Ramp launched Router on Wednesday, an AI model routing service that lets developers and companies switch between large language models through a single API. The service opens with access to eight providers — OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai — and is free to use for the remainder of 2026, with a $26 launch credit against inference costs. Ramp did not disclose 2027 pricing.
The move puts Ramp directly into a market that Stripe is reportedly consolidating with a $7B+ acquisition of OpenRouter. Ramp, which raised $750 million at a $44 billion valuation in June, says it has been running Router internally for the past three years to handle its own AI workloads before productizing it. The service is currently US-only.
Router's core pitch is routing strategies. One option lets customers prefer model providers' flex-usage tiers to cut cost. Another picks a model based on up to three user-specified benchmarks. A third routes only hard queries to expensive frontier models while sending easy calls to cheaper ones — the kind of arbitrage that has driven OpenRouter's growth.
Key facts
- 01Router offers access to eight model providers: OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai.
- 02The service is free for the remainder of 2026 and launches with a $26 credit; 2027 pricing has not been disclosed.
- 03Ramp raised $750 million at a $44 billion valuation in June, giving it capital to enter the inference-routing market.
- 04Router's default one-year data retention is opt-out, with Ramp scrubbing personally identifiable information before using logs.
- 05Stripe is reportedly acquiring rival OpenRouter for $7B+, validating routing as a strategic layer in the AI stack.
Customers get a dashboard tracking token spend, cost, latency, and fallback attempts. That telemetry hook matters more than it looks: Ramp already sells AI token usage monitoring and spend management to its corporate customers, and Router feeds the same finance-ops surface that Ramp's core expense product lives on.
Data retention is where Router draws a line. Router logs model inputs, outputs, and tool calls for one year by default, with opt-out available. Ramp says it will remove
“personally identifiable information before using that content to improve the product.”— Ramp, company statement
The retention default is aggressive compared to enterprise-focused inference offerings — OpenAI, for instance, recently expanded Zero Data Retention to its frontier models, which we covered last week. Enterprises evaluating Router against direct API access from providers will weigh whether the routing convenience is worth an opt-out retention regime.
For Ramp, the strategic logic is two-sided. Every routed token is a new revenue line in a market that shows no signs of slowing, with valuations like Etched's doubling to $21B in a month on the strength of inference demand. And Router positions Ramp as a buyer-of-record for AI labs, potentially opening commercial relationships with providers worldwide that its expense-management sales team can leverage.
The comparison to OpenRouter is unavoidable, and Ramp is behind on catalog breadth. OpenRouter aggregates dozens of models, including long-tail open-source variants; Router's eight-provider launch is narrower. If Stripe's acquisition closes at the reported $7B+ mark, OpenRouter gets a distribution engine Ramp cannot match on payments plumbing alone.
The counterweight is that inference routing is not really a moat business — it is a distribution business. OpenRouter grew because developers found it first. Router will need to convert Ramp's existing corporate customer base into API users, which is a different motion than winning individual developers on GitHub or Cursor. Ramp has not yet said how it plans to bridge that gap.
Ramp's Router bet is really a bet that AI inference spend is becoming a finance-department line item, not just an engineering one. If that read is right, bundling routing, monitoring, and spend management under one vendor is a defensible package. If AI inference stays in engineering budgets and gets bought like cloud compute, Router competes against OpenRouter, hyperscaler gateways, and every provider's direct API — a much harder fight for a company whose brand equity sits with CFOs, not CTOs.
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.
