Skip to main content

US government awards millions to HPE, Intel, and others in hopes they’ll build next-gen supercomputers

US government awards millions to HPE, Intel, and others in hopes they’ll build next-gen supercomputers

/

US ramps up exascale digital arms race

Share this story

HPE

To try and edge the United States ahead of other countries in the arms race of building supercomputers, the Department of Energy today awarded a total of $258 million in funding to HPE, Cray, AMD, Intel, IBM, and Nvidia. The money will be put toward developing exascale computers, which are capable of a billion billion calculations per second. 

The $258 million in funding will be allocated over a three-year contract period. Each company has also agreed to provide at least 40 percent of their individual total project costs, for a total investment of $430 million. Exascale computing can transcend the limits of regular computing. These limits include big data too large to ingest, outdated systems, and large energy consumption.

Engineers are concentrating on increasing the energy efficiency of these machines while boosting memory capacity for overall higher performance levels. They estimate they’ll need systems that are 10 times faster than today’s very best prototypes. The sheer performance capabilities of exascale computing could potentially deliver breakthroughs in many areas, ranging from the financial sector to scientific discovery.

Titan supercomputer

Bill Mannel, VP and general manager of HPC Segment Solutions at HPE, told The Verge that this research grant had been in the works as part of HPE’s continued work with supercomputers. “We expected the grant as we’ve been working on this for a while,” said Mannel. “It did not come as a surprise to us.” 

In May, HPE announced the largest single-memory computer to date and the company’s biggest R&D program, calling it The Machine. HPE estimates that the memory-driven computing architecture could scale to 4,096 yottabytes, or 250,000 times the size of the digital universe today. 

This isn’t the first of the DOE’s investments in this competitive global landscape. In 2011, it awarded a separate contract of $97 million to Cray and Nvidia for building the world’s fastest supercomputer named Titan.

While the DOE’s aim is to develop at least one exascale-capable system by 2021, the current model of supercomputer is still lagging behind. Titan is ranked third in the world, two ranks behind China. China currently estimates finishing an exascale supercomputer by 2020.

The other US-based companies that won the grant besides HPE will be in friendly competition, with the exception of one ally which Mannel wouldn’t name. Although developments may be decades away from commercialization, the final plan is to make these technologies available to the public, not just for military and government use.

Correction: June 15th, 8PM ET: This story has been updated to correct the definition of exascale computing. It’s a billion billion calculations per second, not a billion calculations per second.