AMD recently revealed its strategy for an open AI ecosystem at its Advancing AI 2025 event, showcasing new chips, software, and systems. AMD highlighted its unique position to power all aspects of AI by combining its leading GPUs, CPUs, networking, and open software for unmatched flexibility and performance.
AMD outlined its comprehensive AI platform vision and introduced a new, open, scalable rack-scale AI infrastructure. Some major announcements included:
- New AMD Instinct MI350 Series Accelerators: These new GPUs significantly boost performance for generative AI and high-performance computing, offering 4 times faster AI computing and 35 times faster inferencing than previous versions. The MI355X also provides better performance for the price.
- Growing AMD ROCm Ecosystem: The latest ROCm 7 software is designed for generative AI and high-performance computing, making it easier for developers to work with.
- Open Rack-Scale AI Designs: AMD showcased its current and future AI rack infrastructure, already being used by partners like Oracle Cloud Infrastructure (OCI). They also previewed “Helios,” their next-generation AI rack, which will feature the AMD Instinct MI400 Series GPUs, expected to deliver up to 10 times more inference performance for complex AI models.
- Improved Energy Efficiency: AMD surpassed its goal of making AI training and high-performance computing nodes 30 times more energy-efficient, achieving a 38-fold improvement. They also set a new goal for 2030: a 20-fold increase in rack-scale energy efficiency, meaning AI models could be trained using 95% less electricity and far less space.
- AMD Developer Cloud: This new cloud platform gives developers and the open-source community a fully managed, high-performance environment for AI development.
Strong Partner Collaborations
AMD demonstrated strong partnerships with major AI companies. Companies like Meta, OpenAI, xAI, Oracle, Microsoft, Cohere, HUMAIN, Red Hat, Astera Labs, and Marvell discussed their collaborations with AMD for AI solutions.
- Meta is widely using Instinct MI300X for its Llama 3 and Llama 4 AI models and is looking forward to the MI350 series.
- OpenAI CEO Sam Altman emphasized the importance of optimized hardware, software, and algorithms, noting their partnership with AMD for AI infrastructure, with GPT models running on Azure with MI300X.
- Oracle Cloud Infrastructure (OCI) is among the first to adopt AMD’s open rack-scale AI infrastructure with MI355X GPUs, planning to offer massive AI clusters.
- Microsoft confirmed that Instinct MI300X is now powering both its own and open-source models on Azure.
AMD’s CEO, Dr. Lisa Su, stated that AMD is rapidly advancing AI through open standards, shared innovation, and growing leadership with its hardware and software partners.
