AMD and Intel, longtime rivals in the CPU market, are joining forces to simplify AI inference on x86 processors. According to a report from Tom's Hardware, the two companies are working together to bring CPU inference optimizations for PyTorch under the unified oneDNN library.
As part of this effort, AMD's ZenDNN library will be folded into Intel's oneDNN. The result is a single PyTorch code path for CPU-based AI inference, rather than separate, competing implementations that developers have had to target individually.
The move is intended to make deployment easier and more consistent across hardware from both vendors. The report notes that this collaboration addresses fragmentation in the AI software stack, giving developers one common optimization layer for PyTorch workloads on CPUs. Since there is only one source, no independent reporting is available to corroborate the details, but no conflicting accounts are noted either.