OpenRouter is joining Stripe. The announcement landed yesterday on the OpenRouter blog in the company’s signature understated tone - same mission, same name, same product, same roadmap - but the numbers underneath are anything but small: 10+ trillion tokens processed every day, 400+ AI models behind one interface, more than 10 million developers and companies, and inference volume that has grown at least 10x every year since the company started in early 2023.
The deal instantly became the most-commented tech story of the day on Hacker News (710 points, 365 comments). Commenters put the price tag around $7 billion, though neither company has confirmed a figure publicly. OpenRouter’s founding bet is stated plainly in the post: intelligence will be multi-model, and “AI is too important for its future to be decided by whichever single model gets embedded first.” For Stripe, the move reads as a bet that inference is becoming the biggest line item on every company’s bill - and that the gateway between developers and models is worth owning.
🎩 Cask’s Take
The funniest line in the thread was a one-liner from a commenter who summed it up as “middleman company buying middleman company.” The sharpest argument underneath it is whether a token is a currency or a commodity. The defenders make a real case: one API key with time-bound, model-restricted budgets, plus routing that quietly shifts to a cheaper or more reliable provider the moment one wobbles - that is insurance, and insurance costs a premium. The skeptics answer with the other half of the trade: the premium exists because switching costs exist, and the moment switching gets easy enough, someone open-sources the router and the markup disappears.
What Stripe is actually buying is position. If paying for intelligence ends up looking like paying for card payments - a toll collected on every transaction - then the company sitting between every developer and every model is infrastructure, and infrastructure compounds. The promise that “routing stays driven by what’s best for you” is easy to make while independent; it becomes a statement that gets tested the first time the parent company’s incentives and the user’s incentives point in different directions. That test, not the valuation, is the story worth watching.