AMD is using its open source ROCm platform to empower AI developers, promote collaboration, and offer scalable alternatives to proprietary GPU software.
AMD is doubling down on open source to make its ROCm platform a go-to ecosystem for AI development, aiming to break Nvidia Cuda’s dominance. The US chipmaker is positioning ROCm as a collaborative software stack, integrating with projects like virtual large language model (vLLM) to accelerate innovation.
“We could try to build something that’s closed source, but we won’t get the velocity of an open ecosystem. Instead, we want to leverage everyone’s capabilities to move the industry forward. It’s like the Linux kernel, where everyone collaborates and moves really fast,” said Anush Elangovan, AMD Vice-President of AI Software.
Targeting the Asia-Pacific region, AMD hopes ROCm will provide a “common baseline” for companies to scale AI capabilities without relying on proprietary tools. Elangovan noted, “I’ve been involved with quite a few companies in the region that are building large datacentres and deploying AMD chips at scale. ROCm will also allow them to compete in both model development and infrastructure.”
The “ROCm everywhere” initiative ensures a unified experience across devices, from laptops to supercomputers, supporting students, startups, and enterprise deployments. AMD’s chiplet architecture improves efficiency for inference workloads.
ROCm 7, released in September 2025, introduces native support for the AMD Instinct MI350 and MI325X GPUs, lower-precision data formats (FP4, FP8) for faster AI inference, broader Windows and consumer GPU support, and day-zero integration with tools like PyTorch and vLLM.
“You can go for a little less density, so you can do air-cooled infrastructure versus liquid-cooled infrastructure, and then still get the capabilities that are top of the line,” added Elangovan, highlighting scalability advantages.
Beyond large language models, AMD supports text-to-image and text-to-video AI workloads, with Luma Labs’ Ray3 model fully trained on AMD platforms.
Elangovan emphasised AMD is increasingly a software platform, stating, “You should think of us as a software platform that developers can trust and build on, one that will outlive generations of hardware.”
