The bubble is getting a lot bubblier, folks

Hi, Ryan here, I’m back from vacation and very sunburned. Still catching up on all the posts I missed while I was away, but it seems like using AI is officially a sin for Catholics. So I think it’s worth taking a closer look at the AI bubble. If you haven’t been following it closely, it’s looking a lot more bubbly these days.

The big headline scaring AI guys this week is from Business Insider, which reported out some comments made by Uber COO Andrew Macdonald on the Rapid Response podcast. To paraphrase Macdonald (he’s not a super concise speaker lol), he said replacing human workers with AI is still too expensive. And he’s not the first Uber executive this year to make the same claim. Uber CTO Praveen Neppalli Naga told The Information something similar back in April. Earlier this month, noted AI critic Ed Zitron wrote a piece putting a lot of this together, titled, “AI Is Too Expensive,” if you’d like to go deeper there.

But the simple fact is that, aside from all the problems AI models still have producing work that is error-free or even useful, it’s also just expensive as hell. And it continues to get more expensive all the time. I’ve seen folks like X super-user Aakash Gupta, who definitely writes almost all of his posts with AI, claim that the recent comments from Uber executives is “the strongest AI bull signal anyone has published this year,” but I don’t think people who use terms like “bull signal” are actually ensouled individuals.

There are other warning signals across the AI industry that are flashing bright red right now. Starbucks had to roll back an AI inventory tool because it kept hallucinating. The Verge reported that Microsoft is rethinking its dependency on Claude Code. And Duolingo’s stock price has tanked in the last year, after the company went all-in on AI. Oh, also, Sen. Elizabeth Warren, who has never met an emergent tech industry she didn’t want to crush into dust (complimentary) is going after AI companies now, as well.

I am going to go out on a limb and say that these are not isolated incidents. And if more companies come out of the woodwork and say that AI is too costly and/or bad to rely on, it puts AI companies in a pretty terrible bind. It would mean that their products are too expensive (for them) to make any real money from consumer-grade products and too expensive (for corporations) to make any real money from enterprise-level products. Here’s a totally unrelated link to a Bloomberg piece about the circular financing driving the AI economy.

Like last week, when I said it was time to imagine what an internet after Google Zero might look like, I think it’s also time to imagine what an AI bubble burst would actually mean. They are, effectively, the same thing, actually. A hard reset to an online wilderness period. And the closest analog from recent history we have for imagining an AI crash is either the dot com bubble of the early 2000s or, if things get really gnarly, the Great Recession. But let’s go with the dot com bubble because it’s a less terrifying prospect. If it ends up being the other one, I’ll find you at a campfire in the wasteland in a few years and tell you what I think.

The biggest losers of the dot com bubble were a whole lot of IPOs you probably don’t remember, but here’s a good Reddit thread with some examples. But the bulk of these companies were web portals — for search, for e-commerce, for proto-cryptocurrencies. The bubble was similar to a land grab and when it popped it wasn’t totally different from a housing crash. The tech is different, but the driving force behind our current AI gold rush echoes it pretty closely. Get in early and colonize the internet. Except the internet is, by definition, infinite, which makes any kind of internet real estate pretty worthless. Newer tech companies have realized that what they really need to colonize is attention, but that creates a myriad of other problems, like understanding taste and coolness and the psychology of mass media. Stuff that has, for the bulk of the social media age, been automated by algorithms because rich loser nerds don’t understand it. The (false) process of AI is that it’s finally a way to colonize online real estate as well as attention.

What caused the dot com bubble to finally pop was speculation and over-spending. Buying up catchy domains just wasn’t the growth industry people thought it was. But the most important takeaway for me, one I have argued here several times, much to the dismay of my own readership, is that the dot com bubble did not kill domain names or websites. In fact, even with the open web’s several anemic state currently, there are more websites than ever. But simply owning one is no longer its own industry.

My most reasonable prediction about all of this is that an AI crash will play out similarly. Investing in an AI model will probably sound as silly as stockpiling domain names. Even if many of the big AI companies of the moment might implode and the term “AI” might even be so poisoned that it ends up being called something else, like “machine learning.” The tech, though, is probably here to stay. It may, like the humble website 25 years ago, even become more central to our daily lives, depending on what people end up doing with open source models, in particular. Before you yell at me, I’m not saying this because it’s what I want to happen, merely that there’s historical precedent. File types, protocols, algorithms, specific features, they all cause a gold rush before just being another thing anyone can add to their particular stack.

And while I’ve already got you hot and bothered over this take, I’ll double down and argue something even spicier. If companies like Google and Meta cannibalize their core product — search and social respectively — and go up in flames chasing AI that might actually be a good thing! Well, not for people who work there lol. But we both found content and one another online before they existed and there are more tools than ever now to do so now without them.

This is a question I keep asking myself about, well, the entire internet lately. “What if this all just doesn’t work this time?” What if AI companies can’t replace search? What if streaming video can’t replace Hollywood? What if short-form video apps can’t replace social media users? What if Silicon Valley can’t brute force their products — and ideology — onto the masses again? To take it even further, what if the dot com crash and the AI crash are actually part of the same 25 years epoch of technological stagnation? I’m not going to lie, I find this line of questioning exhilarating.

Anyways, the fact that crypto.com CEO Kris Marszalek bought the domain AI.com for $70 million earlier this year, in what is considered the most expensive domain sale of all time, feels worth mentioning here.