Snowflake and NVIDIA Launch AI Model Routing Solutions as Gartner Warns of Fivefold Surge in Inference Costs by 2028

Snowflake and NVIDIA are launching AI model routing solutions to combat skyrocketing inference costs, as a new Gartner report warns enterprises could face a fivefold surge in AI inference expenses per agentic workflow by 2028.

• Snowflake is launching a dynamic model routing solution within its Cortex AI Gateway and flagship AI products, CoCo and CoWork, matching AI models to task complexity to cut enterprise costs without sacrificing quality.
• NVIDIA is also entering the space with NeMo Switchyard, an open-source routing library that enables enterprises to direct tasks to smaller, domain-specific models for faster responses, greater token efficiency, and higher accuracy.
• A new Gartner report warns that AI inference costs per agentic workflow will rise more than fivefold by 2028, driving the industry toward efficiency strategies like model routing, fine-tuning, and prompt compression rather than reduced AI adoption.