In terms of specifications, the AMD Instinct MI430X is based on the next-generation AMD CDNA architecture and comes with 432GB of HBM4 memory with 19.6TB/s of memory bandwidth. Thanks to HBM4 memory and high bandwidth, it is capable of training large language models or running complex simulations without bottlenecks. Thanks to FP4, FP8, and FP64 precision support it provides balanced performance for both AI and scientific applications. AMD was keen to note that, combined with AMD ROCm software, the MIX430X provides full-stack compatibility and scalability across data center and supercomputing environments.

As noted and detailed by AMD, the new Instinct MI430X will power both the United States' first AI Factory supercomputer, called Discovery, and the recently announced Exascale-class system in Europe, called Alice Recoque. The Discovery will use next-gen AMD EPYC Venice CPUs on HPE Cray GX5000 platform, while the Alice Recoque is based on the same CPU but on Eviden's BullSequana XH3500 platform.