Why Arthur Hayes Thinks These Two Catalysts Will End the AI Bubble

So again, we're in this. There's an excellent paper written by two folks from China University of Hong Kong and University of Macau based on this model phenomenon of the game theory behind capex spending in AI, and the authors call it the Red Queen effect. Where they basically make a distinction between ordinary capex where you build a factory and regardless of if your competitor comes up with a competing product, your factory still has use and doesn't become obsolete immediately. But in AI, the best model wins and the best is the best model, right? If you spent all this capex and your model is not as good, then nobody uses your model, right? It's like Anthropic versus OpenAI. Claude is marginally better. Everybody uses Claude, right? And so they stop using OpenAI. So the hundreds of billions of dollars you spend on capex is worthless. So what do you do? Well, you constantly need to be spending money to create the frontier model. Otherwise, your entire capex debt is worth zero. And this generates this perpetual motion of spending. And we're seeing this now in earnings for a lot of companies, which is why you're seeing this sort of vertical ascension in price for a lot of these inputs to AI data centers and chips and other things in the US and North Asian markets because people are realizing yes, okay, if the model is good enough, people are going to use it. Anthropic's revenue has gone up like something like 10x or something crazy in the last less than one year basically because Claude is just good at doing stuff and so there is a demand for compute and these hyperscalers have to keep spending this money. Now obviously this can't go on forever. There's no such thing as sort of an infinite perpetual motion machine in this universe. And so I, you know, some others have come up with this idea like okay well what usually ends a frenzy? There's some sort of IPO or some sort of mega merger that's just so big and so ludicrous in terms of the underpinning financial assumptions that the market just can't adjust it. And so like you know the stock lists at 100 and the next day it's already down at 50. You're like oh shit, you know, if people aren't buying into this IPO and it's just sell only, this has to be the top of the market. And so, I think there's going to be a situation like that. I don't know when it's going to occur, but it's going to be pretty obvious at the time like this mega merger, this mega IPO, it debuts and it just completely flops. And at that point, people are going to start asking questions. Oh, is the trillions of dollars we spent on all this capex really worth it? Have we generated enough use cases for this or will we generate enough use cases in the immediate future? Why do I own this company at you know 1 million times earnings if the category leader gets IPOed and the stock's down 20% when it should be you know multiples higher on the first day after listing. So, I think these psychological qualitative things feed into the let's ask some questions about all this money that we spent and oh shit guess what? I don't really think that I spent the right, you know, this money was well spent, so maybe I should start selling down my holdings as well. And that's the end of this market, this bull market bubble, whatever you want to call it. And the second is politics. At the end of the day, most people around the world, especially in the United States, don't own a lot of stocks. Yes, you know, people are making lots of money hand over fist buying some random memory company and it's up 40% pre-market because everyone discovered it overnight, but the majority of Americans don't own a lot of stocks, but their electricity prices are going up. The price of labor is going up to build these data centers. The price of raw material inputs into electrification and the construction of data centers is making their lives more expensive and they get nothing back for it. And guess what? They're worried about losing their job at the same point. And so I think there's going to be a political backlash that's going to hand the opposition party, you know, the team blue Democrats a sort of a narrative on a plate which is we are anti-AI, we're for the worker, we're for humanity. Like there's so much good rhetoric that you can put around this that sort of catalyzes the fear that people are feeling and the inflation that they experience on a day-to-day basis and saying vote for me. I'm going to take us back to pre-AI or I'm going to do responsible AI, whatever the fuck that means. It doesn't matter. It's just going to be a political slogan and it's going to be up to the team red Republicans to come up with a counter of okay well everyone's making so much money no one gives a fuck but again 90% of Americans in the United States don't own a lot of stocks. So I think again it's going to make for a very interesting 2028 presidential election and again whoever wins it doesn't matter. The rhetoric is going to be very good from the opposition because it's true to some extent and therefore the investors be like well okay well you know 55 45 whatever you handicap the odds of which side winning you still have to de-risk your portfolio because there's a significant pretty much 50/50 chance that the team blue Democrats take control and all of a sudden this rhetoric could become policy oh and all of a sudden all this money that you spent at these ridiculous valuations is no longer worth anything because they're going to tax you. There's a tax proposal put out by some leader in Korea overnight caused a Kospi to fall 5%. So investors are very concerned that the you know 95% of the population who's not benefiting at all from AI yet is going to be like fuck this tax these fuckers give me my money back.

So, at the end of the day, you know, you create credit to do stuff, right? To the perpetual motion of the economy. Like a recession is when GDP falls. Doesn't have to fall that much, but if the pace of the economy slows, you can't service your debts. And so if the pace of the economy falls in the short term and I you know I agree that in the long term AI is going to be this awesome productivity tool hopefully we as humanity use it wisely and create an awesome economic and life experience for everyone around the world. But we are all humans. We are selfish. We don't share. And in the short term there's going to be some pain. It's not as if there's immediately a job in some sort of AI field for the knowledge worker who just lost her job at Coinbase, right? There isn't. And so that person who was making a few hundred thousand a year is now at least in the United States on unemployment insurance where I think usually lasts about 26 to 27 weeks usually caps out around annualized $40,000 a year depending on your state. So you take somebody who was making $150, $200,000 now they're making 40. Okay, they have to make a lot of adjustments. They're not buying the sweet green salad. They're not buying the mocha loca fucking Frappuccino from Starbucks for $10, right? They can't make their car payment. They have to downsize their dwelling. Maybe they're missing credit card payments and all these debts are assets on the banking balance sheet. And so the banks have a growing hole in their balance sheet due to the productivity gains of AI not being passed on fast enough to ordinary workers. And so that is the it's a timing mismatch. I'm not saying that AI is not going to create this utopia of abundance in you know medium to long term. All I'm saying is right now the way that we have set up our economy globally and this is not just United States, it's every economy around the world. It's fractional financial services sector that sort of needs a constantly growing GDP activity. Well, this AI knowledge intelligence boom goes completely against that and will destroy the financial system unless they print enough money. And so I believe that they're always going to print enough money. And that's sort of, you know, my investment thesis for my entire portfolio.

So you know if you think Elon believe everything he says, right? So we're going to have robots, right? And the cost of labor is going to go to essentially almost zero. Right? The cost of intelligence of doing things is going to be moved to very very little as well. And so we've got to automate all these things, right? Your babysitter is a robot. You know, you're not using a lawyer, an accountant to do random stuff, right? So anything that we do can be automated and done in an intelligent fashion. And that leaves us as humanity free to pursue, you know, hanging out, doing arts and crafts, playing sports, whatever it is, right? Whatever it is that makes you happy, apart from doing your job at some like, you know, company that looks like The Office, the television series. So, we're going to remove all that nonsense and just focus on, you know, communing with each other and producing things that are fun to consume. Again, that's the utopian future. And how do you get there? Would probably have to have some sort of taxation mechanism on whatever the entities that control this intelligence or these robots. Again, I don't know what that situation is, but we're going to have to have a conversation about that. Otherwise, we're going to have such hyper inequality that it's going to cause, you know, potentially a very very big social strike.