​​​​​​​In this article
What is an execution provider?
Execution Provider AI components are Windows AI components used on Copilot+ PCs to enable hardware‑accelerated execution of machine‑learning models. These components act as the hardware abstraction layer between AI models and the underlying compute engines—such as CPUs, GPUs, and NPUs—allowing Windows and applications to run AI workloads efficiently on the best available hardware.
An execution provider is a modular component that integrates with the ONNX Runtime to deliver hardware‑specific optimizations for AI inference. Execution providers handle tasks such as graph partitioning, kernel selection, and operator execution, while abstracting the complexity of vendor‑specific acceleration libraries. This design enables a single AI model to run across diverse hardware configurations without requiring application‑level changes.
On Copilot+ PCs, execution provider AI components support multiple hardware platforms and silicon vendors. Examples include execution providers optimized for Intel, AMD, Qualcomm, and NVIDIA hardware, each enabling AI models to take advantage of specialized accelerators such as NPUs or GPUs when available. When hardware acceleration is not supported for a given workload, execution automatically falls back to a compatible compute backend, ensuring reliability and broad device compatibility.
Execution Provider AI components are a foundational part of the Windows AI platform. They are used by Windows features, Copilot+ experiences, and developer applications that rely on local AI inference. By dynamically selecting and managing execution providers, Windows can deliver low‑latency performance, improved power efficiency, and consistent AI behavior across different device configurations.
MIGraphX execution provider
The MIGraphX execution provider is an AMD execution provider used with ONNX Runtime / Windows machine-learning to deliver hardware‑accelerated inference by offloading supported ONNX model operations to AMD GPUs. It is based on AMD’s MIGraphX graph inference engine, which accelerates machine‑learning model inference and enables hardware‑specific optimizations when running ONNX models on AMD GPU hardware.
NVIDIA TensorRT‑RTX Execution Provider
The NVIDIA TensorRT‑RTX Execution Provider is an ONNX Runtime / Windows machine‑learning execution provider designed specifically to accelerate ONNX model inference on NVIDIA RTX GPUs for client-centric (end‑user PC) scenarios. It leverages NVIDIA’s TensorRT for RTX runtime to generate and run RTX‑optimized inference engines on the local GPU, enabling Windows and apps to take advantage of RTX hardware acceleration.
Intel OpenVINO Execution Provider
The Intel OpenVINO Execution Provider is an execution provider used with ONNX Runtime / Windows machine‑learning to enable hardware‑accelerated inference on Intel platforms. It accelerates ONNX models on Intel CPUs, GPUs, and NPUs, allowing Windows and applications to take advantage of Intel hardware optimizations for machine‑learning workloads.
Qualcomm QNN Execution Provider
The Qualcomm QNN Execution Provider is an execution provider for ONNX Runtime (and Windows machine‑learning scenarios that use ONNX Runtime) that enables hardware‑accelerated execution on Qualcomm chipsets. It uses the Qualcomm AI Engine Direct SDK (QNN SDK) to construct a QNN graph from an ONNX model, which is then executed by a supported accelerator backend library.
AMD Vitis AI Execution Provider
The AMD Vitis AI Execution Provider is an execution provider used with ONNX Runtime / Windows machine‑learning that enables hardware-accelerated AI inference on AMD platforms. Vitis AI is AMD’s development stack for hardware-accelerated AI inference, and it targets AMD platforms including Ryzen AI, AMD Adaptable SoCs, and Alveo Data Center Acceleration Cards.
