A compilation of all discussions about OpenRouter from the More or Less podcast, featuring insights on AI model routing, API services, and the evolution of LLM infrastructure.
The More or Less podcast hosts discuss OpenRouter repeatedly across multiple episodes, examining its role as an AI model routing service that helps companies optimize costs by directing tasks to cheaper, open-source models instead of expensive frontier models. The conversations explore OpenRouter's rapid growth, its partnership with Stripe for payments, and the broader question of whether routing services will become commoditized or develop defensible moats through smart logic and enterprise features.
Sam Lessin: I think this is like the really interesting place we're in right now, 'cause like, look, I wa- I mean, even this morning on CNBC we were talking about this 'cause, you know. And I was like, "Look, here's the deal." Right now, Anthropic tokens cost like 100 times as much as like the open source tokens, right? Like just i- in terms of what they g- And like the open source stuff is catching up, and Open Router and whatever, like obviously you don't need a PhD to do your IQ 100 work, right? So like obviously the future is not gonna be like you just send all of your stuff to the most expensive model, right? So this stuff should rationalize really quickly, and the fact that it's so frictionless means that it should rationalize financially way faster than like a normal market with more friction in it, right? So like naturally speaking, like we're just multiplying big numbers and like it's fine. The only caveat on that for a more or less conversation is ironically the fact that there's no dem- demonstrable, measurable ROI is in some ways this weird moat for a while. 'Cause you're like, "Well, I can pay a hundredth as much, but I can't actually measure the value of this token versus that token in, from an ROI perspective." So the fact that it's so opaque in some ways like makes it kind of defensible for a while, right? Like 'cause I can't actually tell the, you know, the, uh, like what the gap is for an average person. So this is a weird market where like the opacity of quality, it creates defensibility sort of. Does that make sense? Like if you actually had an advertising ROI function, like think about the ads market, which is totally the opposite at this point. It's incredibly transparent. You can literally just look at the ROI and it's perfectly efficient at this point, and so you're gonna buy Meta ads when they're more efficient. And the second they're not worth their extra cost because they're tar- like be, from an ROI perspective, you can run shitty other ads. Like it doesn't ma- It's like a efficiency thing. Whereas with the tokens, you're like you ask for an intelligence task and it's like hard enough to benchmark what you're getting, right, without redoing the work twice in a lot of ways all the time, that you're like, "Well, is there some defensibility in the lack of measurability?"
Sam Lessin: I know, but are they doing it?
Jessica: I think so, but like not in like... I, I think it's hard to measure is the honesty. I w- I will say that like OpenRouter, which is probably the poster child-
Sam Lessin: Yeah
Jessica: ... for this, is growing like crazy, and the valuation's way up and is looking good. There's companies like DigitalOcean. Talk about another... Sam, I'm so bad at trading stocks. It's like wild how bad I am at trading stocks. You have been talking about DigitalOcean for like eight months, and how like it's the core engine I use and whatever else.
Brit Morin: I can give you the exact rundown.
Dave Morin: That's an under.
Jessica: So a lot of companies, Dave, u- like OpenRouter uses Stripe for payments. So there's a lot of-
Sam Lessin: Everyone does, 'cause it's... Look, the, the negative on s- here, let me give you the, the story I ha- this is a very interesting take someone made. They're like, "Stripe really fucked up," right, was the story that someone was kind of giving me recently. They're like-
Brit Morin: Wait, what acquisition?
Jessica: No, this is, Sam is m- mixing things up. But the Information story currently on The Information website is what people should reference on this point. But I think there's a lot of interesting deals they could do, and they have interesting partnerships in AI, like a partnership with OpenRouter that we talk about in the article on the website.
Sam Lessin: But again, they're a 2026 much smaller company than they could be, and the question is whether that's a fuck-up or not. And like, the interesting problem with payments, right, is historically they... I mean, it's great. Like, anyone who's anyone will just use them out of the box 'cause it's so easy, right? Certainly my bot uses them for anything it charges people for, because it's so easy. And like, they, they made... That was the win there, which was so smart, and they just coasted on internet growth. The problem is once you get really big, they charge a lot. And like, payment processing's a low margin business, and so you start, once you're big enough, you just push them down on margin so much that, like, it's still okay, but, like, it's a hard box to be in.
Jessica: I agree with that.
Dave Morin: Like, Jess, you said they have a big partnership with OpenRouter. Is that growth that they're reporting, like, coming from OpenRouter?
Sam Lessin: OpenRouter's not that big.
Jessica: Uh, part, part of it... I, no, but it's an example of how AI is boosting their payments processing business. Um-
Dave Morin: I see
Jessica: ... according to Yuchi, our awesome crypto and fintech reporter.
Dave Morin: I think the other thing here, you guys, is that though there is a narrative about agents and money out there, agents and money and payments is not even remotely a thing yet.
sam lessin: And that becomes their defensibility.
Sam Lessin: Okay. Can I ask, can I pivot the conversation a little bit, unless people wanna go further on it? There's another tack I'm curious people's take on, which is open router. Like, talk a little bit about enterprise and things like that. If we're saying that your data is so valuable, right? And but at the same time, enterprise really cares... Oh my goodness, a guest visitor, Jess Alessi.
Rachel: Yeah. It does.
Peter: I mean, there's a bits of data that I think, in some ways, Sam, I think you and I saw this at Facebook. It's like, what did Facebook do? It just digitized the information that was in our head. Who's your friend? What did you do last week? Yeah.
Sam Lessin: Open router. Routing. Everyone wants cheaper stuff. AI, the, the m- matrix multiplication commodity. There's no reason to pay a PhD to fold your laundry. But you also don't wanna give Open Router all your data. Or do you? Or do you not care? Like, how does that play out?
Guest: I, I do think that the bigger question here, which is gonna also bifurcate, it's not gonna be one takes all, is open versus closed. And, um, I am saying this obviously with a bias, as I've said in the past, that where you stand on the issue depends on where you sit in the cap table, and we're large investors in Hugging Face, which made a lot of news over the past few weeks. But I think that the future is for the more sophisticated users who serve for the enterprise, it's gonna be your longitudinal proprietary repository silo of data that is yours, that is, exactly as we were just talking about, uh, not ingested by the machines. And if you have that and can run open source models on that time series of data that is exclusively yours, whether you're a pharma company, a finance company, insurance company, a retail company, a venture firm, increasingly you are realizing that to get the benefit, you're giving a benefit, and you're gonna say, "I'm not gonna give that data," especially if I'm not gonna give that data and watch as one of the closed labs competes me away. And you saw that first with Figma, which was sort of scary, and I think the push now from Anthropic and OpenAI to go into pharma and pharma companies saying, "Whoa, whoa, do we wanna do this? Do we wanna give them our data and then potentially open up competitors?" And I think there's a big move right now for them to try to ingest as much information from as many non-internet sources that they can, and I think that there's gonna be a movement underway, whether it is industrial, manufacturing, defense, pharma, biotech, and all of our personal data, that we're gonna say, "You know what? The models are good enough, they are performative enough, and I'm gonna retain the sovereignty of that in the same way that Sam does, and I'm not giving it over." So I do think that that's where the value's gonna accrue.
Scott: Sounds like she's got a better agent than yours.
scott: Lockdown data. Lockdown data. So routing. Let's talk routing.
sam lessin: He can Alright, so I wanna go back to open router. Open router. Open router. Routing. Everyone wants cheaper stuff. AI the the mate multiplication commodity. There's no reason to pay a PhD to d to fold your laundry. But you also don't want to give open router all your data. Or do you? Or do you not care? Or like how does that play out?
Josh Wolfe (Lux): Sam tier tier.
Guest: I don't think anyone wants to, like, share their company data with a company. Yeah, I, I do know a number of startups who were kind of using Claude individual accounts and, you know, using their ramp card type stuff, and then now have to, for whatever reason 'cause of scale, go to the enterprise plan, and they realize that it's, like, 10X more expensive. And then they're realizing that they're using a lot of tokens for things that are pretty mundane, and then someone on the engineering team is like, "I think we can do this with this, like, open model, but we can... As Josh said, there's no risk. It's, like, local. We can mess with the weights and do whatever we need." And so they're now, the next question is, okay, what router should we use? And it seems to me like there's a lot of router options out there, and it's gonna be probably pretty commoditized pretty quickly. And so you're gonna all have routers that direct all sorts of spend to lo- more local, more efficient models. Like, this is gonna become the default. My question is actually, like, what more opportunities do routers have to not become commoditized? Like, I thought you were gonna go in that direction. Like, routers can start to ca- carry logic around which models you should use. Why-
scott: anyone wants like share their company data with a company that but I I do I I do know a number of startups who are kind of using Claude individual accounts and you know using their ramp card type stuff and then now have to for whatever reason because of scale go to the enterprise plan and they realize that it's like 10x more expensive and then they're realizing that they're using a lot of tokens for things that are pretty mundane and then someone on the engineering team is like I think we can do this with this like open model that we can, as Josh said, there's no risk. It's like local. We can mess with the weights and do whatever we need. And so there are now, and then there the next question is okay, what router should we use? And it seems to me like there's a lot of router options out there, and it's going to be probably pretty commoditized pretty quickly. And and and so you're going to all have routers that direct all sorts of spend to look more local, more efficient models. Like this is going to become the default. My question is actually like what more opportunities do routers have to not become commoditized? Like I thought you were gonna go in that direction. Like routers can start to ha carry logic around which models you should use.
Sam Lessin: So like a Series A.
Scott: Yeah, exactly. Your next Series A. No, they put, like, 8 billion, which was a big number back in 2001, and they put it on their balance sheet, right? So they basically did vendor financing. They said, "Okay, everybody's gonna need fiber." And they were right. What they were wrong, which is exactly what you were just talking about with Open Router and some of the other developments, is the whole concept of wave division multiplexing. Nobody baked that in, and all of a sudden you get 100X improvement on fiber. And so your demand and your pricing fall through the floor. Let's just assume for a second it doesn't fall through the floor, and we've got a half a trillion dollar financing facility thanks to the six amigos that lined up to lend their balance sheet. None of this goes back to NVIDIA. This is why Jensen's brilliant. He basically set this up as SPVs that will basically borrow against the take or pay agreements of the off-takers, of the Neo cloud guys that are saying, "We're all gonna need $3,000 a month for our personal agents." And they are signing take or pay agreements. So independent of whether or not, Sam, you keep your $3,000 agent alive, they still have to pay for the demand. So for a couple years-
Scott: That's a problem
Guest: ... that's the problem. Yeah.
Scott: Right. So they basically did vendor financing. They said, okay, everybody's gonna need fiber. and they were right. What what they were wrong, which is exactly what you were just talking about with open router and some of the other developments, is the whole concept of wave division multiplexing. Nobody baked that in. And all of a sudden you get a hundred at a hundred X improvement on Fiverr. And so your your demand and your pricing fall through the floor. So Let's just assume for a second it doesn't fall through the floor and we've got a half a trillion dollar financing facility thanks to the six amigos that lined up to lend their balance sheet. None of this goes back to NVIDIA. This is why Jensen's brilliant. He basically set this up as SPVs that will basically borrow against the take or pay agreements of the off takers, of of the neo cloud guys that. are saying we're all gonna need three thousand dollars a month for our personal agents. And they are signing take or pay agreements. So independent of whether or not, Sam, you keep your three thousand dollar agent alive, they still have to pay for the demand. So for a couple years, the the
Rachel Holt: Yeah, I mean, they're assuming, right, these are gonna act like any asset that's gonna spit off cash for a extremely long duration, and, like, this isn't a toll road. You know? This isn't something that, like, inherently is gonna have value 30 years from now. And I think that's, to Josh's point, like, the fundamental risk and question is in that, you know, how much money can you spit off versus the duration that these are useful for.