LLMs are becoming commodities — and that changes everything

Ben Evans makes a clean case that frontier LLMs are heading toward commodity pricing. Models are getting cheaper by the month — the kind of deflation we usually see in hardware or shipping. The gap between the top tier and the rest is closing fast. What used to cost a dollar a million tokens is now a fraction of that. The big players are racing to lower prices, and the pressure is real.

The implication isn't just that AI is cheaper. It's that the moat is moving. When the base model is a commodity, the advantage shifts to what you build on top — tools, workflows, data, distribution. The companies that only sell raw model access are in trouble. The ones that bundle it into something people actually use are the ones that'll stick around.

For la comunidad, this is the setup for a wave of cheap AI tools. Not the flashy ones with demos, but the boring ones — for small businesses, for side hustles, for the cousin who runs the bodega and needs inventory done without hiring someone. The models are free enough now that the real question is what the tools are and whether they solve a real problem.

Why this matters for us: cheap models mean the next wave of useful tools will be built for working people, not just VCs — and the ones that actually help us win are the ones we'll grab first.