Will the AI bubble pop with a bang, or go out more like a sad balloon?
Anthropic was preparing to soon hold a record-breaking IPO, just months after Elon Musk’s SpaceX held its own. But Reuters reported on Tuesday that the San Francisco-based AI giant is now lowering its valuation target amid dramatically widened net losses. It’s yet another symptom of the AI bubble spectre, as investors question when, if ever, AI demand will justify the cost of the associated buildout.
To understand that question for ourselves, BetaKit brought in the big guns: economic commentator and New York Times best-selling author Kyla Scanlon. She joined the BetaKit Podcast earlier this month to help us finally understand what’s going on in the AI and labour markets, as well as ease (or intensify) the AI bubble fears.
An excerpt from that conversation is below.
The following interview has been edited for clarity and length.
BetaKit editor-in-chief Douglas Soltys: About a month ago, BetaKit walked through a series of our favourite AI bubble red flags on this podcast. Do you have a preferred AI bubble red flag, or are you optimistic?
I’m really watching the debt markets. The companies are borrowing quite heavily; they’re spending a ton of money, and I worry about that because once you involve debt, that’s when things get really serious. That’s when systemic risk can start to spread because of who owns that debt. If insurers are buying up that debt, that exposes them, therefore, it exposes households to the AI bubble bursting.
BetaKit managing editor Sarah Rieger: That was something we definitely wanted to talk about today; KKR issued a warning last week. Do you think we’re at the point where all this over-leverage is close to tipping?
There was a really good paper out of Brookings called Financing the AI Buildout. The author said that the AI companies are going to need to make, like, $10 trillion by 2032. It’s really hard to conceptualize that much money. Jared Bernstein, who is the former chair of [former US president Joe] Biden’s council of economic advisors, summarizes it as Amazon and Google would have to build a whole other Amazon and Google on top of the existing Amazon and Google just for AI within the next 10 years in order to make the app buildout work.
RELATED: “The math is not mathing”: How AI bubble fears are changing Canadian VCs’ investment approach
If you think about it, it took 30 years to build Amazon and Google the first time. We’re not totally sure who’s going to be paying for all these AI products. All of a sudden, they’re just spending all this money, and you can’t even plug in some of the data centers. There’s just not enough power, particularly in the United States. So it just seems like the AI industry is a little bit over its skis.
DS: Do you think AI is propping up the stock market?
I think the stock market is not even the stock market. The stock market is AI companies. Like, 75 percent of companies in the S&P 500 are below their 50-day moving average. So, it’s really just a couple of guys pulling this thing up. If you look at the S&P equal weight, that thing’s down four percent. So the S&P is really being driven by these AI companies.
I think it’s just the classic signs of a bubble, where people are like, ‘Okay, cool. This is really exciting and awesome, and I’m going to put money into this buildout because I think that there’s no other alternative because of the weak points in the other parts of the economy.’ The United States’ economy definitely has a cold right now, one might say, and when the US sneezes, the whole world catches a cold, and we’re certainly seeing that.
DS: With AI, we’re seeing both people spinning up companies fast, as well as young people having the worst possible time finding their way into the labour market. How does that play into all of this?
I think for young folks right now, there’s this sense that you studied really hard in school, you did everything right, and it didn’t matter. We haven’t really provided much guidance for anybody to navigate this new economy, and I think for a lot of corporations the number one goal seems to be saving money, and the way you save money is by not investing in the future, it seems, in terms of your employees. I think it’s creating downstream effects that a lot of companies are underpricing. If you don’t hire juniors now, you’re not going to have seniors later.
RELATED: Waving the AI bubble red flags
At the end of the day, somebody else has to buy your product, and ideally you want people employed so they can buy your product, just from a functional perspective.
SR: If we are in a bubble, is there a possibility here of a managed release of air from that bubble, or could this all just pop? Is there a way out that is less dramatic than that?
If the bubble does pop—if it is a bubble—it seems like it’s going to be pretty slow just because of how the money is wrapped up and because there are so many big banks involved. People are trying to protect against the downside here, so it seems like it would be a slow release of air.
I think the highest probability scenario is that AI demand doesn’t totally match what the buildout expected; we have a bunch of empty data centers, and the companies have to figure out how to pay back all that debt. Some of them might not be able to. They’ll go bankrupt and default on those loans. I don’t think it’s going to be like a big implosion, but who knows? That could change depending on how much more they borrow and who they borrow from. Once you get life insurance involved, that gets dicey real fast.
With files from Douglas Soltys and Sarah Rieger.
Feature image courtesy The BetaKit Podcast.
The post Q&A: Why economist and author Kyla Scanlon thinks the AI bubble will pop slowly first appeared on BetaKit.