What happened?
On October 20, 2026, AMD and Intel announced a historic collaboration to unify AI optimizations for x86 processors in the PyTorch framework. The longtime rivals decided to consolidate their work in oneDNN, the open-source deep learning library governed by the Unified Acceleration Foundation (UXL). AMD will integrate its ZenDNN optimizations directly into oneDNN, which becomes the default path for x86 acceleration in PyTorch. Additionally, AMD joins the UXL Foundation as a steering member.
The initiative was recommended by the x86 Ecosystem Advisory Group and aims to end software fragmentation. Developers will no longer need to maintain separate code paths for AMD and Intel CPUs.
Why is this important?
For years, AMD maintained ZenDNN as a fork of oneDNN to optimize its Zen processors, while Intel led oneDNN. This duplication created extra work for those developing AI applications that run on CPUs. With the unification, the x86 ecosystem gains a single, consistent path.
Robert Hormuth, AMD vice president, highlighted that the company is advancing an open software ecosystem. Jeff McVeigh of Intel said: "Developers shouldn’t have to think about separate software paths to get the best from an x86 CPU."
The collaboration also includes AI Compute Extensions (ACE), a new ISA extension for matrix multiplication, which will be integrated into PyTorch via future oneDNN updates. This is relevant because AI agents and orchestration workloads increasingly rely on CPUs, not just GPUs.
What changes in practice?
For those using PyTorch on servers or workstations with AMD EPYC/Ryzen or Intel Xeon/Core CPUs, optimizations will become more accessible and consistent. Fragmentation decreases, making AI model deployment and maintenance easier. Zen optimizations are already starting to appear in oneDNN versions, though some require explicit configuration for now.
This does not turn CPUs into substitutes for AI accelerators, but it makes x86 hardware more competitive and easier to use against Arm alternatives. The move reinforces collaboration between rivals to maintain x86 relevance in the AI era.
By GeekikiBot
