The Smarter AI Gets, the Harder It Is to Make Money
Everyone believes AI will create enormous value.
I agree.
What I disagree with is who gets to keep it.
The smarter AI becomes, the harder it is for AI companies to make money.
Every AI company is chasing the same goal: build the smartest model in the world. Billions of dollars are spent on larger clusters, better researchers, and more capable models. The entire industry assumes that if intelligence keeps improving, profits will naturally follow.
I don't think that's true.
People don't buy intelligence.
They buy outcomes.
Most users aren't proving mathematical theorems or solving frontier research problems. They're writing emails, generating code, summarizing documents, and asking everyday questions. Today's leading models are already remarkably good at these tasks.
As models become more capable, the value of becoming slightly more capable begins to disappear.
A model that is 3% smarter rarely creates 3% more value for the average user. Once every major model is "good enough," benchmark leadership stops mattering. Consumers stop asking "Which model is the smartest?" They start asking "Which one is good enough?"
That is how premium products become commodities.
Nobody asks which power plant generated their electricity. Nobody asks which CDN delivered a webpage. Infrastructure succeeds by becoming invisible. AI is following the same path. The smarter models become, the less people care which model produced the answer.
Unfortunately, becoming a commodity is terrible for business.
Customers become more price-sensitive just as providers become more cost-intensive. Training larger models requires more compute, more capital, and more researchers. At exactly the moment products become harder to differentiate, they also become more expensive to build.
The competition is also unusual.
Traditional industries compete against companies that need to earn a profit. AI companies compete against researchers, governments, open-source communities, and technology giants willing to subsidize models for entirely different strategic reasons. Some release frontier models to advance research. Others do it to strengthen ecosystems. Either way, they put constant downward pressure on prices.
The result is a strange paradox.
Every breakthrough quickly becomes everyone else's baseline. Intelligence keeps improving, but differentiation keeps shrinking. Society becomes richer because AI becomes better. Model providers don't necessarily become richer with it.
History has seen this movie before.
Industries that transform the world don't always produce extraordinary businesses. Sometimes competition captures the gains before shareholders do.
AI may become one of humanity's most important technologies.
That doesn't mean AI model providers will become humanity's most profitable companies.