← Back to blogYouTube Video

Published September 17, 2026

AI is a bubble, but not for the reasons you'd expect

Can't play the video or having issues? Here's the direct link.

Summary

AI is an economic bubble, not a technical one: OpenAI's $850 billion valuation sits on $6.7 billion in revenue against $12 billion in losses, Anthropic's near-trillion valuation is just as disconnected from profit, and neither will ever recover its investment once Chinese and open source models keep commoditizing the product and the subsidies end. But the technology isn't going anywhere — nobody's going back to manual coding, and Jevons Paradox explains why rising developer productivity increases demand for code and likely grows jobs rather than killing them. My take: the labs may go bankrupt, but if you're not using AI to write code, you'll get left behind.

Key takeaways

  • Separate the economic bubble from the technical bubble — AI companies are wildly overvalued, but the underlying technology is here to stay.
  • Don't confuse revenue with profit — an $850 billion valuation means nothing when the company loses $12 billion a year and has never turned a profit.
  • Commoditization destroys pricing power — when cheaper Chinese and open source models do the same task at a fraction of the cost, LLMs compete on price, not on revolutionary value.
  • Treat unsustainable subsidies as a red flag — a $200 plan delivering $5,000–$8,000 of inference is roughly a 50x subsidy, and the real price will change everything.
  • Use AI to write code or get left behind — nobody is going back to manual coding, guaranteed.
  • Expect jobs to grow, not shrink — Jevons Paradox means more productivity expands human ambition, so demand for code and the people to build it rises.
  • Scaling AI costs grow exponentially, unlike software — adding users gets expensive fast, which is a core reason these labs' economics don't work.

Description

AI is an economic bubble but not a technical one, and this video breaks down why. I cover how OpenAI and Anthropic's valuations don't match their actual profits, why Chinese and open source models are wrecking the pricing power of US AI labs, and why Jevons paradox means coding jobs are more likely to grow than disappear even as AI makes developers more productive.

Transcript

AI is a bubble. There, I said it. AI is definitely 100% just a bubble, but not for the reasons you expect, right? There's a bit of nuance to it, and I'm going to go through that in this video, just as soon as I'm done cleaning up my glasses. I'm going to do that, and then I'm going to go over to my screen, where I'm going to show you some shitty little PowerPoint presentation I made, and I'm going to prove to you why AI is definitely a bubble. Let's get into it. So let's talk about the AI bubble, right? When people talk about the AI bubble, the big problem is that they don't distinguish between the two kinds of bubbles over here, right? And for the purposes of this video, I'm going to divide the AI bubble into the economic bubble, if I can spell it right. How do you spell economic? There we go. And the technical bubble. Because, like I said, this is a bit nuanced, and I don't think people understand that. So let's first talk about the economic bubble, right? The economic side of AI. This is most certainly a bubble. And I'll fight anyone who disagrees with me over here because the economics of AI just don't make sense, right? Firstly, the entire product is just extremely overvalued. You have companies like OpenAI and Anthropic that are worth, I don't even know anymore, hundreds of billions of dollars. I never checked. Actually, let's go check. So OpenAI is literally valued at $850 billion. And a company worth $850 billion is not profitable. If I Google its profit, they get $6.7 billion in revenue, but that's just revenue, right? That's not profit. When you actually count the expenses and the losses, which are literally $12 billion, you realize OpenAI isn't even profitable. So a company that isn't even making a profit, that hasn't made a single dollar in profit ever since it came into existence, is worth almost a trillion dollars. Make that make sense to me. I could do the same thing with Anthropic as well. Let's Google Anthropic's valuation. $965 billion. So it's even more than OpenAI, right? Almost a billion-dollar valuation, and they only made almost $12 billion in revenue. So I think it's safe to say that these companies are extremely overvalued for what they're offering. I'm not saying AI and LLMs aren't a very good, very valuable product, but it's not worth trillions of dollars if your company can't even turn a profit. And now the second argument is that, well, they're just offering these models for cheap right now, but later on, when we're all dependent on these models, they're going to increase the prices, right? Yeah, they could do that, except we then have Chinese models that are just destroying OpenAI's valuation entirely, right? Like, if an OpenAI or Anthropic model does something at a hundred or a thousand times the cost of a Chinese model, right, which one do you want to pick? Especially if both of them offer similar kinds of performance. Because look, 99% of businesses, like the vast majority of businesses, and the vast majority of people, don't really need the most intelligent frontier model in every single task, right? Almost all tasks that a company or an individual is trying to do with AI can be done with cheaper Chinese models or even cheaper open-source models from the US, like Gemma 4 from Google, right? So how exactly do OpenAI and Anthropic plan to recover all of their investments when nobody's even going to use their product for the actual price that they should be charging? When all of these subsidies and AI investment is gone, when the real price of these models comes out, nobody's going to use these guys because they have so much cheaper and better options available. That right there is the economic bubble of AI in a nutshell. These companies, these OpenAI and Anthropic, they seem to think that this product is so revolutionary, it's going to change the world and it's going to, you know, help them recover their trillion-dollar investment somehow. It's not. It's really not. LLM models have become commoditized. Like, we treat these models as a commodity now. We don't treat them as some revolutionary product, right? If I can get a cheaper model from some other company that does this exact same thing, I'm going to go for it. These things are a commodity now. They don't compete on value, they compete on price. And this is largely because open-source models have caught up with the closed frontier models, and that has completely destroyed the valuation of all of these other companies. So AI is an economic bubble, and nobody can disagree with me on that. But I encourage you to give it your best shot in the comments below. Anyway, as if all of that wasn't enough, there's also a huge difference between scaling, let's say, AI models and scaling other tech products. When you try to scale a software, right, the costs go up very, very slowly, right? Something like, like I could start some sort of a SaaS product or company, right? And I could get to like 10,000 users, and my costs would be very, very low, right? I could then get to 100,000 users, and my costs would still be very, very low, but they'll slowly go up, right? Like, the cost of adding new users in software is very low, right? But when it comes to AI models, because they're so expensive, the cost of adding more users just keeps going up exponentially, right? So it just becomes so hard to scale these products and get millions of users with it because their expenses go up so rapidly. And remember, these companies, OpenAI is not charging the pricing that it should be charging. Anthropic's individual subscriptions are just full of subsidies. Like, the $200 Anthropic plan for Claude Code gives you like $5,000 to $8,000 worth of inference. That is insane. That, like, I don't even know, like, 50x subsidies. This is obviously unsustainable. And once these companies start charging the real costs, nobody's going to use them. So the whole thing is just one big bubble that is being propped up by all of this hype in the media and overvalued valuations and figures in the stocks. And it's just not going to last. Sooner or later, it's going to pop. However, even though AI is economically just one huge, big bubble (huge and big mean the same thing), it is not a technical bubble. Even though the economics of it don't make sense, even though these companies will probably go bankrupt, or at least they'll never make back their investment because these models have become commoditized. So even though AI is an economic bubble, it is not a technical bubble. And let me explain the difference between both of those things. Firstly, nobody's going back to traditional handwritten code. Nobody, certainly not me, and nobody I know will ever go back to writing code by hand, because it's just so darn inefficient and slow, right? I can build things with AI now that would take me just an hour, or more realistically, 10 minutes. That would have taken me weeks if I tried to write the code by hand. So nobody's going back to traditional coding, right? This technology is here to stay. All of us are going to continue using AI models for coding. That's not going anywhere. But interestingly, jobs are going to increase, not decrease. Why? Because of efficiency gains. And this is a little paradox that you guys need to understand. But let me just emphasize one more time. AI is not a technical bubble. It's an economic bubble. The technology is here to stay. Nobody's going back to traditional coding anymore. We're all just going to use AI models to write the code, even if we have to use really cheap Chinese models like Kimi or Qwen or anyone else, right? We're not going to go back to writing code by hand. This technology is here to stay. So for those of you who are, you know, on the fence about AI, you don't write code using AI, you're still writing code by hand, I want you to know, no matter what your feelings are towards AI, you're gonna get left behind. Even though OpenAI and Anthropic are not going to make a profit for the next 50 years, even though this technology is a huge economic bubble, even though all the hype in the media is completely false, despite all of those things, if you're not using AI to write your code, you're going to get left behind, guaranteed. Even if all these AI labs were to go bankrupt tomorrow, the technology would still stay and everyone would still use AI models to write their code. So if you're not using them, you're gonna get left behind 100%. Anyway, let's talk about the paradox over here. So this is something called the Jevons paradox. I hope I'm pronouncing that correctly. I just learned about this like yesterday. And basically, it's a paradox because you would expect that as the, let's say, and actually, I should call this supply instead of productivity. So it makes a bit more sense. So when you have some resource like code, and yes, in this case, code is a resource, you would expect that as the productivity of each developer goes up using AI models, because they can supply more code because code has become cheaper. As the supply of code increases, you would expect the demand for code to go down, right? Because supply and demand, if the demand increases, supply decreases, if the supply increases, demand increases, right? No, wrong. And this is the paradox, right? What ends up happening is, instead of the demand going down, it goes up. Why? Because you can now accomplish more things, right? Human ambition doesn't stop just because you're getting better and better at something. When you can produce more code, you want to produce more code. Just look at how much custom software people are building for themselves nowadays. Just look at how much code the average developer is producing nowadays. Look at how many more features each company is shipping nowadays. I'm sure I could find some news article about this as well, but basically, instead of reducing the working hours, AI models often cause humans to work more. And, you know, there's a big bunch of reasons for this. But one of the reasons is that we have just become so much more capable to do so many more things that we are then doing those new things. And that takes up so much more time. So even though the supply has increased so much, because these things have made it so much easier to produce code, we are producing way more code now. Because it's just so much easier, right? And that right there, ladies and gentlemen, is the Jevons paradox in action. So if I were you, I wouldn't worry too much about jobs disappearing. Because the thing is, when you can do more, you just do more. And to do more, you need more employees, more resources, more people. That's how human economics have always worked. So I wouldn't worry too much about jobs disappearing anyway. But yeah, in a nutshell, the technology is here to stay. Nobody's going back to traditional coding. All of us are going to use AI models for coding, and jobs will probably increase, not decrease, due to the efficiency gains that we're getting from AI models. And these American companies are probably never going to make their investment back because AI is one big economic bubble that's being propped up by subsidies, unrealistic valuations in the stock market, and the scaling issues of AI because it's such an expensive technology. Not to mention Chinese and other open-source models completely destroying OpenAI and Anthropic's valuation because if you can use a cheaper model to do the same task, why would you use a more expensive model? So yeah, just sharing some thoughts on economics. Hope that was useful and gave you some insights. If not, I mean, you already watched the video, so what can I do, right? See ya.

Share this article

Hassan Aziz pointing up

Dude, you’re at the bottom of my landing page.

Just book a call already if you’re that interested.

You scrolled all the way here.