TL;DR

Bloomberg reports that Stripe has closed a deal to buy OpenRouter, the gateway developers use to route requests across competing AI models, at a price above $7bn. Talks were first reported by the Wall Street Journal last month. Stripe says it does not comment on rumour or speculation.

What Stripe is buying

OpenRouter sits between an application and the model providers, letting a developer pick a different model per task according to capability and cost without rewriting integrations. It claims 8 million users worldwide and access to over 400 models. Its chief executive Alex Atallah has described the company as Stripe for AI — a single point of access that keeps customers from being locked to one vendor.

The valuation trajectory is steep. A $113m Series B in May priced the business at a reported $1.3bn, with Alphabet’s growth arm alongside Menlo Ventures, Sequoia and Andreessen Horowitz. The acquisition figure is more than five times that within four months.

Why a payments company wants a model router

The analogy Atallah used is the strategy. Stripe’s business has always been an abstraction layer over fragmented, awkward infrastructure that developers would rather not integrate individually — card networks then, model providers now. Both are metered, per-request services where the value sits in the routing and reconciliation rather than the underlying supply.

That is also what should give buyers pause. The pitch for a neutral gateway is that it prevents lock-in, which is a harder claim to sustain once it belongs to a company with its own commercial interest in what flows through it.

What it means for UK teams

Any British business running a multi-model architecture through this gateway now has a vendor relationship it did not choose, and should be checking what its contract says about assignment, pricing changes and data handling under new ownership. The practical mitigation has not changed: keep the routing layer replaceable, and know what switching would actually cost.

Looking forward

This is consolidation of the plumbing rather than the models. If the pattern holds, the layer that decides which model handles a request may end up more commercially valuable — and more concentrated — than the labs competing to answer it.