Buying Supercomputer Pioneer Cray, Hewlett Packard Enterprise Ups Its High Performance Game


Hewlett Packard Enterprise’s $1.3 billion purchase of Cray, the pioneer supercomputer maker, intends to expand its reach into the high performance computing-as-a-service market for enterprise customers.

The dynamic market involves data-intensive workloads that are growing in number and volume, in turn forcing many large corporations to migrate their on-premise data centers to the cloud. The enterprises find the economies of as-a-service cloud offerings difficult to ignore.

Hewlett Packard Enterprise — commonly known as HPE, the servers and storage firm spun out of the 2015 breakup of Hewlett Packard — last week agreed to acquire the Seattle-based Cray, the manufacturer of massive rack-and-server systems for crunching terabits of data at speeds surpassing most systems.

Despite recently winning a $600 million contract to build the world’s fastest supercomputer for the Oak Ridge National LaboratoryOpens a new window in Tennessee and a $500 million contract for a system at the Argonne National Laboratory in Illinois, Cray’s business has weathered sharp swings.

Such big-ticket contracts for supercomputers involve long time frames buffeted by various economic and political winds. Cray has posted a string of quarterly losses, and with more of the same forecast going forward, its board approved the sale.

Nevertheless, the deal stands to benefit both Cray and HPE.

For HPE, the acquisition expands the San Jose, California-based company’s reach into government and academic markets and boosts its considerable presence in the markets for on-premises storage and computing and the cloud platforms that connect them with edge devices.

HPE, which also builds supercomputers, reported $30.7 billion in annual revenue for 2018 and has formed partner relationshipsOpens a new window with Amazon Web Services, Microsoft Azure and Google Cloud.

For Cray, the dealOpens a new window paves the way for a new level of service offerings such as high-speed securities trading, pharmaceutical research and artificial intelligence-machine learning.

Cray’s sales have focused on powerful exascale machines for U.S. intelligence agencies, government laboratories and universities.

Last June, HPE dedicated $4 billion for investments in its Intelligent Edge product and services suite. Its $1.3 billion Cray investment compliments the $275 million HPE paid for SGI, another legacy maker of high-performance computing engines.

SGI, formerly known as Silicon Graphics, was born in 1982. It pioneered graphics processing architectures that have become centerpieces of the contemporary artificial intelligence and machine learning applications that HPE aims to execute for its enterprise customers. The OpenGLOpens a new window graphics language-and-library API that SGI released 26 years ago remains an industry utility.

Cray’s history, which includes a period as an SGI subsidiary, dates back still further. The company continues to manufactures its Shasta and XC product lines in Chippewa Falls, Wisconsin, where it was founded by engineer Seymour Cray in 1972.

Cray’s Shasta architecture, which can be cooled by air or water, is capable of supporting both complete and reduced instruction-set CPUsOpens a new window , as well as GPUs with or without field-programmable gate arrays.

Shasta’s compatibility with chipset designs and standards, as well as a variety of storage media, provides customers with the ability to tailor machines for dedicated workloads, the company says.

Tying together and shepherding data flows is the Slingshot interconnectOpens a new window , an adaptive Ethernet with a congestion-control feature to optimize processing performance. Cray’s Shasta machines can achieve exascale computation speeds of a quintillion calculations per second.

HPE says it plans to target enterprise users with data-intensive workloads, where high performance computing is needed to process multiple models and simulations simultaneously.

In a press releaseOpens a new window announcing the deal, the company said it anticipates compound growth in the market for high performance computing services of 9% per year, to $35 billion, by 2021.

HPE’s as-a-service portfolio has issued a slew of dedicated offerings in the last 18 months, including those for blockchain and edge processing.

The corporation intends to push high performance computing in the medical and climate research areas as well as build on Cray’s presence in the energy and security industries.