Warren Buffett Has a Lesson for AI Investors: Don't Ignore History.

In 1989, Warren Buffett put $358 million into the preferred stock of an airline called USAir. By the end of 1994, Berkshire Hathaway had written that stake down to $89.5 million, or 25 cents on the dollar.

Not the best move by one of the greatest investors of all time. Indeed, Buffett later said that it was "to [his] shame" that he had "participated in this foolishness" by purchasing the airline's shares.

OK, that's great, but what does that have to do with artificial intelligence (AI)? I'll tell you.

What Buffett's airline disaster teaches us

The debate at the heart of the AI trade, the one largely splitting bulls and bears, is whether AI is truly a revolutionary technology. Will AI fundamentally reshape our world and continue to drive the S&P 500 (SNPINDEX: ^GSPC) higher?

While this is certainly an interesting question, I want to propose something that I think a lot of investors miss: The answer can be "yes," and investors can still lose money.

This is where the airline industry becomes instructive. How many technologies have been more revolutionary than flight? I would argue very few. And yet, as Buffett put it in his 2007 letter to shareholders: "[I]f a farsighted capitalist had been present at Kitty Hawk, he would have done his successors a huge favor by shooting Orville down."

By Buffett's accounting, the industry as a whole had lost money on a net basis over its entire existence, going all the way back to the Wright brothers, the first to take flight. Many of the biggest carriers have either entirely folded -- Pan Am -- or been bought out in a moment of deep financial distress -- Continental.

The point is, flying obviously changed the world. It just didn't make much money for the people who financed it. Why?

Two primary reasons: One, it is an extremely expensive business, and two, it is more or less a commodity. A ticket from San Francisco to New York is a ticket from San Francisco to New York. While the experience may differ at the margins, the core product -- getting you from point A to point B in the air -- is exactly the same.

Why AI might be the new airline industry

Admittedly, I cannot say for certain this will hold true of AI, but the two hallmarks of what Buffett argued made a business "gruesome" look like they apply. First, the scale of investment in AI dwarfs anything we've seen before. And second, there doesn't seem to be a significant difference in model quality between the frontier labs over time.

Yes, one lab may come out with a new model that takes a leap forward, but it's closely followed by a competitor's latest offering, which is then matched -- often just a few months later -- by an open source model available for free.

Image source: Getty Images.

Now, my argument is obviously targeted at the model makers like OpenAI and Anthropic. The companies selling hardware or services to them, like Nvidia, are obviously making money hand over fist. But these companies are reliant on the model builders. If they continue to be "bottomless pits" for capital, they will fail. And if they fail, the whole market contracts.

What this means for your portfolio

So what should investors take away here? I think it's not about getting carried away by flashy narratives and instead focusing on the boring financial facts of a business, not just the promise of its technology. So while AI may well turn out to be as important as its boosters say, that doesn't mean it will make for a good investment.

I think this is a great time to evaluate your portfolio and make sure most of it is made up of companies that make real cash and aren't entirely dependent on customers who might not be solvent in a year or two.

Johnny Rice has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Berkshire Hathaway and Nvidia. The Motley Fool has a disclosure policy.