In 1999, Warren Buffett shocked his fellow billionaires at the annual Sun Valley conference by making this note about the airline industry:
“As of a couple of years ago”—as in the mid-90s—he said, “there had been zero money made from the aggregate of all stock investments in the airline industry in history.”
That’s right—from the beginning of the airline industry in 1919, to more than 70 years later, in the 1990s—stock investments in the airline industry had, in the aggregate, lost money.
Airplanes are an incredible invention, but airlines were terrible businesses.
Buffett called himself an “Air-coholic,” someone who invests in airlines even though he knows better.
He made this point—notably in 1999, right before the bursting of the dotcom bubble—to warn tech founders and investors that great inventions aren’t always great businesses.
That hype doesn’t necessarily translate into profit.
I think AI and Large Language Models are going to be very, very similar. And we’re going to see a lot of “AI-coholics,” buying up or spending on anything AI that seems promising.
Like airplanes, though, AIs are incredible inventions, but they might be awful businesses.
They’re all competing in a race to the bottom on speed, efficiency, and end-user cost.
They’re actively avoiding making money so they can be first to market. That’s great for customers and short-term VCs, but terrible for long-term investors, employees, and, probably, society.
All of this is to say: Use AI and large language models if you like, but be wary of investing in them or building your own.
I could very well be wrong—airlines and AI are very different technologies, and extraordinarily different businesses. But the incentives are largely the same—scale economies from market domination—which makes speculative over-investment almost a certainty.
So, while directly investing in AI may be tempting, there’s another option: Let VCs and stock market gamblers foot the bill while you create value on top of their investment.
You can buy a plane ticket, not a plane.