Big Technology Podcast
00:48:27 · Ranjan Roy, Alex Kantrowitz
[00:48:27] S1: Well then [00:48:28] S0: You're it's gonna potentially wipe us all out and you have to develop it because you can develop it safely. But then you're also walking away with some of like the richest people in history. [00:48:37] S1: No, but that's what I've always said. And that's actually, again, why politicians, you got four days right now. This is a pretty easy talking point and you might as well push it but I also actually on the existential risks to humanity, I also have to admit I actually wasn't that interested in the whole risk factor side of it. I know I saw you tweeting them out and you were discussing them with Dan, but, like [00:49:04] S0: Yeah. [00:49:04] S1: To me, they were... I don't know. I... They almost just feel like they're, like, just fucking with us. Like [00:49:09] S0: Dude, I have a conspiracy theory on that. Alright. They had the longest risk factor section I've ever seen in my whole life. I tweeted 80 of them, and then I was like... I tweeted one post with 80 of them. And the funny thing is, I stopped at 80. I don't even remember how many there were because there were so many. And I think what they did was they wanted to bury the existential risk factor in the most boilerplate shit imaginable so you would miss that and just be like, Oh, well, they have all these other capital risks.
00:49:09 · Alex Kantrowitz, Ranjan Roy
[00:49:09] S0: Dude, I have a conspiracy theory on that. Alright. They had the longest risk factor section I've ever seen in my whole life. I tweeted 80 of them, and then I was like... I tweeted one post with 80 of them. And the funny thing is, I stopped at 80. I don't even remember how many there were because there were so many. And I think what they did was they wanted to bury the existential risk factor in the most boilerplate shit imaginable so you would miss that and just be like, Oh, well, they have all these other capital risks. [00:49:39] S1: But we know the existential risk and everyone covered the existential risk. I guess to me that was like, normally S1 risk factors, the whole point is to recognize risks that exist that especially likely disclose things that investors don't already understand and legally also that you are recognizing them. So, I mean, it has to be in there, but he's been saying it for years. They've all
00:50:06 · Alex Kantrowitz
[00:50:06] S0: been No, just saying it was so long that they were just like, All right, you'll get exhausted and no one will pay it. Well, anyway, I think you're right that they have... It's been talked to death. All right, here's a risk factor for us to talk about before we leave today, which is the price... Know, we talked about a price war coming for LLMs and man has it happened. This is from Max Wolfe. OpenAI Anthropic cut their LLM prices by up to 80% within a few months. Why? I'll just read a little bit for Max. 07/30/2026, OpenAI announces that they're going to cut GPT 5.6 Luna by 80%. The price is 08/10/2026: Anthropic says their temporary Sonnet five price cut is permanent. 08/21/2026: OpenAI announces they are cutting the price of GPT-5.6 SOL by 20%. September 2026: OpenAI announced GPT-six SOL at a shocking price that halves the input and output price relative to the temporary discounted GPT-5.6 SOL. They simultaneously released GPT-six Luna also at half the input and output prices of the previous version. On the same exact day Anthropic announced Claude Opus 5.5 and the total cost reduction for that max calculates was 40 to 60% for a model that's even better than Fable five. Is not just a price war. Is not just the sorry, this is not the rumor of the price war. This is the price war. Some reasons for Max. He says, If you've been testing agents with new LLMs as they are released, it can be very difficult to differentiate them outside of subjective vibes. Since last November that launched the hypergrowth of Vagintic LLMs, the models have become more of a commodity, and it's now trivial to try a new model on your code base with just a toggle harness like OpenCode or using something like OpenRouter. As a result, there is little moat. That's just part of the reason that he gives. We talked a lot about the commoditization of LLMs and how that could lead to a price war. Is this it actually happening, Ranjan?
00:52:09 · Ranjan Roy, Alex Kantrowitz
[00:52:09] S1: I mean, it is... That... That's the other part again. Like, the the moment we're in in the next few months, we're already seeing the actual like cost. Yeah. I mean, we were... We've talked about the cost of compute. We've talked about the actual ability of them to charge, but also, I guess the... If the aggregate demand increases and everyone is identifying everything, this does become moot. And I think that's what they're gonna have to kinda pitch on the story. But I agree that this is... This doesn't help the overall business story that like say frontier models at high margin are going to be what differentiates them. And the main thing, I don't... I still don't like... Again, in the enterprise side of it, the number of people talking about or working on routing models across whatever the lowest cost best model is for the task is definitely increasing. So so I think it's still... It certainly presents a threat, but it hasn't affected them yet, it looks like. [00:53:09] S0: Another potential interesting reason is infrastructure. Here he writes this: Another less cynical factor that could explain the price cuts is that LLM infrastructure has improved over the past few months. When Anthropic said OPUS 5.5 requires less compute to serve than OPUS five and its pricing reflects that. I believe it. One outcome that would be genuinely ironic is that if these models are more compute efficient, it reduces the need for these massive data center build outs, which could cause cascading economic failures. Interesting. Your thoughts.
00:53:09 · Alex Kantrowitz, Ranjan Roy
[00:53:09] S0: Another potential interesting reason is infrastructure. Here he writes this: Another less cynical factor that could explain the price cuts is that LLM infrastructure has improved over the past few months. When Anthropic said OPUS 5.5 requires less compute to serve than OPUS five and its pricing reflects that. I believe it. One outcome that would be genuinely ironic is that if these models are more compute efficient, it reduces the need for these massive data center build outs, which could cause cascading economic failures. Interesting. Your thoughts. [00:53:41] S1: Yeah. I mean, it's a great point because right now, the... It feels like the... We just said the entire first part of this podcast was around big tech and data center investment. And like the economics of that, no one has any idea. And if it's... If everything becomes far more compute efficient, which it looks like it is becoming, that fundamentally changes the entire data center story. If the whole, again, the whole world gets identified, maybe it all works out but I think all of this is still, there's just so much risk and clearly the market with big technology firms has already recognized it but it is a bit terrifying, cascade economic failures.
00:53:41 · Ranjan Roy, Alex Kantrowitz
[00:53:41] S1: Yeah. I mean, it's a great point because right now, the... It feels like the... We just said the entire first part of this podcast was around big tech and data center investment. And like the economics of that, no one has any idea. And if it's... If everything becomes far more compute efficient, which it looks like it is becoming, that fundamentally changes the entire data center story. If the whole, again, the whole world gets identified, maybe it all works out but I think all of this is still, there's just so much risk and clearly the market with big technology firms has already recognized it but it is a bit terrifying, cascade economic failures. [00:54:26] S0: Don't like the sound of that at all. Mean the other, the final explanation is like the IPO was coming and they want to show usage increases, but like I know they also they really want to show revenue increases. So I don't know. Sometimes companies try to show real growth ahead of the IPO, but I don't think tokens is the growth that Wall Street carries. [00:54:46] S1: No, It's revenue and low costs and mask the true costs as much as possible in a GAAP compliant way. [00:54:54] S0: Okay. [00:54:55] S1: She's gotta do. [00:54:56] S0: Couple things before we leave. First of all, I love that you said Cheesecake Factory earlier in this episode because our SEO is now gonna be off the charts giving. [00:55:05] S1: But no one actually searches anymore, so it doesn't matter. [00:55:09] S0: Oh, that's true. Okay, you got me. You got me there. Do you want to spend a couple minutes on you can see everything?