors able to train at the frontier. The decisive inputs are therefore access to vast computing capacity, elite research talent and the capital required to finance increasingly expensive training runs. These assets determine who can push the underlying model’s capabilities furthest.
But if general-purpose models become sufficiently capable and widely accessible, value is likely to migrate from model creation to industrial adaptation. Success will then depend less on marginal improvements in benchmark performance than on embedding AI into factories, energy systems, logistics networks and specialised equipment. This requires proprietary industrial data, domain expertise, systems integration, physical infrastructure, supply-chain control and the capacity to manufacture and deploy AI-enabled products at scale.
