Just a few short months ago, Hewlett-Packard Enterprise (HPE) announced its $1.3 billion acquisition of supercomputing leader Cray. The acquisition marked a curious venture for the HPE, who — as the lead supplier in the global high-performance computing market — had traditionally set their ambitions towards more low-key, less expensive HPC solutions, servicing mostly commercial markets. Meanwhile, Cray has had their hand in the very highest of the high-end of HPC implementations, even lending a helping hand in the development of the world’s first ‘Exascale’ supercomputer for the US Department of Energy, known as Aurora.
However, their difference in clientele ultimately amounted to a match made in heaven, with HPE receiving access to high-end markets and Cray’s advanced technologies, such as its new high-speed interconnect, known as Slingshot. Although, as much as Cray’s acquisition highlighted an interest of HPE to enter into the public-sector dominated high-end HPC markets, it also provides a bit of a boost and increased validation for its private sector offering and – along with it – the commercial/private HPC market in general.
This is particularly evident on a technological level, as Cray’s Shasta architecture (i.e. the one that powers Aurora) provides capacity for the use of multiple processor types, including x86, ARM, and, significantly, GPUs. It’s easy to see how the private sector would benefit from such technology, given the increasing prominence of machine learning technologies in enterprise organisations, and noting ML’s dependency on GPU-accelerated processing.