Comment by tyre
4 hours ago
To people asking why, this is a good lesson on the Collison’s ambitions. Stripe is one of the best API companies in the world. They know how to serve high volumes of latency and availability sensitive requests. They’ve abstracted the financial rails for payments and now want to abstract the rails for LLMs.
They’re the perfect company to own OpenRouter.
Tokens are simply a lightweight valuable asset. Stripe can serve as the middleman as well as anyone. They know how to route to many providers (payment rails) with huge differences in service characteristics. LLM providers are far easier.
Then they can work this into an offering where users can subscribe to tokens and use them across services. It solves one of the core monetization challenges of every AI company: how do you price when your costs are variable on usage, but nobody can make sense of charging by token?
From here, they can start hosting their own models and competing as an AWS for tokens. They can be the best provider of $OPEN_MODEL, or their own, and optimize for you.
You'd be surprised how little volume of api calls payments companies get compared to advertising companies or consumer internet companies. Stripe aren't especially well placed to own this asset on a technical level. That doesn't make this a bad acquisition per se, but there is no magic stripe can inject here as far as infra/technology is concerned.
Agree, even in the payments space usage-billing metrics providers like Chargebee / Lago etc get order(s) of magnitude more data than payment.
Granted payments and associated ledges require a level of consistency that other systems do not require, but that is orthogonal to high volume.
You'd be surprised how antiquainted many of them are. I don't mean Stripe or Braintree but the long tail of terrible terrible credit card processors. If I hear "Chase Payment Gateway" I shudder.
Do people feel Stripe APIs are really that good? I’ve always found them unintuitive, tons of options, weird edge cases, breaking occasionally, etc.
I’m sure payments are convoluted, but I’d still imagine they could be meaningfully easier for the bulk 80% of use case?
Yes, Stripe is extremely good at what they do. Payments are complicated.
Payments themselves are so much complicated. Payments to support multi-currency, jurisdictions, countries, scenarios, that's complicated.
1 reply →
I have implemented the Swift payment Spec and I can tell you having this abstracted in a nice API is so much better.
Stripe's APIs are very good. They could be better, but it's difficult to change the underpinnings of financial infrastructure. A lot of risk, little incentive to move, and you're designing something that has to last (and make sense) for 15+ years in the future.
I worked on this at Stripe in 2022. We were the first teams to start building v2 APIs and data models to solve exactly this problem. The first target launch date (in Feb 2022) was November 2022. It was launched in May of 2025.
My reporting line, as an EM, was Netflix, Oracle, Oracle. No one had startup experience. It was drenched in politics. The engineers were largely brilliant, kind, and hardworking.
I still love the company and believe in Patrick. Believe me, he deeply understands what you're saying and wants it to be the best it can be. But it was clear to me, even then, that they'd lost a lot of what made them special. They could maintain it, but I wasn't sure they could do it again. Banking-as-a-Service was one opportunity, Link was another, and now this will be a third. We'll see. (I say this with a lot of love for Stripe and Stripes.)
Coincidentally, I had a conversation with a recruiter at Anthropic and saw them doing something very similar. They were starting a new team in a new vertical and wanted someone with experience running an org of 100+ people. I would bet real money that it will be a fraction of the product/impact it could be (though still probably make money!)
I expect many of the warts in the API surface are there because they want to avoid changing things for fear of breaking someone else's business.
I mean. They used to be good. When they started out their docs were absolutely far above the competition. They aren’t now, and Stripe has lived long enough that they’ve seen themselves become the villain.
Feature bloat definitely made their API and product docs harder than the early days. Stripe Checkout would have saved my teams days, but I agree, they need to go back to their roots and make the 99% use case very easy.
Stripe has the worst payment APIs, except for all the others.
> Tokens are simply a lightweight valuable asset
(Not picking on you here, you just provided a well-written peg for a popular narrative. I'm aiming to sharpen my own thinking here & perhaps learn something.)
This seems like a stretch given the rise of local inference, especially the Prism Labs rumors from a few weeks ago.
One way I think about LLMs is they are akin to fancy databases in that they are software of which you can ask questions and get answers if you ask properly. Oracle & SQL Server are akin to OpenAI and Anthropic, and there are analogues for MySQL, PostgreSQL, SQLite, MongoDB, PlanetScale, etc. (This is an analogy, it's not going to be a perfect fit.)
In that view, would it make sense for someone to say that their credit card processor just bought the company that makes their ODBC driver? Would anybody suggest that the TPS of their RDS instance is a lightweight asset?
I don't see this as strategic beyond the obvious idea that Stripe wants to get closer to AI, and they haven't been able to get the market to care about their natural linkage to AI (Radar).
> they can start hosting their own models and competing as an AWS for tokens
The financial pressure of doing this has caused cuts to core product teams at the richest companies (which Stripe isn't!). I would not want my payments processor to go down this road and to get worse at processing payments.
Your simplification itself is compressing away the meaningful point, and then you're confused about your own over-simplification.
Stripe is not a "credit card processor", and that's not the point of OP's comment. The point is that being the intermediary between merchants and processors is the valuable expertise in terms of an OpenRouter acquisition.
It's been a few years, but I used to work on almost exactly this (not at Stripe, but we processed hundreds of billions annually). If you're a big enough merchant (or Stripe itself), you can choose to send a credit card transaction to any one of potentially many processors, each of which have their own performance characteristics (one may approve at a higher rate, the other may charge better fees, etc). All of this subject to attributes of the transaction itself (ticket size, geography, card type, many others). You know quite a bit about the transaction itself before you send it out, so you can build up routing knowledge to optimize for whatever thing you care about (usually transaction success rates or fees).
See how this starts to look a lot like OpenRouter with money instead of tokens? I'm not sure I 100% believe that's how it'll shake out, but there is a transferable skillset.
I hear your argument, I do.
I think the "money instead of tokens" is the important part. Money and tokens are fundamentally different was part of my argument (that I did not make well). It's not clear that it is a good thing for money movers to get into the token business (this apples to Ramp as well).
Banks are also intermediaries between parties (at scale, really between any parties). One could use similar logic to say that it therefore makes sense for Wells Fargo to start buying homebuilders because the home buyers will eventually be paying Wells anyway.
BankRate similarly processes volume of consumer mortgage quotes in real-time communication with lender APIs. They also do not have a reason to own a token router, even though their business involves similar processes to what you outline for Stripe.
Being an intermediary in a value chain does not mean you're critical path, or that it makes sense for you to be in the critical path.
Funnily enough, you argue that this is a natural fit for Stripe while a peer reply argues that it's a change in the business a la Amazon->AWS.
Anyway, I appreciate your thoughts.
> would it make sense for someone to say that their credit card processor just bought the company that makes their ODBC driver? Would anybody suggest that the TPS of their RDS instance is a lightweight asset?
Would is make sense to say that their online bookstore now sells Ethernet cables, bidets, and delivers groceries? Sells _cloud infrastructure_??
Amazon made two transitions:
1. Amazon the online bookseller => Amazon "The Everything Store"
2. Amazon the online retailer => Amazon the Cloud Services company
I believe what Stripe is doing here is closer to (1). "We are good at high-throughput APIs that wrap complexity with thin margins. We did it for credit cards, then ~all payment methods, now other digital bits."
In this context, tokens are much, much easier than international payments.
Now: Is it outlandish for their customers?
Not really. Every engineer knows stripe as an engineering company. I don't think Stripe is what it once was, but it's certainly a generational company. You're asking engineers—who broadly have a positive impression of Stripe—to use this product they already know. OpenRouter gets the positive brand association (trust) of Stripe; Stripe expands into a new domain whose technical needs are extremely similar.
Appreciate your thoughts.
re: the Amazon transitions, the first shareholder letter lays out the plan to go beyond books. As I remember it, books were always only supposed to be the entry point.
AWS transition was more around building the platform Amazon.com needed to grow, and also to monetize the same platform.
Neither of these really fit with Stripe.
Stripe can obviously operate OpenRouter, they have the tech skills. The risk to the core business is that OpenRouter's growth path will distract from the core financial business and/or require a very different capital stack. (Someone already suggested Stripe scale out first-party model running, which can get very expensive.)
Appreciate your taking the time to respond.
Ok, at first I didn’t get it and thought it would make sense if stripe just wants to build payments for agents, but on a second thought your idea about “once buy tokens — use everywhere” is very good one!
why did they miss the market that is OpenRouter?
You're asking how someone missed llms. On this website? That demonized them for years and is only now fighting the cultural battle to accept them as useful?
unless it's all just tulip mania, then they just spent $7B on nothing
Here are some things that are abundantly clear to me.
1.) LLMs are useful for programming
2.) Open models are excellent and will continue to improve
3.) Economies of scale and ease of access mean self hosting is out of the question for a large number of users
This means that even if the largest labs are not worth trillions and a large amount of the data center build out is not as valuable as the builders project and GPU/RAM prices plummet, it will not matter at all for this business. People will want to buy cheap open source tokens from a centralized trusted provider.
$7 billion for a business with little overhead that is already within their core competency and has strategic growth potential seems like a very good deal.
What does openrouter actually do though? They buy a bunch of tokens from token providers and allow you to easily route queries between token processors?
This seems like the kind of thing that you can have Claude write in an afternoon for whatever service you're running. I don't see the value.
1 reply →
Unfortunately we don't really know anything about OpenRouter or Stripe's numbers. They're both private companies. AFAIK there really isn't much information publicly available about how profitable or not OpenRouter actually is, or whether it has a little or a lot of overhead. So I can't really say whether $7B is reasonable or inflated. It seems like an enormous number, and my hunch is like all the other enormous numbers floating around in the AI space it's probably inflated by the hype bubble. But we just don't know.
Broadly I agree with you. It seems like they have an in-demand product and there could be a sustainable business there at least in principle. But whether it's a $7B business or a $70M business I can't say.
depends what the tulips are. tokens? private growth stage shares?
openrouter is involved with a lot of scammy crypto personalities. maybe they are the tulips people have been manic about for too long. maybe the collisons are the tulips.
[dead]