AMD expands AI portfolio with EPYC 9006, Intuition MI400 and new bodily‑AI {hardware}
AMD has introduced new additions to its knowledge‑centre portfolio at Advancing AI 2026, introducing each the sixth Gen EPYC 9006 Sequence server CPUs and the Intuition MI400 Sequence GPUs, geared toward next-generation AI infrastructure.
The EPYC 9006 Sequence is described by AMD as a workload‑optimised CPU lineup constructed to help agentic AI, the place CPUs tackle a bigger share of labor by working brokers, coordinating duties and managing knowledge flows. AMD claims the brand new technology is designed to deal with this elevated load by way of increased core counts, expanded reminiscence bandwidth and help for subsequent‑technology I/O requirements akin to PCIe Gen 6 and MRDIMM.
The EPYC 9006 Sequence consists of 4 CPU households. The SP7 tier is the utmost‑density choice, with core counts as much as 256 cores and 512 threads, together with excessive‑frequency variants reaching as much as 5GHz. SP8 fashions give attention to effectivity, providing as much as 128 core configurations designed for enterprise, edge and energy‑constrained deployments. For HPC and reminiscence‑delicate workloads, AMD factors to the SP7X variants, which it says provide triple the L3 cache per core in comparison with SP7. The LP tier serves as an AI host‑node CPU, utilizing LPDDR reminiscence and better frequencies to maintain accelerators provided with knowledge in multi-GPU racks.
On the GPU facet, AMD launched the Intuition MI400 Sequence, cut up between the MI455X for frontier‑scale AI deployments and the MI430X for sovereign AI and HPC. The MI455X is AMD’s highest‑finish accelerator for giant‑mannequin coaching, wonderful‑tuning and excessive‑quantity inference, whereas the MI430X is geared toward workloads requiring numerical accuracy. AMD says the MI430X can ship as much as 288 TFLOPS of FP64 efficiency, alongside HBM4 reminiscence and excessive‑bandwidth interconnects to help bigger fashions.
The MI400 Sequence additionally incorporates safety and system‑degree options for delicate AI and HPC environments, together with SecureBoot, encrypted GPU‑to‑GPU hyperlinks and {hardware}‑based mostly protections. These GPUs help AMD’s ROCm software program stack for AI growth, which can be receiving updates this week by way of ROCm.ai – a brand new AI‑native developer surroundings that provides a unified CLI, AMD‑authored steerage for coding assistants and AI‑pushed workload optimisation.
Lastly, AMD can be increasing into bodily AI and robotics, introducing new Ryzen AI Embedded X100 processors and Kria AI system‑on‑modules for autonomous machines and industrial methods. These platforms mix Zen 5 CPU cores, built-in RDNA graphics and an NPU, with shared entry to reminiscence to cut back latency in actual‑time workloads.
KitGuru Says: What do you consider AMD’s push into the AI area?