Hello, Reader.

In February 1893, America’s railroad empire began to collapse.

Railroad companies that had spent a fortune building the infrastructure found themselves drowning in debt.

The Philadelphia and Reading Railroad collapsed under the weight of its debts in February. Others soon followed. Between 1892 and 1897, railways operating roughly 60,000 miles of track passed into receivership. It was a period of financial distress that hit about one-third of the national network.

Investors lost fortunes, and the failures helped fuel the Panic of 1893, a financial crisis that sent stocks tumbling, banks failing, and the economy into a severe depression. Industrial production fell more than 15% between 1892 and 1894.

Meanwhile, other businessmen were making a fortune on the same railroad tracks.

In 1903, Ladies’ Home Journal became the first American magazine to reach 1 million in circulation.

Its publisher, Cyrus H.K. Curtis, was delivering magazines to households from New York to California. And he managed this remarkable feat without laying a single railroad tie, purchasing a locomotive, or borrowing millions of dollars to build a railroad.

While his publishing empire was growing through the railroad infrastructure, the railroad industry that made it possible was suffering one of the worst financial crises in its history.

Curtis didn’t have to pay for the bridges, tunnels, locomotives, or thousands of miles of track that made his business possible. Other investors had already paid for all that, and many lost fortunes doing it.

That’s one of the great, underappreciated lessons of investing in a technological revolution: The companies that build the infrastructure aren’t usually the ones that make the most money from it.

Sometimes, the biggest winners are the businesses that put that infrastructure to work.

Today, I’ll explain why the railroads went broke, how similar financial pressures are threatening today’s AI giants, and where investors might find the next generation of winners.

I’ll also give you a healthcare stock recommendation that I believe is already benefiting from AI.

How the Railroad Boom Went Off the Rails

Railroad mileage exploded during the 1880s. Investors poured money into new lines, convinced that whoever controlled the tracks would control the movement of goods and collect the profits.

But building railroads was enormously expensive. Companies had to acquire land, lay track, build bridges, and buy locomotives, often with borrowed money. Those debts still had to be repaid whether the trains were full or half-empty.

Meanwhile, the technology kept improving. Steel rails replaced iron ones, allowing trains to carry heavier loads. Westinghouse air brakes made longer trains safer. Better signaling systems let railroads move more traffic over existing tracks.

These innovations were wonderful for the American economy. But they also increased transportation capacity and intensified competition, putting pressure on freight rates.

The railroad companies had borrowed fortunes to build their networks. Then technological improvements helped make the services those networks provided cheaper and cheaper.

That’s a fine arrangement for anyone shipping freight. It’s a considerably less attractive arrangement for the folks trying to pay off the locomotives.

Many railroad companies couldn’t survive. The railroads themselves, however, remained indispensable.

Curtis Publishing was among the businesses that benefited. It could use the rail network to reach customers nationwide without financing the network itself.

The railroad builders spent the money. Curtis got to use what they’d built.

And I’ve seen this pattern before.

In December 2000, I warned investors away from Corning Inc. (GLW), a major supplier of the fiber-optic cable being laid to build the internet. Barron’s picked up my warning, and Corning subsequently plunged 87%.

That same month, I recommended an altogether different kind of Internet investment:

Humana Inc. (HUM). Humana wasn’t laying fiber-optic cable. It was using the internet to make its existing health insurance business more efficient, moving claims and records online and reducing costs. While the internet’s infrastructure builders struggled, Humana eventually gained more than 4,000% following my recommendation.

Corning helped build the internet. Humana found ways to use it. One stock collapsed 87%. The other eventually gained more than 4,000%.

The internet worked beautifully. That didn’t mean every internet investment would.

And now I believe we’re seeing a similar divide emerge in artificial intelligence.

Wall Street May Be Making the Same Bet Again

Microsoft Corp. (MSFT), Amazon.com Inc. (AMZN), Alphabet Inc. (GOOGL), Meta Platforms Inc. (META), and other technology giants are spending hundreds of billions of dollars building AI infrastructure.

Amazon alone is planning roughly $220 billion in capital expenditures this year, much of it tied to AI infrastructure and the computing capacity needed to support it.

There’s a good reason for all this spending. AI is transforming businesses, and demand for computing power remains enormous.

But these companies are making a very expensive bet about what that demand will look like years from now. And I believe they’re beginning to face the same two financial pressures that squeezed the railroads.

First, building infrastructure keeps getting more expensive. Data centers require enormous amounts of electricity, water, land, and equipment. Communities are pushing back against new projects, particularly when residents worry about higher utility bills.

According to real estate firm Cushman & Wakefield, data-center construction costs have risen 21% per megawatt since late 2024.

Second, the technology itself keeps getting cheaper and more efficient. In January 2025, China’s DeepSeek demonstrated that advanced AI models could operate much more efficiently than many investors had assumed. Meanwhile, AI developers are finding ways to deliver more computing power for less money. New models, chips, and software are steadily improving what each dollar of AI spending can accomplish.

Cheaper computing is wonderful for businesses that want to use AI. But it could become a problem for companies that have spent billions building infrastructure on the assumption that computing power will remain scarce and expensive.

What happens if tomorrow’s chips can do substantially more work than today’s? Or if new software dramatically reduces the computing power needed for a task?

The data centers don’t become useless. But their owners may struggle to earn the returns they expected.

Now, there’s an obvious objection to this argument: Cheaper AI could encourage so much additional use that demand for computing power continues to soar. In fact, history offers plenty of examples of falling costs producing explosive growth in demand.

But explosive demand doesn’t guarantee explosive profits. Especially when the companies chasing that demand must continually spend billions to keep up.

And that’s why I’m increasingly interested in companies on the other side of this equation.

The Next Cyrus Curtis May Be in Healthcare

Developing a new drug is notoriously expensive and uncertain. Researchers can spend years identifying promising biological targets, testing thousands of molecules, and deciding which candidates deserve further development.

AI could accelerate that work dramatically. It can already analyze genetic and clinical data, identify drug targets, and guide experiments before scientists enter the laboratory.

And pharmaceutical companies don’t need to become AI infrastructure providers to benefit. They simply need to put the technology to work.

One example is Bristol-Myers Squibb Co. (BMY).

I’ve recommended BMY to my members since February 2024, when I first advised readers to buy shares at $56 or below. It hasn’t exactly been a Wall Street favorite. Investors have worried about the company’s earnings outlook and future drug pipeline.

But Bristol Myers is investing heavily in AI-assisted drug discovery. In July, it announced plans to become the first life-sciences company to purchase a supercomputer powered by Nvidia Corp.’s (NVDA) latest Vera Rubin technology. The company intends to use that computing power to design medicines, identify drug targets, and guide experiments before laboratory work begins.

Here’s the distinction. Bristol Myers may be buying sophisticated AI equipment. But its business isn’t selling computing power. Its business is developing and selling medicines.

I call companies like this AI Appliers.

If AI helps Bristol Myers discover promising drugs faster, reduce failed experiments, and improve research productivity, those benefits could eventually show up in its earnings.

Meanwhile, the stock remains relatively inexpensive. In late September, BMY traded at less than 12 times forward earnings. I continue to recommend BMY to my members up to $70 per share.

Of course, AI won’t magically transform every experimental drug into a blockbuster. Drug development remains an expensive, failure-prone business, and Bristol Myers still faces challenges in its existing portfolio.

But that’s precisely why the stock interests me. Wall Street isn’t pricing BMY as an AI superstar. It’s pricing the company as a pharmaceutical business with problems to solve. If AI helps solve even a few of them, investors could be pleasantly surprised.

And BMY is hardly alone. Across healthcare, manufacturing, energy, and other industries, companies are discovering ways to use AI to improve their businesses. Many of the most promising are still being overlooked by Wall Street.

That’s where I’m spending more and more of my research time.

Who Gets Rich After the Boom

The railroad bust of 1893 didn’t end the railroad age. Nor did the dot-com crash end the internet revolution.

In both cases, the technology survived. And investors who understood how businesses could profit from it found extraordinary opportunities.

I expect AI to follow a similar path.

Some of today’s AI infrastructure giants may become spectacular long-term investments. Others may discover that building the future is a very expensive way to make a disappointing return.

Meanwhile, thousands of companies will have the opportunity to use that infrastructure without paying to build it. That’s the part of the AI Revolution that interests me most.

Bristol Myers is one example. I’ve identified several others that I believe offer even greater upside potential.

In fact, I’ve just recorded a free special presentation explaining why I believe November 4 could mark the beginning of a major shift in the AI market — and which companies I expect to benefit.

During the presentation, I reveal my No. 1 AI Applier, including its name and ticker symbol. You don’t need to buy anything to get that recommendation. Click here to watch my new presentation and get my No. 1 AI stock pick for free.

Cyrus Curtis didn’t need to own a railroad to build a publishing empire. He just needed the trains to run.

Regards,

Eric Fry

Senior Global Macro Analyst, InvestorPlace