AMD and Intel agree to combine PyTorch CPU optimizations in one library

AMD and Intel are putting their PyTorch CPU optimization work on a shared path. The companies announced a collaboration on October 20, 2026, to combine that work in the open-source oneDNN library. They say the collaboration will make AI applications easier to build, maintain and deploy across both companies’ x86 processors.

An advisory recommendation becomes a shared library choice

Following a recommendation from the x86 Ecosystem Advisory Group, AMD and Intel have aligned on oneDNN as the default open-source library for x86 acceleration within PyTorch. oneDNN is a deep-learning software library that helps AI workloads run efficiently on CPUs. It is governed by the UXL Foundation, placing the joint optimization effort in an existing open-source project.

Developers shouldn’t have to think about separate software paths to get the best from an x86 CPU,

Jeff McVeigh, Intel vice president and general manager, Data Center Strategic Programs

AMD brings ZenDNN work into the common path

AMD’s contribution is to integrate its ZenDNN optimizations into oneDNN. The company says it will continue optimizing for AMD processors while doing so. The common software path therefore does not mean an end to processor-specific optimization: AMD’s stated approach is to bring that work into the shared library. The integration is underway, rather than a completed rollout.

Future processor instructions will use the same route

The collaboration also covers AI Compute Extensions, or ACE, a next-generation x86 instruction set. These processor instructions are designed to accelerate matrix multiplication, a core mathematical operation in AI workloads. AMD and Intel say ACE support will reach PyTorch through future oneDNN updates. That part of the effort concerns support for a future hardware capability, not an acceleration feature available with this announcement.

Neither milestone has a disclosed delivery date: the announcement gives no completion date for ZenDNN integration or release date for ACE support.