OpenRouter has raised $113m in a Series B led by Alphabet's Capital G, which is an interesting choice of investor for a company whose core value proposition is that no single AI provider, including Google, should have a monopoly on your workloads.
The product is straightforward. OpenRouter gives customers access to more than 400 AI models through a single gateway, letting them route different tasks to different providers based on cost, performance or capability.
The picks-and-shovels pitch
The bull case is simple. The AI model market is fragmented and will stay that way. Different models are better at different things. Enterprises do not want to be locked into one provider. OpenRouter sits in the middle and takes a cut.
It is the classic infrastructure play: do not bet on which model wins, bet on the switching layer that connects them all.
The consolidation risk
The bear case is equally simple. If the market consolidates to two or three dominant providers, the switching layer becomes less valuable. Nobody needs a gateway to choose between two options.
The 400-model catalogue is impressive today but misleading as a guide to the future. Most of those models are niche, undifferentiated or unlikely to survive. The models that matter, the ones enterprises actually use for production workloads, number in the single digits.
The real business might be consulting
Where OpenRouter could build durable value is in helping customers figure out which model to use for which task. That is a genuinely hard problem, and the data OpenRouter collects on usage patterns across hundreds of models gives it an information advantage that no individual model provider has.
How many models does the average customer actually use? What are the switching patterns? Which workloads are price-sensitive and which are performance-sensitive? That data is worth more than the routing infrastructure itself.
Whether OpenRouter builds that into a consultative business or remains a commodity switching layer will determine whether the $113m was well spent.
