Nvidia, Tech Giants Embrace Open AI Models as Inference Demand Surges

The economics of generative AI are shifting from costly model training to inference at scale. Microsoft Azure and Amazon Web Services are expanding access to both open and proprietary models, allowing companies to route routine summarization, classification and coding tasks to cheaper open systems while reserving advanced reasoning for frontier closed models. The approach gives developers more control over performance, data and spending as AI moves into daily operations.

Nvidia’s latest earnings commentary and the strategies of Microsoft and AWS point to wider adoption of this mixed-model architecture. Nvidia is also releasing models and software more openly to draw workloads into its CUDA-based ecosystem. Quantization, distillation and faster inference can reduce the cost of each request, but lower prices are encouraging heavier usage and more frequent AI-agent activity, supporting continued demand for Nvidia GPUs and growth in its data-center business.