AI model routing draws Cursor, Ramp, Meta and Stripe interest
Cursor, Ramp, Meta and Stripe-linked deal talks show demand for software that assigns AI requests to lower-cost models.
By Rafael Ortiz · Fintech Correspondent
· 3 min read
AI model routing has become a focus for companies trying to reduce the cost of running artificial intelligence at scale, with Cursor, Ramp, Meta and Stripe-linked deal activity all surfacing in one week in July. The software can cut AI spending by 30% to 50% when used effectively, and some workloads can see savings of as much as 85%, according to Mindstudio.
Routing tools assess a request before it reaches a model, then send it to a system judged sufficient for that task. That avoids sending routine queries to frontier models that are more expensive to use, while reserving higher-end systems for harder work.
What is AI model routing?
AI model routing is software that decides which artificial intelligence model should handle a specific prompt or workflow. The business case rests on matching cost to complexity: simpler requests can be handled by cheaper models, while more demanding tasks can still be routed to more capable systems.
Cursor launched Cursor Router on Tuesday, July 21, according to AlphaSignal. The coding tool evaluates requests and directs them to a model suited to the job. AlphaSignal reported that about 60% of Cursor users had been relying on one AI model by default across different tasks. Cursor said internal testing showed the Router produced the same output quality at 60% lower cost, while early enterprise users reported savings of 30% to 50%, according to AlphaSignal.
Ramp has also moved from internal cost control to external distribution. The spend management company said in a launch announcement that it spent three years building a routing tool and used it across more than 100 of its own AI features. Ramp said the tool reduced its internal AI costs by 30%. The company opened the product to other businesses without charging for the routing software, while users still pay for the underlying AI consumption.
Ramp has said customer demand for AI tools is rising sharply. Business AI spending among Ramp customers increased nearly tenfold between January 2025 and April 2026, according to figures the company shared publicly on social media.
Why are payments and billing companies watching routing?
Model routing sits before the invoice. A routing decision determines which model handles a request, which then affects the usage charge that a business or customer may later need to reconcile.
That makes the category relevant beyond software engineering teams. PYMNTS reported that Stripe was negotiating to acquire OpenRouter, a startup operating AI model routing at large scale, for about $10 billion. Stripe has its own exposure to AI billing through payments and usage-based pricing tools. Stripe research found that unpredictable AI pricing is the main barrier to adoption for 46% of IT leaders.
Industry analysis from billing platform Flexprice found that AI infrastructure costs have risen from about 10% of a typical AI company’s budget to as much as 35% to 40% as usage grows. Flexprice also cited 2025 research from Metronome, which Stripe acquired in January, finding that 92% of AI companies that began with usage-based pricing changed that model at least once.
Meta is pursuing a similar cost-control idea internally, according to The Information. The report said Meta is building a system called Switchboard to direct simple requests to cheaper models and reserve costlier models for more complex tasks. Meta has not confirmed the report or said whether such a system would be sold outside the company.
Taken together, the moves show how AI cost management is shifting from after-the-fact budgeting toward decisions made at the moment a request is routed. For companies handling large volumes of AI queries, that layer can shape both operating costs and the billing records attached to each use.
This story draws on original reporting from PYMNTS.
