28 June 2026 · LinkedIn
AI is generating real value that almost no one keeps. It's turning into lower prices rather than profit, which is why the loudest question, whether this is a bubble, keeps missing it.
Start with the money going in. People spent around $110bn on AI over the past year, growing about three times faster than internet or mobile spending did at the same age. Yet the companies selling it are barely making money: on a bottom-up reconstruction of the industry, the AI revenue at the big cloud providers only just covers the depreciation on the chips it runs on.
The buyers don't obviously keep it either. What AI has delivered them so far is mostly efficiency, the same work done by fewer people or in less time. So where does that saving go? What follows is a supposition rather than a measurement, but a logical one.
It can't easily stay as profit, because the capability is the same one every competitor can rent. Everyone has to adopt just to keep pace, a competitive arms race some economists have begun to model, and the first to use it to cut prices forces the rest to match or lose the work. The saving ends up with the customer rather than on anyone's books: a law firm trims its fees, its rivals follow, and the client pockets the difference. Where a firm holds something rivals can't copy it keeps more of the gain, but where it doesn't, which is most places, the saving competes away.
If that is what is happening, it would explain a pattern the figures keep showing: for all the money going in, the profit spike never quite arrives. Value that turns into lower prices would never show up as profit, or in any tally of what people pay for AI. It would spread through the economy as cheaper services that no revenue line records.
So the useful question isn't whether the AI economy is a bubble. It's whether that deflation is reaching you as your own efficiency yet, or as a competitor's lower price.
Join the conversation on LinkedIn →
← All writing