Bloomberg reported on August 16, 2026 that Stripe had finalised a deal to acquire OpenRouter, the AI gateway that routes requests across more than 400 models for around 8 million users, for more than $7 billion. Stripe told TechCrunch that it does not comment on rumours or speculation. The following day, OpenRouter shipped something considerably more concrete.

The facts: on the acquisition, OpenRouter raised a $113 million Series B in May 2026 at a $1.3 billion valuation, which would make the reported price roughly a 5.4× markup in three months. The Wall Street Journal had reported talks in July. On the product, OpenRouter published an Activity Dashboard on August 17 with five sections — Overview covering spend, requests, token volume, cache hit rate and blended cost per million tokens; Trends sorted by movement rather than size; Explore for building custom charts grouped by up to two dimensions; Guardrails showing prompt-injection blocks and sensitive-data redactions; and Logs with per-request detail. It also shipped a beta Analytics API with two endpoints requiring a management key, GET /api/v1/analytics/meta and POST /api/v1/analytics/query, plus saved views shareable across an organisation, CSV and PDF export, and an openrouter-analytics agent skill.

Automation Squad's take: these landed a day apart and the smaller one is the one to act on. Nobody reading this can influence whether Stripe buys a gateway, and the number will be litigated by people with better information. What you can do today is find out what you are actually spending and where. The blended cost per million tokens is the figure most teams have never seen, because their spend is smeared across models, keys and applications in a way no invoice untangles. The cache hit rate sitting next to it is usually the largest available saving and the one nobody has looked at. The acquisition does raise one legitimate operational question worth writing down — how much of your traffic runs through a single gateway, and whether you have a fallback if its terms change under new ownership.

Run this now: open the Activity Dashboard and read two numbers before anything else — blended cost per million tokens, and cache hit rate. Then go to Trends and look for something rising that you did not decide to increase, because that is where unapproved spend shows up. Deep-link from your most expensive slice into the logs and read ten real requests; retries, tests and forgotten jobs turn up more often than anyone expects. Finish by writing down what fraction of your LLM traffic depends on this one gateway, and what your fallback is.