Comment by Gecko4072
5 hours ago
How can a middle man for api calls be worth so much? Their market share can’t be very large right? For comparison, $7B is more than market cap of Lyft, Dolby, and Alaska Airlines. What is happening?
5 hours ago
How can a middle man for api calls be worth so much? Their market share can’t be very large right? For comparison, $7B is more than market cap of Lyft, Dolby, and Alaska Airlines. What is happening?
It's an inflated number due to the AI market, but their selling point is distribution.
Many people and businesses want to experiment with different models, but they don't want to sign up for a dozen different services. Businesses can make it difficult to approve new vendors. If your company is looking at 5 different vendors for tokens and teams can't agree to switch together, OpenRouter comes along with a unified interface and a single billing point.
They also become the point to add value-add services on top in a portable way. They already offer some things like automatic JSON repair, but I can see them adding functionality like leak detection tools, monitoring, alerts, and other patterns that a company can set up once and use with all of the models theirs teams need.
What I'm not so sure about is their moat. They have the brand recognition, but it seems rather easy for someone else to build what they've built. I'm a little confused about why Stripe didn't just build the same thing internally. Acquiring this company gives them an instant boost of 10 million customers for their AI business, which might be key to some financial goal they've got.
>It's an inflated number due to the AI market, but their selling point is distribution.
>Many people and businesses want to experiment with different models, but they don't want to sign up for a dozen different services. Businesses can make it difficult to approve new vendors. If your company is looking at 5 different vendors for tokens and teams can't agree to switch together, OpenRouter comes along with a unified interface and a single billing point.
Isn't that ripe for being picked off by aws or azure? Both already have marketplaces where vendors can offer whatever cloud services they want. Both already offer first party inference service, and have contracts with all the stodgy corps where it's "difficult to approve new vendors". Not to mention they have IAM and SSO built in. Good luck bolting that onto a third party vendor like openrouter.
I can actually restrict how much is spent using a sensible user interface with openrouter. That’s a real pain with aws
1 reply →
Yeah, but at least currently, openrouter has the latest open source options way faster than AWS does.
As a nobody I like OpenRouter because it lets me quickly test out all of the emerging open weight models. I normally use Claude Code but if I run out of usage I'll pull up GLM 5.2 or Deepseek V4 in OpenCode via OpenRouter. Occasionally I'll run Claude's work by a Google/OpenAI model through OpenRouter. If Claude Code becomes too expensive I now know I'll be pretty comfortable with open weight models.
I'd suspect it's related to the rise of good/cheap Chinese models, and OpenRouter is the best way to use them without jumping through a ton of hoops.
Exactly , not even small startup anywhere in us , eu or nearby will be ok to send data to China, but with open router you can create a no China provider selector with one llm call and start using Chinese llms without paying to vendors a single cent .
What’s the angle for stripe , electrify over tokens exchange is the new money flow , and stripe wants to monetize it. 5% tax on any llm token is an amazing deal
> but with open router you can create a no China provider selector
why on earth would you trust no data is sent to China? just because you clicked a toggle?
6 replies →
I prefer using OpenRouter because the money I pay them in premium is far less than the money I lose by not using it on other providers. On OpenRouter, I can switch dollars between models. But for the providers, after evaluating I'm stuck with some number of credits on ones I don't use. This kind of intermediation is actually super useful to me. I must imagine that they can also provide other things that are of value like provider quality measures, ability to swap providers during downtime, etc.
You can't really compare market caps like that. Addressable market or market share is only one piece of the puzzle, and the other companies you mention are in very different kinds of markets, with very different kinds of products, with very different kinds of costs and margins.
A valuation just reflects what someone thinks about the future cash flows of the business.
But yeah, it does feel a bit crazy; unsurprisingly, AI hype affects valuations of AI companies too. On the other hand, I can see the idea that some people might be betting on the idea that the big US labs are bloated and spend too much money, and that the real money is going to be in serving open-weight models, and/or in automatically combining and routing to different models based on the task at hand.
LLM traces are supposedly very valuable. I imagine OpenRouter has one of the most extensive and diverse set of traces in the world.
openrouter prompt and i/o logging is off by default https://openrouter.ai/docs/guides/privacy/data-collection
the providers you route to will of course have their own policies, which openrouter surfaces to you through the webui and api. you can even configure automatic routing to select providers based on your policy preferences.
Oh that's very interesting. Thanks for the link.
If your data is leaking to OpenRouter, how many companies would be OK with that? So I doubt OpenRouter are peeking inside the workloads that flow through their proxy. However, they do publish "Top Models by Task" ranking[1] so maybe you have a point.
[1] https://openrouter.ai/rankings#task-spend
The top models by task is self reported. It’s based on the referrer header or some custom one I forget at this moment.
If you don’t include that it just doesn’t count those tokens.
Optionality. This positions Stripe well to compete in AI compute, model development, apps... This is not a simple 1+1=2 merger.
If any of those companies put Ai into their pitch they too would be worth billions
You've already seen Long Island Blockchain, now get ready for... Alaska AI Lines!
I landscape gardens... With AI.
Pay me.
Cursor got bought for $60B. This is simply how AI companies are valued.
You could just as easily ask yourself how can a middle man for people finding web pages be worth so much, but here we are, and Google is worth something like $4 trillion, mostly on the back of search.
As it turns out, it's very valuable to be the intermediary between a large number of people who want something and a commoditized market of providers. It's the middleman who captures most of the margin.
Now, AI models are not a commodity yet. But things seem like they might be heading that direction. And in a world where they are, Stripe probably wants to be that guy sitting in the middle.
When there is a bidding war for a company, I don’t think fundamentals matter as much anymore..
A good year for Alaska Airlines is 5% sales growth. OpenRouter is growing that much per week I'm guessing.
Their market share is as big as it gets for that type of AI business.
The market cap of any business operating in the physical realm has to account for enormous liability, especially for one like an airline. Also, their operating costs are huge with much less ability to scale, resulting in much lower potential increase in margins.
[dead]
[flagged]
Dude I looked at your post history and every single comment is linking to your blog.
[dead]