00:28:19 · Jordi Hays, John Coogan, S2
[00:28:19] S0: They're going whole hog. [00:28:20] S1: They're going whole hog. They're using everything everything they possibly can. Every lab is using every chip and every data center and every hyperscaler and every Neo Cloud. They're using it all for all different things, and different chips will be better at different models and different part... Pieces of the inference stack, and, like, everything will be used. Very good for the... Like, if you've been in semis and and, you know, Neo Clouds and all this stuff, rising tide will, like, lift all boats as everything becomes... Like, there was a moment where it was, like, oh, if you made the wrong architectural decision, like, you would get smoked. That was the theory. And now it's, like, oh, actually, these models, at least a piece of the models can be, like, used in niche areas for different things. And so even if even if one chip is better at something, faster, but that... What's less smarter smarter but slower, like like, there will be an optimal way to to get a model running on that. And with AI agents, it's pretty easy to port to different things. I I think we saw that from the AMD coverage we did with Lisa Su where Anthropic vibe coded their AMD implementation, like, pretty quickly. So good for anyone who's betting on oligopoly in silicon, rougher if you were like, this will not be a car like market, but still very, very early. Still a lot of dependence on NVIDIA, obviously. So nothing too scary there, but just interesting that everything is getting used by everyone. [00:29:44] S2: So Yeah.
00:29:44 · S2, John Coogan, Jordi Hays
[00:29:44] S2: So Yeah. [00:29:45] S1: That's kinda my update. Anyway, open source open source... [00:29:50] S0: Yeah. So Dax over at OpenCode posted yesterday. He says, to put what I'm saying about demand in context, OpenRouter is the most successful model model router. You can almost serve a 100% of their deep sea traffic yesterday on a single GB 300 rack. This is worth 6 to 7,000,000. Even if there's 10x this demand, secretly, that's still less than a 100,000,000 of compute. [00:30:14] S1: This is crazy. This is crazy because I don't believe... I mean, I don't know. Maybe people are inferencing DeepSeek from Neo Clouds directly and not going through Open Router, but I would be very skeptical that there are large enterprises out there that are saying, oh, yeah. We have a million dollars a day of inference. Yes. A lot of [00:30:36] S0: people are saying, well, this doesn't... Yeah. This doesn't factor in people that have [00:30:40] S1: Their own CPUs on prem. [00:30:41] S0: Yep. But the the elephant in the room is the the base tens, the Yep. The models, the fireworks, the together AI. [00:30:50] S1: So it's going through that and it's not going on [00:30:51] S0: at Lambda's, you know, like a very very long list of Okay. Of other clouds that can [00:30:56] S1: Okay. [00:30:57] S0: That can serve these models. So I would expect that that is where that is where the that's where the billions of dollars of of inference spend is happening.
00:30:57 · Jordi Hays, John Coogan, Ben Liu
[00:30:57] S0: That can serve these models. So I would expect that that is where that is where the that's where the billions of dollars of of inference spend is happening. [00:31:06] S1: Sure. Do we have a do we have a reference point for what percent of DeepSeek inference is going through OpenRouter versus going direct? [00:31:14] S0: Oh, I think it's impossible to know. [00:31:16] S1: Okay. Well, if you wanna get in the game, there's a 300 rack AI supercomputer powered by 9,600 NVIDIA a 100 GPUs. It's on eBay. You can just buy it [00:31:27] S0: on eBay. If [00:31:28] S1: you were thinking about starting a Neo Cloud, getting in the token selling business, hit up your local banker, get a credit line for 60 mil, go on eBay, click it. Fair market value for this? [00:31:42] S0: EBay up 16% over the last six months likely because people are realizing that this... It's a compute marketplace. [00:31:49] S1: Apparently. Yeah. Fair market value for this would be closer to 50, maybe 55,000,000, but you could still [00:31:56] S0: I love that. The screenshot. It says in six carts. [00:32:00] S1: In six [00:32:01] S0: carts? Like six Only people [00:32:02] S6: have added it.
00:32:02 · Ben Liu, John Coogan, Jordi Hays, S2
[00:32:02] S6: have added it. [00:32:03] S1: Added to cart. Two available. Wait. They have two available of these? Wow. Yeah. Interesting. Pretty pretty solid depreciation too. This this hardware originally sold for something like, what was what was this originally originally went for? Let me let me find it. It's selling for 60,000,000 now. The original hardware was almost 300,000,000. So there is depreciation, but there's still some value there. Of course, there's... The a 100 is a seven year old chip now, something like that. And so you're banging the table. You're you're you're hitting by? Are you bidding? [00:32:47] S0: I got... I was getting excited. [00:32:48] S1: Oh, you're just getting excited. [00:32:49] S0: Okay. No. Very old chip. Still worth a lot of money. [00:32:53] S1: Yeah. Totally. [00:32:54] S2: It's the price of one dinosaur. [00:32:56] S1: Oh, yeah. Which one would you do? Jordi, which one are you going with? [00:33:00] S2: Are you going Gus, the the T Rex? T Rex? Or are going these g p... [00:33:03] S0: 20... That's that's the 20... Think I think I'm still, to to my oldest dismay, I'm still going I'm going for the GPUs. [00:33:12] S1: You're going for the GPUs? [00:33:13] S0: Just because I think Malibu Inference would be Really good. A great name [00:33:17] S1: Yep. [00:33:17] S0: For a a Neo Cloud. [00:33:19] S1: Yeah. And everyone's been making that claim. Oh, like, put the data centers in, like, luxury towns. And you're like, yeah, got one right here. [00:33:26] S0: Yeah. Also, you could Literally in my backyard.
00:33:26 · Jordi Hays, John Coogan
[00:33:26] S0: Yeah. Also, you could Literally in my backyard. [00:33:28] S1: GPUs up in the shape of a dinosaur Yes. So that you have a beautiful architecture in your foyer. You'd of course need to add a lot of cooling. Because all of a sudden your entryway is the hot aisle. That'd be rough. But it would make a statement. People would come in, they know you mean business. They know they know you do [00:33:47] S0: That's right. [00:33:48] S1: You do local compute. You're not paying API pricing, you know. You're saving dollars every... [00:33:54] S0: Why on earth does Malaysia have so much compute, John? Can you take [00:33:59] S1: a Nobody knows. [00:33:59] S0: It's a mystery. Nobody knows. [00:34:00] S1: Nobody knows. [00:34:01] S0: Nobody knows. It is a total mystery. [00:34:03] S1: This was the first story we talked about on this show. The very first show. Wow. Very first show I opened up The Wall Street Journal and and there was an article right on the cover about Malaysia being the new the new data center hub. [00:34:18] S0: Tyler, we gotta send you down there.
00:34:18 · Jordi Hays, John Coogan
[00:34:18] S0: Tyler, we gotta send you down there. [00:34:19] S1: They are the Switzerland of AI. That was sort of the framing was that they can buy compute from The US and resell it in China. And so that was an incredible advantage. And they had a lot of power between the different locations there, and the government was sort of set up to allow a lot of compute build out. Pretty pretty interesting. They they have a lot. It's not that much compared to America. I mean, is not the green bar here. They're not the red bar. The red bar is China. I'm actually surprised They're not even the blue bar. [00:34:54] S0: I'm I'm surprised that the EU has more compute than Malaysia. Okay. Yeah. You're back. Aren't you? [00:34:59] S1: No. Because the EU is huge, and Malaysia is like a pretty small country. [00:35:04] S0: I know. But, yes. They they... Malaysia seems to be going so long.