Stripe reportedly paid $7.5 billion for a company that never built a frontier AI model.

Look at how it makes money.

No margin on the models. None. Provider prices pass through at list. OpenRouter charges about 5.5% on the money you load onto the platform.

No revenue from being right about models. All of it from being where the money enters.

Two companies made opposite bets on that.

Martian launched in November 2023 with $9 million and the smarter product. Send us your prompt, we pick the best model. Pay us for judgment.

OpenRouter sold no judgment at all. One key. One bill. 400+ models. The choice stayed with the developer.

OpenRouter sold no judgment at all. One key.
The choice stayed with the developer.
One bill. One invoice , One vendor , One Procurement for 400+ models merged into one OpenRouter



One tried to be the smartest thing on the road. The other became the road.

Twenty four years ago I was the road, for a much smaller market.

A MERA softswitch was quoted to me at thousands I didn't have, so I stayed up nights and wrote my own. Every call had to find the cheapest carrier that still sounded clean. Ten cents in from the customer. Eight cents out to the carrier. The two cents that stayed was the entire business.

Nobody paid me for choosing the right carrier. They paid because every call had to pass through me.

And I had one thing AI routing does not. A bad route was audible in three seconds. A bad model pick, you never see the answer the other model would have given.

Smart routing cannot be proven. Only claimed.

Would you back the router that makes the best decisions, or the one nobody can afford to route around?


Krishna Lakamsani | Entrepreneur · Venture Studio Founder · Investor. Writing at the intersection of AI, capital, and the future of work. For founders who build what's next. #MyLifeLessons #Reflections #InspireFounders #BuildInPublic #AiImpact