HPE Launches New AMD Helios Rack to Increase Large-Scale Computing Power

The new rack combines advanced hardware, strong networking, and efficient cooling to help global businesses scale high-performance systems by 2026.

The AMD Helios AI rack-scale solution, a full rack for large-scale inferencing and AI training, has been announced by HPE. This new AI rack contains 72 AMD Instinct MI455X GPUs, 260 terabytes of bandwidth, and 2.9 AI exaflops of FP4 performance. The first AI rack to use an Ethernet network, and HPE is dedicated to open standards. The new AI solution will launched worldwide in 2026.

HPE and AMD are working together to create AI infrastructure that will enable massive inferencing. The new rack mixes hardware, software, and network connectivity in a single system.

According to HPE CEO, Antonio Neri the launch of the new AMD Helios and their specially designed HPE scale-up networking solution they are giving their cloud service provider clients quicker installations, more flexibility, and lower risk when scaling AI computing in their businesses. Also highlights the importance of open standards.

The partnership with HPE and AMD has been in place for over a decade. Both parties have already produced many exascale systems for supercomputing. The Open Compute Project (OCP) specifications serve as the basis for this architecture. The Open Rack Wide (ORW) design blends energy-efficient liquid cooling with a standard maintenance structure. HPE Services provides the system with straight liquid cooling and experience with exascale installations.