Every tech company now seems to have its own AI: Google Gemini. ChatGPT by OpenAI. MetaAI. Spending on AI is reaching record levels, which is fueling a major boom in the stock market. Even the White House wants in on the fun.
So are we in an AI bubble, a period of hyped investment that is sure to deflate? Yes, again Paul Kedrosky, partner at SK Ventures and member of the MIT Digital Economy Initiative. But not the bubble for everyone. think We are in. “AI is obviously a hugely important technology,” Kedrosky said. Today, explained co-host Noel King. And then what?
It’s the money going into AI infrastructure, like data centers, that worries Kedrosky: “We’re spending this prodigious amount of money on the underlying infrastructure for AI with probably no chance of recovering most of that cost, and with a significant chance that most of those assets will lose their value because of the rate at which they depreciate.”
What happens when the bubble bursts? And can the bubbles of the past tell us something about what is to come?
Below is an excerpt from their conversation, edited for length and clarity. There’s a lot more in the full episode, so listen Today, explained wherever you get podcasts, including Apple Podcasts, Pandora, and Spotify.
How much money is allocated to these data centers?
Now they will be on the order of billions. Forecasts exceed $2 trillion in future data center spending. But a growing fraction of the money spent on all these things that allow us to deliver AI, like electricity, comes from debt. And debt comes with obligations. You can’t just walk away from it. That makes this moment even more dangerous.
If AI is so important, why doesn’t it make sense for trillions of dollars to come? Isn’t this what we should be doing?
We should be. But the problem, of course, is that there is this idea of what is called a rational bubble. Everyone thinks they’re doing the right thing, but when you add up everyone’s “rightness,” you end up with a prodigious amount of waste.
It is no different than if we go back to the railway bubbles of the 19th century in both the United Kingdom and the United States. There were simply too many tracks, too many enthusiastic railway builders building tracks almost adjacent to the same places. And this caused an incredible amount of waste. But it also caused business bankruptcies and several market crises throughout the 19th century in the United States and repeatedly in the United Kingdom. It’s not as simple as saying, “Well, this is important, so we should build it and not worry about what it costs or the consequences.”
If so many smart people think we’re in a bubble, why does money continue to flow into data centers and other AI infrastructure at the rate it does?
I’m not convinced that many people think this is a bubble. When I talk to people in the technology industry, the most common response I get is not only that it’s not a bubble, but that it’s probably the most important technology of our lifetime. We have the opportunity to build a superintelligence, a godlike intelligence on top of all these chips and buildings and this electric AI thing that we’re creating. And to say that we should slow down at this point, according to the tech community, is simply a big mistake. But there are people outside of technology who say, “Oh, this is an incredible amount of expense.” The Bank of England said it. Other people are cautious about it, but not within the technology.
The United States and humanity in general have had no shortage of bubbles throughout history. You mentioned railroads; Walk us through some famous American bubbles.
The railroad is probably among the most prominent in the United States and that, again, generated enthusiasm for the idea. The same thing happened in the 1920s during electrification. In the 1920s we went from a single-digit percentage of rural areas with access to electricity, [to] by the end of the decade it was more or less ubiquitous. Everyone had access to electricity. But at the same time, that cold rise of the proliferation of utility companies, of companies that were doing all kinds of questionable things in terms of excessive spending. One could argue that electrification and the frenzy surrounding it gave rise to the stock market boom of the 1920s, which led to the crash of ’29 and helped precipitate the Great Depression.
People are quite familiar with the telecom and dotcom bubbles, but the closest historical analogy to what is happening now is railroads and electrification. In the same way that we don’t need to have two sets of roads to Philadelphia, we probably don’t need the same number of companies delivering what are called these big language models, these artificial intelligence models that people are using. These will shrink naturally.
How destructive are bubbles and what do they tend to destroy?
All of them cause immense damage. It’s a question of how big the bubble is and where the damage is going.
So if you just have an index fund and you think you’re being too conservative, you’re actually deep into AI right now. If everything is reversed, goes 20 or 30 percent in the other direction, you are much poorer than you were. That will change your expenses. And that has implications for recessions.
Isn’t it always the case that the bubble bursts and then what it leaves behind is, perhaps not something beautiful, but something viable?
That’s kind of the pattern of the tech community. But the reality is that almost every financial and technological revolution has caused enormous damage and it may be decades before we get back to where we were before. And as the famous phrase in economics says, in the long run it may work, but in the long run we are also all dead.