Value Created vs Value Captured

The analogy

Eric Glyman uses air conditioning to defuse a common paralysis of the AI era: that the sheer scale of AI's potential can blind people into wondering whether it is worth building anything else, or whether to build a company at all if a much more general intelligence is right around the corner.¹ He encounters engineers at both extremes, some genuinely afraid of catastrophe, others fixated only on the frontier labs.

His analogy: it's 95 degrees outside, and nobody is sitting around discussing the great robber-baron families of the air conditioning industry. There are no legendary air-conditioning fortunes to speak of. And yet air conditioning made Las Vegas possible, his own birthplace, since that city would not exist without it, and it made Miami possible, along with entire ecosystems built on top of climate control. "The value created out of the ecosystems far exceeded the value captured by the people who invented this technology."

Applied to AI

Applied to AI, Glyman agrees it is one of the most revolutionary transformations of a lifetime, on a scale beyond the industrial revolution, since where that revolution made goods plentiful, this one makes services plentiful, and he agrees some of the frontier labs will become among the best businesses ever built. But he argues the more interesting question is different: in a world where intelligence is plentiful, accessible, and cheap, what should get built, and what kinds of businesses are now possible that were not possible before. Just as entire cities became viable only after the air conditioner existed, whole categories of business become viable only once cheap intelligence exists underneath them. He grounds this in a specific observation about payments: a general intelligence "likes to pay for things and hires other intelligences to do work," so cheap intelligence produces an explosion in the volume of payments and a genuine need for the substrate that manages them, which is the business he is building at Ramp.

Why it matters

This functions as a strategy heuristic for navigating the AI transition, and as an antidote to tunnel vision fixated only on the labs themselves. It argues that the durable opportunity is frequently one layer removed from the general purpose technology itself, not the inventor of the intelligence, but the businesses and even the cities that intelligence makes viable. It pairs naturally with the observation that cheap intelligence expands total demand rather than shrinking the market for intelligence-adjacent work: demand grows on one side while most of the created value shows up in newly possible downstream businesses on the other, rather than in the commoditizing input itself. This is also why Glyman can describe the frontier labs as his real long-run competitor and still remain energized about building a fintech company: the ecosystem value sitting atop cheap intelligence is the actual prize, and being positioned inside that ecosystem, in the flow of money and payments specifically, is how a company captures a piece of it.

A cap-table complement

A separate ranking of founders by wealth created for shareholders other than themselves, built around a methodology Jeff Bezos proposed, offers a narrower, quantified version of the adjacent idea.² That ranking measures the shareholder residual, market capitalization minus a founder's own stake minus the outside capital raised to build the company, and by its own admission still leaves out the wider ecosystem and spillover value this concept describes. The two ideas nest inside each other: the value created for an entire economy or ecosystem is larger than the value captured by outside shareholders, which is in turn larger than the slice any individual founder personally keeps.

Tensions and open questions

The analogy functions as much as a comfort as an analysis. Some general purpose technology providers, railroads and oil majors historically, and arguably today's hyperscalers, did capture enormous value rather than merely enabling it for others. Whether the frontier AI labs will turn out to be air conditioners, in the sense of low value capture relative to the value they unlock, or railroads, in the sense of high value capture, remains genuinely unresolved, and Glyman himself concedes that some labs will likely be among the best businesses ever built. The claim about what new businesses become possible is also directionally right but underspecified: naming the opportunity space is not the same as identifying which specific downstream businesses will actually capture durable value, and ecosystem value exceeding captured value, while historically common, is not guaranteed, since a provider that controls a genuine bottleneck can invert the pattern entirely.