AMD and Intel Unite to Simplify AI Development on x86 CPUs with PyTorch
AMD and Intel are deepening their collaboration to make AI development easier, faster, and more consistent across x86 processors. As part of their work through the x86 Ecosystem Advisory Group, the two chipmakers are aligning around oneDNN as the default open-source library for x86 acceleration in PyTorch.
This move is designed to reduce complexity for developers building AI applications on x86 CPUs. Instead of managing separate optimization paths for different processors, developers will be able to rely on a more unified software foundation that supports modern CPUs from both AMD and Intel.
A major part of the announcement is AMD bringing its ZenDNN optimizations into oneDNN. ZenDNN has been AMD’s performance-focused library for accelerating deep learning workloads on AMD EPYC and Ryzen processors. By contributing these optimizations to oneDNN, AMD is helping create a broader and more standardized solution for AI acceleration across the x86 ecosystem.
AMD has also joined the UXL Foundation as a Steering Member, further reinforcing its commitment to open software development and cross-industry collaboration. The UXL Foundation focuses on building open standards and tools that help developers create high-performance applications across different hardware platforms.
For PyTorch users, this collaboration could bring meaningful benefits. PyTorch is one of the most widely used frameworks for AI research, machine learning development, and production deployment. With oneDNN serving as a common acceleration layer, developers can expect a more streamlined experience when running AI workloads on x86 CPUs.
The goal is simple: make it easier to build and deploy high-performance AI applications without forcing developers to deal with vendor-specific software fragmentation.
Intel emphasized that developers should not need to choose separate software paths to get strong performance from x86 CPUs. By working with AMD through oneDNN, Intel says the PyTorch community gains a more consistent approach for modern processors, while also creating a path for future capabilities such as AI Compute Extensions.
AMD also highlighted the importance of an open software ecosystem. The company said this collaboration gives developers more choice and makes it easier to create efficient AI applications across x86 processors. By working through the x86 Ecosystem Advisory Group, AMD and Intel are aiming to bring the industry closer to shared technologies that simplify software development.
Red Hat also welcomed the move, noting that open-source innovation works best when developers have a unified and seamless experience. Standardizing x86 CPU optimizations through oneDNN in PyTorch is expected to reduce ecosystem fragmentation and help teams build, optimize, and deploy AI workloads more consistently.
The broader importance of this partnership goes beyond a single library or framework. AI workloads are becoming more demanding, and organizations need flexible hardware options as they scale machine learning and generative AI applications. While GPUs often dominate AI training discussions, CPUs remain essential for inference, preprocessing, data handling, and many enterprise AI workloads.
By improving CPU acceleration in PyTorch, AMD and Intel are strengthening the role of x86 processors in the AI landscape. A unified oneDNN-based approach can help businesses get more value from existing infrastructure while making it easier to support future hardware advancements.
This collaboration also signals a shift in how major x86 companies are approaching software. Instead of competing only through separate optimization stacks, AMD and Intel are working together on shared foundations that benefit developers, enterprises, and the wider open-source community.
For developers, the biggest advantage is simplicity. Less time spent managing hardware-specific workarounds means more time can be focused on building models, improving applications, and deploying AI solutions at scale.
For businesses, the benefit is flexibility. A more consistent PyTorch experience across AMD and Intel CPUs can make infrastructure planning easier and reduce concerns about software compatibility when choosing x86 platforms.
Ultimately, the expanded AMD and Intel collaboration marks an important step for AI development on x86 CPUs. By combining AMD’s ZenDNN optimizations with oneDNN and making it the default open-source acceleration path in PyTorch, the companies are helping create a more open, efficient, and developer-friendly AI ecosystem.
