High-performance computing (HPC) is seeing stronger adoption than ever, thanks to the rise of HPC-as-a-service offerings from the major cloud providers, Dell, Hewlett Packard Enterprise, and Lenovo, according to GlobalData analysts.
“HPCaaS made history during the pandemic, making the incredibly quick turnaround of a COVID-19 vaccine possible,” wrote Beatriz Valle, senior analyst for enterprise technology and services at GlobalData, in a research note.
Not long ago, HPC was limited to those with deep pockets and a steady supply of scientific workloads necessary to justify such an expense.
Today, scientists rely on HPC clusters to track and predict climate change, sequence DNA, and manufacture safer vehicles. And, according to GlobalData, the ability to rent these resources on a whim without the upfront cost has made this possible for a wider range of customers.
“Setting up and maintaining on-premises HPC facilities is very expensive, which means the majority of HPC deployments have traditionally taken place in the government sector, research, or academia,” Valle wrote. “HPCaaS unlocks HPC to private enterprises through flexible consumption models and through the cloud.”
GPU and custom-silicon accelerated cloud instances have become commonplace in recent years as demand for artificial intelligence (AI) training and inference have gained mainstream appeal.
This consumption model has fueled similar offerings from leading original equipment manufacturers (OEMs) like HPE, Dell, and Lenovo, which have all launched their own HPC-focused products.
Demand is strong enough that HPE, which acquired supercomputing legend Cray in 2019, raked in more than $1 billion from HPC-related sales during the fourth quarter of 2021.
“HPE is at the center of several compelling megatrends: the explosion of data at the edge, the mandate for a cloud experience everywhere, and the need to extract value from data to generate insights,” CEO Antonio Neri said on the company’s Q4 earnings call.
However, HPE isn’t the only OEM in the game. Fujitsu is the latest to join the party with a custom CPU design pioneered on the Fugaku supercomputer in Kobe, Japan – the world's fastest supercomputer in terms of computing speed. Fujitsu’s Cloud Service HPC offering enables customers to run applications on the company’s PrimeHPC FX1000 servers.
Valle expects demand for these services to drive worldwide sales of high-end servers, with the market expected to reach $9 billion by 2024.
“Demand for HPCaaS continues apace, and this is underpinning the high-end server market,” Valle wrote, adding that this will only accelerate as hyperscale vendors roll out new HPC offerings and Arm gains a stronger foothold in the data center market.
Comments