Is AI giving you dot-com déjà vu? Here’s what’s different about these investments. | Expert Opinion

Is it déjà vu all over again? The massive flow of capital into the AI sector looks a lot like the incredible level of investments in dot-com companies during the 1990s.

That raises the question: When the arms race for AI supremacy settles down and the inevitable shakeout occurs, will earnings be significant enough to justify those investments, or will firms crash and burn again?

In the second half of the 1990s, the only thing investors could think of was how to make the next million off the next dot-com stock. I used to “joke” that investors assumed if they bought a stock in the morning, by the afternoon they could buy their dream foreign sports car, or if they waited a day, they could buy a new house.

Unfortunately, most dot-com stocks turned out to be “rice cakes”: Something that appeared to have form and substance until you bit into them and discovered there was nothing there.

When the bubble burst, there was widespread pain.

Since it took relatively modest capital to create browser software, small companies could get into the competition.

Ultimately, there was one big winner: Google Chrome, which now has nearly 70% of the global search market. In second place, Apple Safari, which lags well behind at 15.5%.

In contrast, AI software developers require immense investments in data centers filled with ever-increasing computer capacity needed to run the software and continue the “learning process.”

Consequently, returns to the tech investment in AI have to cover not only the cost of the software development, but also the build-out of the data centers to keep the software at the cutting edge.

On top of that, they need lots of capital to outbid competitors for the consumer-oriented products that will inevitably be created by independent software developers.

Put simply, we are not talking about a world where Stanford University students Sergey Brin and Larry Page could receive a $100,000 influx of capital and develop Google in a garage.

Only the wealthiest tech companies have the financial wherewithal to compete in this market.

If the AI sector consolidates similar to the web search sector, the returns to the “Google of AI” will be well worth the expenditures.

But for the others, much of their spending could yield little.

If the market is fractured, the profitability outlook for almost every top-end investor is still uncertain.

With a variety of survivors and a relatively even distribution of the market share, intense competition is likely. Margins would be constantly under pressure.

In addition, as we have seen with the software industry over the past 25 years, independent entrepreneurs are capable of creating products that readily compete in smaller ways with the big firms. That puts additional pressure on profits.

Unless there are only very few winners, profitability in the AI sector is not a given.

Regardless, there are likely to be a significant number of firms whose huge bets go bust.

The latest data on the economy’s economic performance, the GDP report, was released recently. While activity moderated in the fourth quarter of 2025, the 2.2% increase for the year was in line with what economists call trend growth.

What was eye-opening in the report was how much AI likely added to overall growth.

AI spending affects a number of different segments of the economy. Given the massive demand for data centers, their construction is one obvious area.

But that isn’t the major place where we see AI powering growth. It is what goes into those buildings that really makes a difference. Purchases of information processing equipment added the most to growth.

Software and research and development are two other sectors where spending is booming.

When you add those four components together, AI may have created about 40% of the GDP growth in 2025.

But there is a caveat: Not all spending is created equal.

Nvidia, the chief supplier of computer chips to AI firms, manufactures most of its chips in South Korea. In GDP accounting, those chips are imports, which reduce growth.

To the extent that computer products used in the data centers are imported, the spending doesn’t directly affect United States demand, making it somewhat unclear how much the sector truly powers growth.

Unless U.S. chip and computer hardware manufacturing expands dramatically, the positive impacts from the sector may be disappointing.

As with any industrial build-out, the race to build new data centers will ultimately slow. When that river of cash flowing into developers and local governments peaks is unclear, but I wouldn’t be surprised if it is in the next couple of years.

As for the stock markets, in 2025, the so-called Magnificent Seven (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla) contributed roughly 40% of the S&P 500’s total return.

To maintain their robust valuations, the AI companies have to start showing soon that the returns from those investments are real.

AI is rapidly becoming embedded in all sectors of the economy. And with that, the risks and opportunities are becoming concerning.

What the AI economy will finally look like, who it will benefit or hurt the most, and what ups and downs the economy will endure remain open questions.

But there is no turning back.

Full disclosure: AI developer Anthropic was sued for training its AI product on copyrighted materials without permission. A book Joel Naroff coauthored, “Big Picture Economics,” was part of the suit, and he is part of the settlement.