AI Boom Reminds Henry Blodget of Dot-Com Boom

In the late 1990s, everyone was convinced that the internet was the future — and that by buying the stock of internet companies like the theGlobe.com you could get really rich.

They were half-right: You’re reading this on the internet. But all that remains of theGlobe is a mothballed web domain.

We could see the same thing this time around with AI, says Henry Blodget. And he says that wouldn’t be a bad thing.

Blodget has a very particular perspective on booms and busts, since he was a central figure in the rise and fall of the first internet bubble. On the way up, he was a celebrated Wall Street analyst; afterward, the Securities and Exchange Commission charged him with fraud — a complaint he settled without admitting or denying guilt.

After the dot-com era, Blodget went on to found the publication you’re reading now (I was his first hire), and is now putting out his own publication and podcast.

Blodget’s bubble argument goes like this: We see them all the time, and they can often be a good thing, by focusing money and attention on important things. But they will also inevitably deflate — so the trick is figuring out how to play them. Which, for most people, involves not playing at all.

Blodget and I talked about the AI boom and his new media aspirations in the latest episode of my Channels podcast. What follows is an edited excerpt of our conversation.

Peter Kafka: You got a very close look at the internet bubble of the 1990s. How does it compare to what we’re seeing now?

Henry Blodget: This bubble phenomenon is the way markets and technology innovate.

It’s often portrayed as a morality play: “It’s gonna end! It’s so stupid. All this money is gonna be burned, and people are losing their minds.”

What’s really going on is it’s a huge R&D lab. Hundreds of billions of dollars, probably trillions at this point, are being thrown at something that most of us recognize as a huge future opportunity, the same way a lot of people recognized the internet as a huge future opportunity in the mid-1990s to late 1990s.

And with any R&D, most of the experiments are going to fail.

What is going to happen is that at some point, this amazing fundamental growth wave — which is amazing, people are buying and using these products, and there is a lot of revenue being generated and so forth — at some point, we are going to get out ahead of ourselves.

We’re going to have a setback, and there’s going to be a huge reordering of everything, and a whole bunch of companies are going to fail. And then the next generation of companies or the survivors from this first crash will go on and inherit the earth, and AI will be bigger than everybody thinks it will be, although different.

That is the consensus. I think it’s probably right.

Then the only question is where are we? Are we in 1996 or 1997? Or are we in the fall of 1999, right on the precipice of the crash? And what I will respectfully submit is that there is no way to be super-confident about that.

The premise here is that AI is real, and it’s not going away. But some bubbles are just bubbles, right? People bought tulips and they went up in value. Or a couple of years ago, very serious people thought Web3 was a big deal worth taking seriously. It appears to have gone nowhere. Are you confident that AI is real and important things will come of it?

I think so. I look at the adoption of Chat GPT and Claude and some of the other models. There is a lot of fundamental adoption there.

As many people have pointed out, there is not enough adoption or revenue to even begin to fund the extraordinary capital expenditures that are being made to provide these services.

And one of the things that is worrisome is that the models seem to be progressing more slowly than people thought. And their usefulness is not increasing as fast as people thought. And most companies are actually really struggling with how to use it.

Like companies that started a website in 1999 and then realized it didn’t transform their business. They just had a website.

Right. But it feels like this is real. If you look at what Open AI is already doing, the fundamental product is going gangbusters.

The problem with the valuations of all these stocks is that they are priced for the same kind of growth continuing for decades. So if you run into a problem like AOL ran into, which was they couldn’t make the transition from dial-up to broadband, their valuation is going to get destroyed.

Unlike the web bubble, you can’t really be a stock-picker this time around: If you want to invest in an AI company, you can buy Nvidia, and that’s about it.

If you’re a normal person and you’re interested in AI and you think this is real, but also a bubble, what do you do?

For most of us, the truth is you have a balanced portfolio of index funds with a good, healthy slug of cash and bonds in there. You don’t bet on anything.

One of the things that makes this look very much like the 1990s bubble is how AI investing is now juicing the rest of the economy. A lot of the economic growth we’re seeing right now is coming from these investments in data centers. Microsoft, Meta, and a few others are investing hundreds of billions in them.

Paul Kedrosky wrote a great piece about how everyone’s saying it’s just like the railroads — you make the big investment up front and then you get the benefits for a long time.

But as Paul points out, this is not like laying a fiber line that is going to be great for the next 25 years, or a set of railroad tracks you can use over the next hundred years. These things have a three-to-four-year life.

Because Nvidia will come out with new chips. The current chips will be outmoded. A lot of this massive investment in data centers is going to be obsolete soon. There isn’t a huge payback time.

So we are really setting up for a situation in which a lot of the fundamental performance is being juiced by every company in the world panicking and saying, “I gotta invest in AI”. And all venture capital money is going there and so forth.

At some point, that party is likely to stop, and when it does, there’s going to be this big reordering. Lots of companies will go.

It’s not going to shock me if in seven years [we realize] OpenAI took a wrong turn two years ago and went down some path, and somebody else comes up with something new that is the Google of the era.

And Google didn’t really appear until 2000. They missed the whole 1990s, and yet they were the one that you wanted, because they got it right.