When a Great Technology Makes a Poor Investment

What railroads, airlines, and the internet can teach us about investing in artificial intelligence.

Throughout history, some of the most important technological advances have produced surprisingly poor returns for the investors who financed them.

Railroads transformed transportation and commerce. Airlines made international travel accessible to ordinary people. Fiber-optic networks helped create the modern internet. Each changed the world in ways that would have been difficult to imagine when the technology first appeared.

Yet the history of these industries is filled with bankruptcies, disappointing profits, and substantial losses for shareholders.

This apparent contradiction illustrates one of the most important distinctions in investing: A wonderful technology does not necessarily make a wonderful business, and even a wonderful business can be a poor investment at the wrong price.

The enormous investments being made in artificial intelligence provide a timely reason to revisit this lesson.

The Railroad Boom

During the 1840s, Britain experienced a remarkable expansion of its railway network. Early railroads had demonstrated their ability to move people and goods faster, more reliably, and at lower cost than existing alternatives. Many were profitable, and investors understandably became enthusiastic about the industry's prospects.

Capital poured into railroad construction. New lines were proposed, existing companies expanded aggressively, and investors anticipated substantial profits from the growing transportation network.

The railroads delivered on their technological promise. They helped transform Britain's economy and remained essential infrastructure for generations.

The financial results were another matter.

As railway companies expanded, they increasingly competed for the same passengers and freight. Construction costs were substantial, and many additional routes failed to generate sufficient revenue to justify the investment. Industry profitability deteriorated, dividends declined, and shareholders suffered considerable losses.

Interestingly, much of the destructive expansion came not from inexperienced promoters but from established, previously successful railway companies.

Their mistake was not believing in the future of rail transportation. That future was remarkably bright. Their mistake was investing more capital than the resulting profits could justify.

The Fiber-Optic Boom

A similar pattern emerged during the internet's early expansion.

Telecommunications companies correctly anticipated an enormous increase in demand for data transmission. Fiber-optic technology offered extraordinary improvements in communications capacity, and companies committed vast sums to constructing networks.

The underlying prediction proved correct. Internet traffic expanded spectacularly, and fiber-optic networks became indispensable to modern commerce and communication.

But the investment economics proved far less attractive.

Companies constructed competing networks while advances in optical technology continually increased the amount of data that existing infrastructure could carry. As capacity expanded, the price of transmitting data declined.

Numerous telecommunications companies ultimately failed or underwent financial restructuring. Much of the infrastructure they built continued serving society, often under new ownership.

The internet became an even greater success than many early investors imagined. Unfortunately, that success did not necessarily translate into adequate returns for those who financed its infrastructure.

Airlines: An Enduring Example

Commercial aviation offers another instructive example, particularly because the industry's financial difficulties have persisted despite decades of technological improvement.

Modern aircraft are vastly more capable and efficient than their predecessors. Passenger traffic has expanded enormously, and air transportation has become an essential part of the global economy.

Nevertheless, airlines have historically struggled to earn satisfactory returns on the capital invested in their businesses.

Aircraft are expensive, operating costs are substantial, and competition frequently limits pricing power. An airline must continually invest in equipment while competing against other carriers offering essentially the same transportation service.

Passengers have benefited tremendously from this competition. Shareholders, collectively, have often fared considerably worse.

The lesson is not that every airline is a poor investment. Rather, it is that strong demand, technological progress, and even growing revenues are insufficient to guarantee attractive profitability.

The Parallels to Artificial Intelligence

Artificial intelligence may ultimately prove as transformative as railroads, electricity, or the internet. Its potential applications in scientific research, engineering, medicine, manufacturing, and countless other fields are extraordinary.

Nevertheless, its investment economics remain unusually difficult to predict.

The first concern is competition.

Large technology companies are committing enormous amounts of capital to computing infrastructure, semiconductor development, and AI models. Each has powerful incentives to establish a leading position or defend its existing business against competitors.

Individually, these investments may appear entirely rational. Collectively, however, they could produce considerably more capacity than the market can profitably support.

The railroad and fiber-optic booms demonstrate how easily this can happen.

The second concern is the pace of technological change.

Building an AI data center requires substantial investment in specialized computing equipment. Unlike railroad tracks or fiber-optic cables, however, advanced computer chips may become economically obsolete relatively quickly.

As newer equipment becomes more capable and efficient, businesses may need to replace existing hardware simply to remain competitive.

This creates an important distinction between growth capital expenditure and maintenance capital expenditure.

Growth expenditure is intended to increase a company's future earning power. Maintenance expenditure is the money required to preserve the earning power it already possesses.

An investment in new computing equipment may initially appear to be growth expenditure. But if the company must repeatedly replace that equipment to maintain its competitive position, what once appeared to be an investment in growth can effectively become a recurring cost of doing business.

For shareholders, that distinction matters enormously.

A business can report impressive revenue growth and accounting profits while continually consuming much of its cash flow in additional equipment.

The third concern is where the economic benefits ultimately accrue.

Artificial intelligence may dramatically reduce the cost of performing many tasks. But lower costs do not necessarily translate into higher profits for AI providers.

Competition may force providers to pass much of those savings along to customers. As happened with telecommunications, rapidly improving technology could make the service increasingly valuable to society while simultaneously making it less profitable to supply.

Indeed, some of the greatest beneficiaries of artificial intelligence may be established businesses that use it to improve productivity without having to finance the underlying infrastructure themselves.

What This Means for Investors

None of this establishes that artificial intelligence is a speculative bubble or that every company participating in its development will prove to be a poor investment.

Some businesses may emerge with durable competitive advantages and exceptional profitability. Others may benefit substantially from incorporating AI into their existing operations.

The difficulty lies in determining which companies will succeed, how much capital they will ultimately require, what their normalized earning power will be, and how much of that future success is already reflected in their share prices.

These are separate questions from whether artificial intelligence itself will succeed.

My investment approach is built around owning understandable businesses with durable competitive advantages, sound financial positions, and the ability to generate substantial cash for their owners. Equally important is purchasing those businesses at prices that provide a reasonable margin of safety.

When an industry's competitive position, future capital requirements, and long-term economics are exceptionally difficult to forecast, estimating a business's intrinsic value becomes correspondingly difficult.

That does not mean the opportunity is necessarily unattractive. It means I may be unable to determine whether it is attractive with sufficient confidence.

There is no requirement that an investor participate in every technological revolution.

The objective is not to identify the industries that will change the world. It is to acquire ownership interests in businesses that can generate satisfactory returns on the capital entrusted to them, at prices that make those returns worthwhile.

History repeatedly demonstrates that these are very different things.