OpenRouter raises $40 million to simplify AI model overload

OpenRouter, a startup helping software developers manage the growing number of AI models, has raised $40 million in venture capital. The company wants to make it easier for developers to choose and use the right AI model for their applications, without having to build their own complex systems.

“There’s been a Cambrian explosion of models. Our business is a one-stop shop for all models,” CEO and co-founder Alex Atallah told The Wall Street Journal.

OpenRouter runs a platform that connects AI apps to different LLMs and cloud providers. It automatically picks the best option based on a developer’s needs, like price, speed, and performance. The platform includes models from AI players such as OpenAI, Anthropic, and Meta, and uses cloud services from Amazon and Microsoft, among others.

The startup raised a $12.5 million seed round in February, led by Andreessen Horowitz, and a $28 million Series A round in April, led by Menlo Ventures. It is now valued at around $500 million, according to a person familiar with the matter.

Developers using OpenRouter buy credits to run AI inferences. That means the responses AI models generate when given a prompt. In May, customers spent about $8 million through OpenRouter, which is 10 times more than what they spent in October. The company takes a 5 percent cut of inference costs.

Chris Clark, OpenRouter’s chief operating officer, said, “We believe that inference costs will eclipse salaries as the dominant operating expense for most knowledge-based companies over the next five to 10 years.” He estimates that AI inference spending will reach $25 billion in 2025, with $15 billion of that coming from third-party apps that don’t own the AI models they use.

Atallah believes more companies will choose to outsource this kind of infrastructure work. “New models are constantly released and their parameters [are] changing,” he said.