OpenRouter closed a $113M Series B led by CapitalG at a $1.3B post-money valuation, more than doubling from the $547M mark it hit a year ago after raising $40M in Series A. The AI gateway startup, founded in 2023, now processes 100T tokens per month across 400+ models from Anthropic, Google, OpenAI, xAI, and DeepSeek, serving 8M global users.
Token volume surged 5x in six months, from 5T tokens per week to 25T per week today. That acceleration tracks the broader shift from training to inference to agentic workflows, where enterprises swap models mid-task to balance cost and reasoning depth.
“The gateway helps enterprises and other AI users select different models for different jobs to control costs or increase reasoning and accuracy for the task at hand.”— OpenRouter, Company statement
The Series A in June 2025 was led by Andreessen Horowitz and Menlo Ventures, with participation from Sequoia. The new round brings CapitalG, Alphabet's growth fund, into the mix, a notable endorsement given Google competes on model serving.
Key facts
- 01OpenRouter raised $113M in Series B at a $1.3B post-money valuation, up from $547M a year ago.
- 02The AI gateway processes 100T tokens per month across 400+ models from Anthropic, Google, OpenAI, xAI, and DeepSeek.
- 03Token volume jumped 5x in six months, from 5T tokens per week to 25T per week.
- 04The round was led by CapitalG, Alphabet's growth fund, with prior backers including Andreessen Horowitz and Sequoia.
OpenRouter's growth reflects a fundamental change in how enterprises adopt AI infrastructure. Rather than standardizing on a single model provider the way they locked into SaaS vendors, companies are treating models as swappable compute layers. OpenRouter's gateway lets them route tasks to the cheapest or most capable model for the job, flipping between providers without rewriting code.
The 400+ model catalog spans proprietary frontier models and open-weight alternatives, giving enterprises optionality as DeepSeek and other international labs close the gap on reasoning benchmarks. Multi-model routing is becoming table stakes for production AI, not a niche workflow.
“Companies have no plans to get locked into a model vendor as they did with their various SaaS providers.”— OpenRouter, Company statement
The startup claims 8M users globally, though it doesn't break out enterprise vs. developer accounts. Token processing volume is the clearer metric, and 100T tokens per month puts OpenRouter in rarefied air alongside the frontier labs' own inference platforms.
CapitalG's involvement is the sharpest signal yet that gateway infrastructure is a durable business, not a transitional convenience. Alphabet's venture arm rarely leads rounds in markets where Google competes directly, yet here they're backing the abstraction layer that makes Google's Gemini one choice among many.
The counterargument is that OpenRouter's margin compresses as model providers cut pricing. Anthropic, OpenAI, and Google all dropped inference costs 30–50% in the past year, and open-weight models run even cheaper on dedicated hardware. If enterprises can self-host or negotiate volume deals directly, the gateway's value proposition narrows to convenience.
OpenRouter's 5x token-volume jump in six months suggests enterprises aren't self-hosting yet. The complexity of managing 400 models, tracking uptime, and routing tasks in real time is steep enough that paying a gateway fee beats building in-house. Whether that holds as model APIs commoditize further is the bet CapitalG just made at a $1.3B valuation.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
