AMD’s Instinct MI350P Targets Enterprise AI GPU Adoption

AMD is pushing to democratize AI compute with the new Instinct MI350P GPU, designed for standard air-cooled enterprise servers. The chip packs 144 GB of HBM3E memory and runs at 450 watts, targeting inference workloads where memory bandwidth drives latency and user experience.

The MI350P is paired with ROCm 7 and AMD Inference Microservices, providing out-of-box performance for popular open-weight AI models without requiring specialized expertise.

To lower entry barriers, Dell is offering the hardware for remote proof-of-concept testing through its Customer Solution Center, allowing enterprises to try before buying.

“You don’t have to buy the huge server to try it and prove that it works,” said Mike Darby, senior manager of business development for Instinct Data Center GPUs at AMD, in an interview at Dell Technologies World 2026.