AMD and Intel, which compete with each other, are cooperating under a shared goal of making AI development easier on x86 CPUs, HotHardware reported on Oct. 23 local time.

The two companies will integrate their PyTorch optimisation technologies into the open-source library oneDNN.

The cooperation will allow developers to use the same software on processors from both companies, reducing the hassle of having to optimise AI workloads separately for each CPU, HotHardware said.

Under the collaboration, AMD will integrate its optimisation technology ZenDNN into Intel's oneDNN.

As a result, oneDNN will become the default open-source library for x86 acceleration in PyTorch. Intel and AMD have also jointly developed a next-generation x86 instruction set extension, AI Compute Extension (ACE), for faster matrix multiplication.

ACE support features will be included in future oneDNN updates.

The cooperation between the two companies is seen as a move that takes into account an offensive by Arm-based competitors.

Nvidia recently released RTX Spark laptops featuring Blackwell GPUs, Grace CPUs and up to 128GB of unified memory. Nvidia is also accelerating an expansion of its Vera CPUs with an eye on the agent market.

Apple's Mac mini and Mac Studio have emerged as compact computers for running local AI models and agents.

AMD and Intel are still competing fiercely in the server market, but both companies want x86 to remain dominant rather than ceding server market share to the Arm camp, HotHardware said.