OpenRouter raises $113M Series B for AI model switching gateway, processing 25T tokens weekly

OpenRouter raises $113M Series B for AI model switching gateway, processing 25T tokens weekly

OpenRouter, the AI gateway that lets developers route requests across 400+ models from 70+ providers via a single API, announced a $113 million Series B round led by CapitalG (Alphabet's growth fund). NVentures (NVIDIA), ServiceNow Ventures, MongoDB Ventures, Snowflake Ventures, Databricks Ventures, AMP PBC, and Pace Capital joined as new investors, with Andreessen Horowitz and Menlo Ventures participating. The round comes roughly one year after its $40 million Series A in June 2025. The company now serves over 8 million global users and processes approximately 25 trillion tokens per week — a 5x increase from 5 trillion tokens six months ago.

Why it matters: OpenRouter sits at the intersection of two powerful structural currents in the AI substrate. First, the hyperscaler-distribution pattern (§5.2) is inverting: rather than a single cloud provider bundling model access, OpenRouter aggregates supply from every major lab — Anthropic, OpenAI, Google, xAI, DeepSeek — and presents a unified endpoint. This effectively commoditizes the provider layer and shifts value to the routing layer, a dynamic the framework identifies as a recurring pattern when inference becomes disaggregated. Second, the capital-compression arc is accelerating: between this round and previous raises from hyperscaler-affiliated VCs, OpenRouter is building a capital moat that mirrors what we saw with early infrastructure consolidators. The 5x token-volume growth in six months validates the thesis that enterprises need multi-model orchestration for production workloads — customer support uses latency-optimized models, contract analysis uses reasoning-strong models, and agents increasingly require dynamic fallback across providers.

Grounded expert take: OpenRouter's real strategic value is as a hedge against provider concentration. In an environment where OpenAI, Anthropic, and Google are all racing toward vertical integration — training their own models, building their own inference stacks, and distributing directly — a neutral aggregation layer becomes indispensable for enterprises that refuse to lock into any single lab. The participation of Snowflake, MongoDB, Databricks, and ServiceNow as strategic investors signals that data-platform companies see OpenRouter as essential middleware for the AI-augmented enterprise stack. However, the core challenge remains: if any one provider delivers a model so superior across cost, latency, and accuracy that it dominates routing decisions, the aggregation thesis weakens. OpenRouter's Series B bet is that the opposite future — fragmented model specialization — will define the next cycle.

#OpenRouter #ModelGateway #InferenceInfrastructure #AIOrchestration #EnterpriseAI #SeriesB