Why Meta Could Benefit Even If AI Models Become Commoditized

Summary

The article argues that Meta’s AI strategy could work even if it does not consistently lead the model frontier. Its return would not have to come from selling access to models: better ad targeting and improvements to Meta’s apps could generate value across the company. The argument rests on a distinction between AI’s usefulness and model developers’ pricing power. A model can remain useful after competitors reproduce its capabilities, while the premium customers pay for direct access approaches zero. In that scenario, cheaper models could hurt model providers but improve the economics of companies that use them. Meta’s potential distribution of 3.6 billion users makes a model that is sufficiently capable more valuable to the company, even if it is not the best available for every task. Owning a competitive model could also reduce Meta’s dependence on platform owners and their permissions, a concern reinforced by its experience with Apple. The article rejects the stronger claim that commoditization would send OpenAI and Anthropic to zero, noting that reliability, integration, and preferred products can retain value even when underlying models become interchangeable. It concludes that Meta’s weaker model position alone does not prove failure, while its large user base alone does not prove that the spending will pay off; the unresolved question is whether advertising, product gains, and independence justify the investment.