# AMD acquires FastFlowLM to accelerate on-device AI inferencing

> Source: <https://www.sdxcentral.com/news/amd-acquires-fastflowlm-to-accelerate-on-device-ai-inferencing/>
> Published: 2026-07-20 10:00:18+00:00

[FastFlowLM](https://fastflowlm.com/) has become the latest acquisition by AMD as the chip giant continues to seek out means to boost performance across its stack.

FastFlowLM formed out of a collaboration between academic researchers, software engineers, and community maintainers. They built a runtime environment for powering AI models locally on AMD hardware.

Underpinned by [Iron](https://github.com/amd/IRON), the open-source close-to-metal Python API tailored for AMD neural processing units (NPUs), FastFlowLM ships without drivers, meaning users can simply run the installer, pull their desired model, and off they go.

AMD [confirmed](https://www.amd.com/en/blogs/2026/fastflowlm-joins-amd-to-advance-ai-inference.html) the FastFlowLM team has joined its AI Group, with their expertise set to “accelerate” AMD’s client and workstation AI software stack.

Among the team behind FastFlowLM are University of Rhode Island professors Tao Wei and Ken Qing Yang, and Zhenyu (Alfred) Xu from Clemson University.

“What started as a bold idea to make AI inference faster and more efficient is now entering an exciting new chapter,” Wei said in a [LinkedIn post](https://www.linkedin.com/posts/tao-wei-734a121b_fastflowlm-joins-amd-to-advance-ai-inference-share-7483901461043511299-f696/?utm_source=share&utm_medium=member_desktop&rcm=ACoAAB77WY0BoVw4HAZ8qXeMBPsQ5oeCHSmHqRI), with his profile now stating his role as a fellow at AMD.

Upon announcing the deal, AMD said it remains committed to “investing in this open ecosystem, and we’re thrilled to build the future of on-device AI together.”

FastFlowLM joins Mext, which [joined AMD in mid-June](https://www.sdxcentral.com/news/amd-acquires-storage-software-startup-mext/). Mext was a storage software startup whose Predictive Memory solution tricks an operating system into thinking flash is DRAM – thereby expanding a system's usable memory capacity using the cheaper and denser option.

The FastFlowLM, however, is part of AMD’s bid to push open alternatives to beat Nvidia’s proprietary stack lock-in. Its flagship Radeon Open Compute (ROCm) software stack gives developers and enterprises the means to run AI and high-performance computing (HPC) workloads with popular open frameworks like PyTorch, TensorFlow, and JAX upstream, as well as an open, Python-first GPU compiler in the form of the OpenAI-developed [Triton](https://triton-lang.org/main/index.html).

No financial details were disclosed.