The AI model routing and inference market is heating up rapidly. Stripe has agreed to acquire AI gateway OpenRouter for over $8 billion, positioning itself directly in the developer transaction path. Simultaneously, Ramp has launched its own Router service to help businesses manage AI spend. To combat intensifying competition from Anthropic and low-cost Chinese models, OpenAI has slashed developer pricing for its frontier GPT-5.6 Sol model by over 20%. These shifts come as enterprises like AT&T aggressively adopt routing tools and open-source models to slash AI costs.
Stripe acquisition of OpenRouter
- ▪OpenRouter closed a $113 million Series B funding round in May 2026 at an externally reported valuation of approximately $1.3 billion.
- ▪Stripe agreed to acquire AI model routing platform OpenRouter for more than $8 billion in cash and stock, following reports of a signed deal above $7 billion.
- ▪Stripe completed its acquisition of the metering engine Metronome in January 2026, which is used by OpenAI, Anthropic, and Nvidia to bill on tokens and GPU seconds.
AI model routing gateways
- ▪OpenRouter's weekly traffic grew from 5 trillion tokens to 25 trillion tokens over a six-month period ending in May 2026.
- ▪OpenRouter runs an AI gateway that allows developers to integrate a single API while the platform authenticates to providers and routes requests based on price, throughput, or availability.
- ▪Ramp's Router service will not charge a separate usage fee through the end of 2026, though customers must pay for the underlying model inference.
- ▪Corporate spend management platform Ramp launched Router, a service that connects developers to multiple AI models through a single API, currently available to United States customers.
OpenAI GPT pricing cuts
- ▪OpenAI previously cut prices in late July 2026, reducing the mid-tier GPT-5.6 Terra model pricing by 20% and the lower-cost Luna model by 80%.
- ▪OpenAI reduced GPT-5.6 Sol pricing to $4 per 1 million input tokens and $20 per 1 million output tokens, down from $5 and $30 respectively.
- ▪OpenAI cut developer API pricing for its frontier GPT-5.6 Sol model by more than 20% for a three-month period starting August 21, 2026.
Enterprise AI cost optimization
- ▪AT&T reduced the costs of coding and advanced AI tasks by up to 56% by using LiteLLM model routers to direct queries to cheaper models.
- ▪AT&T's use of LiteLLM model routers resulted in only a 2% decline in AI performance quality, according to AT&T Vice President Mark Austin.
Open-source model adoption
- ▪AT&T aims to increase the share of employee queries powered by open-source models from 40% to between 60% and 70% in the coming years.
- ▪AT&T is utilizing open-source models including Nvidia's Nemotron, Meta's Llama, and Google's Gemma, while evaluating potential risks of Chinese models.
AI inference market competition
- ▪Chinese-origin AI models held above 30% of United States token volume on OpenRouter weekly since February 8, 2026, peaking near 46%.
- ▪Anthropic prices its frontier Claude Fable 5 model at $10 per 1 million input tokens and $50 per 1 million output tokens.
Story comments
Loading comments…