AI Token Prices Hit Record Lows - OpenAI is DEAD

Eli the Computer Guy (0:00)

Coming from CNBC, AI token prices are hitting new record lows, which is basically something that I would expect, right? The longer a technology product has been around, basically the value of that product is going to go down, right? The first computers that were ever invented cost millions and millions and millions of dollars, and now you can buy a Raspberry Pi for only one or two million dollars. No, anyways, you can buy a Raspberry Pi for, like, I don't know, a hundred bucks, I suppose, at this point in time, right? So as the technology has been around, the cost, the price of that technology goes down. One of the weird things we've been dealing with, with the whole artificial intelligence revolution, is apparently none of the people who run these AI frontier labs seem to understand, you know, basic technology business, that this is a normal process. And so there has been the idea, that the technology is going to get only more valuable as time goes by. And that's why they've written the contract, these absurd contracts for hundreds of billions of dollars. If you look at OpenAI, $300 billion contract to Oracle alone. One single company, $300 billion contract. You know, Anthropic, they just signed another $35 billion contract. They are just throwing money around, because the idea is the value of this technology is going to continue to get better and better and better. One of the arguments that I've made for a long time is that I think these frontier models basically hit the level for most people they needed to hit probably one or two years ago. Now, with that, there could be a whole bunch of arguments. Did it hit that level two years ago? Did it hit that level six months ago? Did it hit that level six months, two weeks, three days, and five hours ago? Look, we can have that fucking argument. But I think the reality is, is at some point in the past, however far away it was, the frontier models basically hit a level that's good enough for most people's tasks. I've talked with CEOs in companies that heavily use artificial intelligence. Some of them are continuing to upgrade the frontier models when they need very specific things done. But a lot of the CEOs at this point in time, they hit a level for these models, and that is basically good enough. And this, again, to be clear, this is completely normal. This is not being snarky, it's not being mean, this is not anti-AI, it's just how the shit works. I've talked about this with smartphones. Smartphones got to a certain level. We can say probably around the iPhone X or the iPhone X era, if you looked at the iPhone X, that was basically about as good as you need. Really, for iPhones, for smartphones, that's about as good as you need. Everything past that, it's an incremental improvement. You're most likely going to buy your next smartphone because you need a better battery, or maybe the bigger screen, or maybe because you dropped it and you broke it. You're probably not going to be buying your next smartphone because of some amazing feature it has. And that's just where we are in the smartphone world, right? And I think the same is true in the AI, the frontier model world, is you may buy the next models, but you don't really care that much. But the issue that these AI companies have, is they have simply made the bet that the value of this technology is only going to go up and up and up and up. They have signed contracts based off of that assumption. And I don't know, it's just not reality. Just not reality.

AI token prices are hitting new record lows from CNBC. And yeah, the price of tokens is just going down. A closely followed measure of AI token prices touched fresh lows this week. The latest signs of deflating prices in an increasingly competitive landscape. The LLM Token Expenditure Index, a key gauge of daily prices from intelligence firm Silicon Data, fell to 97 cents on Monday. That marked the index's lowest reading since its creation late last year, and has fallen by more than half from the high recorded earlier in the summer. So it's up about $2, and now it's gone down. So this is per million tokens, so we talk about this. So it's 97 cents per million tokens. And so that's one of the things, when you look at all the models, Jesus Christ, they have so many different prices out there. Right, so OpenAI, depending on what model you use, they have a different price per million tokens. Again, they're all the way down. One of their latest models is 20 cents per million tokens. Some of their higher level models, though, require more, might be $1.52 per million tokens. And so what this is trying to show is basically what is the average people are paying. So some people are paying 20 cents per million tokens for certain tasks. Other people are paying $2.50 per million tokens for other tasks. But when you average all of that out, where is the price going? Down, down, because that's the only place it was ever gonna go.