AMD Joins UXL Foundation to Standardize AI Software
AMD and Intel are unifying acceleration libraries to simplify PyTorch performance across x86 CPUs.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Do you believe standardizing open-source software libraries across different chip brands benefits end users?
AMD has joined the UXL Foundation as a steering member, marking a shift toward unified AI software development. Additionally, AMD and Intel have agreed to adopt the oneDNN library as the default standard for x86 CPU acceleration within the PyTorch framework.
Why it matters
This move aims to reduce software fragmentation, potentially simplifying how developers optimize AI models for x86-based hardware by standardizing the underlying acceleration libraries.
The oneDNN library provides a unified API for deep learning primitives, allowing developers to target x86, Arm, and RISC-V CPUs via a single interface. This standardization replaces varied implementation methods for frameworks including PyTorch, TensorFlow, ONNX Runtime, and OpenVINO.
The players
AMD
A semiconductor company that designs high-performance CPUs and GPUs for data centers, gaming, and AI workloads.
Intel
A major chipmaker focused on x86 processor architectures and specialized hardware for AI acceleration.
UXL Foundation
A cross-industry group operating under the Linux Foundation to develop open-source AI software standards.
PyTorch
A leading open-source machine learning library primarily used for developing and training deep learning models.
The details
The integration uses oneDNN — a specialized library for deep learning operations — to handle computation acceleration on x86 CPUs. By adopting this as the default in PyTorch, the initiative aims to streamline the way deep learning primitives are executed across hardware, rather than requiring separate, custom code paths for different chip architectures.
Timeline
September 2023: The UXL Foundation was formally established.
October 20, 2026: AMD officially assumed its role as a steering member of the UXL Foundation.
The Tech Race
The UXL Foundation is attempting to build a vendor-neutral software ecosystem for AI, similar to how the Linux Foundation fostered cross-platform operating system standards. AMD's steering membership directly challenges the proprietary software silos that have previously fragmented AI performance optimization.
AI researchers and engineers can expect simplified workflows for optimizing models on x86 CPUs as PyTorch transitions to the new default library. Compatibility improvements will roll out as part of the broader effort to align deep learning frameworks with standardized API primitives.
The takeaway
The move suggests a long-term shift toward a unified open-source software layer that minimizes hardware-specific bottlenecks for AI performance. Observers should track the next updates to the PyTorch repository to verify the speed and adoption of the oneDNN transition.
Further reading
For broader trends in cross-platform development, visit our Artificial Intelligence section.
Live Poll
Do you believe standardizing open-source software libraries across different chip brands benefits end users?
