Prof G Markets (Live)
Episode Title: Is AI Worth the Cost? The ROI Reckoning and the Coming Market Correction
Location: The Castro Theatre, San Francisco, CA
Hosts: Scott Galloway & Ed Nelson
ED: We're going to talk about a topic not enough people talk about called AI.
Nearly 50,000 workers have been laid off this year supposedly because of AI — that's almost as many as in all of 2025. For companies adopting AI, the thesis is simple: AI is supposed to do much of the work that humans do. In recent weeks, however, that thesis has hit a roadblock. More and more companies are reporting that despite the enormous power of AI, the technology is actually more expensive than the humans it is supposed to replace.
Uber, for example, just blew through its entire 2026 AI budget in just four months. According to the COO, it is now getting harder to justify AI costs within the company. Microsoft is cancelling its Claude Code licenses across multiple divisions because it's simply gotten too expensive. And over at Nvidia, one executive said that the cost of compute is now "far beyond the cost of employees."
Which all raises a crucial question for the AI industry: at what point does AI actually stop being worth it?
This has blown up basically in the last 48 hours, with many companies coming out and saying they're not as confident about this whole AI thing as they used to be. ServiceNow is another company that just blew through their entire Anthropic budget. Technical staff at Stripe are reportedly spending nearly $100,000 on AI tokens every day. Salesforce is on track to spend $300 million on Anthropic tokens this year. Shopify said their earnings were "partially offset by increased LLM costs." We heard similar things from Meta, Spotify, and Pinterest. One Anthropic employee said his Claude Code bill came out to $150,000 in a single month.
In some cases, it's getting very, very expensive.
We've also seen an incentive — especially among tech companies — to use AI as much as possible. There was this idea that employees would engage in what we call "token maxing," where you use as many tokens as possible from your AI API. Companies like Meta and Amazon have even created internal leaderboards tracking how many AI tokens employees are using. The people using the most tokens are seen as the most AI-forward, the most AI-deployed — the ones who are going to get recognized, maybe even promoted. And this has resulted in extraordinary costs on the AI front.
Now we're starting to see the next phase of this, Scott, where companies and their executives are beginning to realize: this is a little expensive. So the question becomes — at what point will AI actually pay off?
I'll pose that question to you: at what point is it too much?
SCOTT: I think we're already seeing hints of it, and I think it comes down to incentives. You were talking about how companies are trying to incentivize people to use AI more — and that's kind of an interesting part of the ecosystem right now. The adoption layer is trying to get people to use it, and companies have put in place the incentives to do that.
But there was a recent survey by a professor at MIT who found that about 5% of the projects people are using tokens for can actually be connected by CFOs to some sort of return. So while I think they're really intoxicated by it — and talking about AI as much as you can in your earnings call is like adding "dot-com" back in the '90s — I think you're already starting to see some fatigue.
And I think the AI companies are trying to get public as quickly as possible to raise that cheap capital before things start to — I don't want to say unwind, but...
You can see how the string gets pulled here. A large company, a CEO who has a lot of credibility in the industry, just comes out and says: "We're dramatically scaling back our AI investment. Let's be honest, folks — we're just not seeing the return we'd initially hoped."
And then Nvidia reports its first miss. Nvidia has beaten its estimates 15 quarters in a row. Nvidia's first miss probably takes the entire market down five or ten percent.
You are seeing some productivity gains from this and quite frankly, they look as dramatic, if not more dramatic, than the internet. But look what happened in 2000. This definitely does feel like '99.
And I'm waiting for the first CEO to come out and say we have to get procurement involved and dramatically scale back our expenses. I don't think it's that romantic, honestly. I think it's just going to be a traditional Fortune 500 company that starts the narrative: okay, this has been fun, but we have to dramatically decrease our AI investment because we're not seeing the ROI we'd anticipated.
ED: Yeah. I mean, we heard a quote this week from the CEO of Match Group — not a huge company — but he said AI is costing them $5 to $10 million a year, and his exact words were: "I think we're benefiting from it, but it's hard to feel." So that's not great if we're supposed to be riding this multi-trillion dollar technology that's going to transform our economy.
[link] [comments]