Hewlett Packard Enterprise is betting artificial intelligence (AI) and machine learning will drive the adoption of high-performance compute (HPC) in the data center. The company today announced it will offer HPC clusters through its GreenLake platform.

The company’s goal is to enable enterprises to run high-performance workloads, like simulation and AI inference or training, without needing to build or maintain a supercomputing cluster.

HPE SVP and GM of GreenLake Keith White described the new offering as “supercomputing-as-a-service.”

“With this offering HPE customers will be able to run their modeling and simulation workloads with a fully managed, pre-bundled, HPC cloud service to operate in any data center or colocation environment,” he said.

GreenLake is the company’s consumption-based cloud portfolio. The platform enables customers to deploy HPE infrastructure wherever they need it whether that be in the data center, a co-location facility, or at the edge.

The platform has seen major success since it launched in 2018. GreenLake orders swelled 82% during the company’s third-fiscal quarter of 2020, and CEO Antonio Neri previously called GreenLake is HPE’s fastest-growing business. At the company’s annual conference last year he pledged to offer HPE’s entire portfolio as a service by 2022.

And true to Neri’s promise, HPE is continuing to expand its product offerings under the GreenLake umbrella. Customers can now essentially order a supercomputing cluster right through a self-service portal, and have it up in running in as little as two weeks.

HPE’s push into HPC with GreenLake is nothing new. The company acquired supercomputing legend Cray last year for 1.3 billion in cash. Founded in 1987, the Seattle-based Cray has, from the beginning, focused on supercomputing, storage, and data analytics. And now, under HPE, Cray’s technologies are getting integrated into HPE’s product portfolio, and GreenLake is just the latest platform to get the HPC treatment.

According to White, the move comes at a time when the public cloud is undergoing strong customer growth thanks in part to the simplicity and automation of the platform. However, many workloads are expected to stay on premises for the foreseeable future.

“IDC expects 70% of our current and future workloads to stay on premises,” he said. “This is because of a variety of factors [like] data gravity, latency, security, app entanglement, compliance, or regulatory challenges.”

And these concerns are also one of the reasons that HPC is apt to stay on premises, he added. This is where GreenLake has an advantage over public cloud providers, White claims. While the platform behaves in much the same way as a public cloud might, the actual infrastructure isn’t locked away in a public cloud data center. Instead, HPE deploys that infrastructure wherever the customer needs it.

This has other advantages such as speed to deploy and total cost of ownership, White explained, claiming the offering would speed up adoption of HPC products by up to 75% and reduce total cost of ownership by a further 40%.

HPE will initially offer HPC services based on its Apollo systems combined with storage and networking technologies that have been purpose built for modeling and simulation. Initial pre-bundled offerings for GreenLake's HPC offerings will launch in spring 2021. Cray-based compute, software, and storage will arrive at a later date.