What's next for AI and has its explosive growth in 2025 created a bubble?

Nick Schifrin:

This year, perhaps the most significant business and economic story has been the development and spending around artificial intelligence.

A.I. spending is driving one of the tech industry's most explosive periods and playing a big role in overall U.S. economic growth. A few companies are churning immense profits, most of all Nvidia, which posted record gains of $32 billion, a 65 increase in one year. But there are very real questions about whether this is a bubble.

Geoff Bennett dived into that in a conversation he recorded recently.

Geoff Bennett:

Experts say this boom exceeds almost anything else you can imagine, spending that goes beyond what was spent on the Manhattan Project and the Apollo missions to space.

But a number of experts are concerned that the trillions of dollars spent and intense competition is creating an A.I. bubble that won't sustain the kind of earnings growth to match all of that investment. And if the bubble bursts, it could affect more than just Silicon Valley.

For more, we're joined now by Cade Metz, technology reporter for The New York Times who covers the world of A.I.

Thanks for being with us.

Cade Metz, Technology Reporter, The New York Times:

Glad to be here.

Geoff Bennett:

So let's start with the basics. What is driving this explosion of A.I. investment? Why does it seem like all these companies are piling in at once?

Cade Metz:

Well since the arrival of the chatbot ChatGPT about three years ago, we have seen this technology steadily progress and slowly work its way into people's lives, into the lives of office workers.

In many ways, it's a transformative technology. It's helping people search the Internet in new ways. And, inside offices, it's helping to transcribe meetings, helping doctors and other professionals do their jobs maybe a little bit faster.

But Silicon Valley sees much bigger things ahead. They see this technology continuing to improve, becoming more and more powerful over the next few years. So they're investing those hundreds of billions of dollars into the data centers that not only allow them to improve this technology, but serve it up to a much larger portion of the population.

Geoff Bennett:

And, overall, you have got trillions of dollars flowing into A.I. Is this growth sustainable or are we in a bubble territory here?

Cade Metz:

That's the question that everybody's trying to answer and no one can quite answer.

This is enormous spending, to say the least. And it's a bet on the future. These data centers are not only expensive. They take years to build. So all these companies not wanting to miss out on this big boom are having to make a bet two, three, four years into the future that they will be able to pull in the revenues to pay for this.

Maybe they can. Maybe they can't. No one can quite agree. It's really a timing issue. The revenues are already coming in for a lot of these companies. It's a question of how quickly it will come in and how many companies can pull in those revenues. There are so many companies in this race. A lot of people believe not all of them can win out.

Geoff Bennett:

Is that why it's so hard for these companies to establish A.I. dominance?

Cade Metz:

In part.

You have got many big players. The Googles and the Microsofts and the Amazons of the world are doing this. And then you have got more nimble start-ups like OpenAI and Anthropic. Meta recently doubled down on this, the company that owns Facebook and Instagram. They have built a new A.I. lab. Elon Musk is in the mix.

People in the Valley talk about FOMO here, the fear of missing out. No one wants to miss out on this technology. So the race has expanded even over the past year.

Geoff Bennett:

Let's return to the data centers, which you mentioned, because the spending there has just been enormous. What risks does that pose to the markets, to energy systems, and to the broader economy?

Cade Metz:

To the broader economy, the concern here is that so much debt is being taken on to build these data centers. For years, the big players I mentioned, the Googles and the Amazons and the Microsofts, built these giant data centers largely with cash on hand or essentially with cash on hand.

Those are companies that pull in billions of dollars in revenues a year. They can afford to build these facilities. But you have got so much demand for this extra computing power that so many other companies are now in the mix and building these data centers for various reasons, either for themselves or for their partners.

And those companies are taking on far more debt than we have seen in the past. And that's where the risk comes in. Eventually, that debt has to be repaid. And if the revenues aren't coming in from these companies by that time, you have a problem.

Geoff Bennett:

So it's this idea of a circular loop. You have the tech giants investing in the A.I. firms, which then use that money to invest in those same tech companies' infrastructure. Is that why folks are raising a red flag?

Cade Metz:

Well, that's part of what's going on. All these companies are partnering up, hoping to kind of bootstrap the whole industry and move it forward. And so you do have deals like that, where, for instance, one company will take investment from one of the big giants and then immediately spend that money with the same company.

They see that as partnership. Others see that as a sign that the market may not be as healthy as it seems.

Geoff Bennett:

What about on the substance, the concerns about how A.I. is being deployed, often poorly with little regulation, a flood of low-quality, misleading content on the Internet? How are the companies responding to those concerns and worries?

Cade Metz:

Well, in some cases, they're working to address that disinformation issue you talk about. These systems do make mistakes. They do cause problems. We have got three years documenting all that.

And there are constantly efforts to improve that problem. But you have to realize this technology fundamentally makes mistakes. This is a technology that is built by analyzing vast amounts of data culled from across the Internet. It looks for patterns in that data. And that's how it learns its skills.

But that also means that, as it learns, it's going to make mistakes. It's going to learn from bad data, but it's also going to make mistakes just because these are probability engines, right? They're doing something based on what they have seen in the data. And that means, fundamentally, a percentage of the time they're going to make mistakes. People have to realize that that's always in the mix.

And that can also -- in addition to causing all sorts of problems in -- just in the information as it's distributed across the Internet, it can cause problems in other ways these systems behave. The hope is that these systems will do more and more important tasks as the years come.

But if those mistakes are in the loop, it becomes harder to complete those tasks.

Geoff Bennett:

Problems, to be sure.

What about the promise, the promise of A.I.? What have you encountered in your reporting that you find most exciting?

Cade Metz:

The most exciting part of this, I think, is in health care.

What we have seen is, the technology that drives something like the chatbot ChatGPT, which many people are now familiar with, that same technology can be used to aid drug discovery. Essentially, it can help design medicines and vaccines that can help deal with illness and disease. That's the most powerful aspect of the technology, in some ways the most promising, and in other ways the most important.

Geoff Bennett:

Cade Metz of The New York Times, thanks again for your time. We appreciate it.

Cade Metz:

Thank you.