We are in the construction phase of an economic transition, when capital often moves ahead of the revenue streams that will eventually test and support it. This has happened before, and the clearest precedent is the one most people remember wrongly. A cold glass of beer and a mail-order catalog explain it better than any forecast.
Before the Tracks
For most of history, production and consumption were local because distance was expensive. Beer brewed in Milwaukee was consumed near Milwaukee. A store sold to the people who could reach it. The size of a business was constrained by the cost of moving its goods, and that cost was high enough that most markets stayed small. Railroads changing the price of distance and with it the structure of the economy is the story, but how it changed, and who captured the value when it did, is where the lesson lives.
The Tracks
The railroad era is often remembered as a bubble, and it may have looked like one. The country built far more track than it had traffic, lines to nowhere, two railroads laid side by side to serve a single town. Capital flooded in. Stock was watered and books were cooked. And when the financing finally outran the freight, the system broke, and by 1893 roughly a quarter of American rail mileage sat in receivership. Yet the track remained. Steel does not evaporate when euphoria does. The lines were reorganized under new owners and kept running, and the network they formed went on to transform the economy.
That gap, between the companies that failed and the network that endured, is the part worth carrying forward. The value such a transition creates and the value an investor can capture from it are rarely the same thing, and rarely held by the same people. It's helpful to separate layers.
There were builders who sold steel, locomotives, equipment, the "picks and shovels" of the railroad boom. Andrew Carnegie† made his fortune supplying steel rails. George Westinghouse made another supplying the air brakes on every train that ran across them. There were operators, the railroad companies themselves, names that everyone knew, like Union Pacific, Northern Pacific, New York Central. And after the correction came consolidators. Businessmen like J.P. Morgan and E.H. Harriman reorganized bankrupt railroads in the 1890s into disciplined systems, capturing durable value precisely because they bought when others were forced to sell.
Steel does not evaporate when euphoria does.
Beers and Sears
Then the network matured, and some of the largest winners were not railroad companies. Not builders of track, not the operators of rail lines, and not necessarily the consolidators. There was a fourth category few had been tracking. They were companies that created business models that only existed because the network did, and those that benefited from the emerging economic structure.
Refrigerated cars and pasteurization turned beer from a local product into a national one. The great shipping breweries of Milwaukee or St. Louis could suddenly sell a thousand miles from the brewhouse. Sears built an empire on the same substrate, launching a catalog that reached rural homes with no store nearby. The rails carried it across the country. After 1896 the post office began delivering to rural doorsteps, and the Sears, Roebuck & Co. Catalog traveled the last mile into the home. It was a business model unthinkable without the tracks. Standard Oil is the sharper version of the same story. Rockefeller owned no railroads, but overbuilt lines competing for freight handed their biggest shipper the power to set his own terms, and the era's excess capacity became a subsidy to his business. Many long-term beneficiaries didn't lay a foot of track or produce an ounce of steel.
Beer and Sears are not a story about lucky downstream winners. They are a story about what the network did. It separated production from consumption and changed the economics of transport. You no longer had to brew where you drank or hold inventory where you sold. The network made both optional and unlocked economic activity that couldn't have existed previously.
The charm of the story can hide something harder. There were investors who lost out, and not because they were fools or latecomers. They were watching the most important, most-discussed, most-obviously-consequential companies of the age. Fixating solely on the "transformative" companies was the mistake, not because the transformation wasn't real, but because the transformative company and the rewarding investment are not always the same thing at the same time.
Many investors believe they are buying the future. They may only be buying one participant in it.
The AI Build-Out
The modern parallel is close enough to be uncomfortable, and yet the roles are not equally visible or as clearly defined.
There are builders supplying the new infrastructure, lines to transmit and process data in the form of semiconductors, networking capacity, power, cooling, data-center equipment. And there are operators, cloud providers, foundation models, and Mega-AI platforms. The other two layers are, so far, positions rather than participants. Consolidation is what a shakeout creates, and so far AI hasn't seen its great consolidators. There have been no J.P. Morgans yet. The beneficiaries, the future Sears, Anheuser-Busch, and Standard Oil, are the businesses that become possible and profitable at scale because abundant computation exists.
The names of that last group are evolving, and many of the models of business that can truly change the system have not yet taken their final form. That can sound like a reason to wait, but waiting mistakes a hidden name for a hidden shape, and only one of them is as obscured as it seems. Sears was unknowable in 1870, but the conditions that would produce a Sears were already in plain view, a network severing production from consumption, and a rural population it could suddenly reach. The role shows up in advance as a position downstream of the network, made possible by it, identifiable long before anyone can say who will fill it.
A similar decoupling is already here. Companies no longer build their own server farms. Computation is no longer only on premises. Machine intelligence is produced at hyperscale data centers and delivered over the network. That is the same move the railroad made, production separated from place, and it points to the same conclusion. The railroad was not only a product, it was a new economic substrate.
Which leaves a single question for anyone deploying capital into Artificial Intelligence. Which role are you actually buying? Each is priced differently, rewarded at a different stage, and exposed to a different kind of risk. Many investors believe they are buying the future. They may only be buying one participant in it.
The network had to be built before anyone could reorganize around it. That reorganization, and why its payoff often arrives much later than the capability does, is the next part of the story.