Qualcomm has announced a custom AI inference chip with Microsoft and Meta as founding customers, validating its alternative-to-GPU strategy with the largest hyperscalers.

Qualcomm has announced a custom AI inference chip, marking a significant push into the accelerator market dominated by NVIDIA. The chip has secured Microsoft and Meta as founding customers, providing critical validation from two of the world's largest hyperscalers.

The choice by Microsoft and Meta signals growing confidence in alternatives to GPU-based inference architectures. By backing Qualcomm's strategy, both companies demonstrate willingness to diversify their chip portfolios and reduce dependency on traditional GPU providers for workload-specific tasks.

This development indicates acceleration in the broader shift toward custom silicon within hyperscaler infrastructure. As model inference scales across multiple platforms and use cases, specialized chips designed for specific workloads become economically compelling, opening room for non-GPU competitors in what has been an NVIDIA-centric market. Qualcomm's validation from the two largest customers suggests confidence in the inference-chip segment as a meaningful lever in the AI buildout.