HPE Unveils AI Supercomputer Cloud Service

Hewlett Packard Enterprise is launching an AI supercomputer cloud service that will enable organizations of all sizes to pursue AI projects without the cost and complexity of having to buy, install, and manage AI-specific hardware themselves.

At its HPE Discover 2023 conference in Las Vegas, the company today announced HPE GreenLake for Large Language Models (LLMs), an HPE-hosted, subscription-based cloud service that will enable enterprises to access HPE Cray XD supercomputers and the AI software they need to build and run large-scale AI models.

"HPE GreenLake for Large Language Models allows our customers to rapidly train, tune, and deploy large language models on demand using a multi-tenant instance of our supercomputing platform — truly a supercomputing cloud combined with our AI software," said Justin Hotard, HPE's executive vice president and general manager of High Performance Computing, AI & Labs, in a news briefing with media.

HPE today also announced a raft of new GreenLake services, including:

Related:5 Ways AI Could Reshape Data Centers

Convergence Between AI and HPC

With the AI supercomputer cloud service, HPE becomes the latest company to announce new AI-specific hardware for enterprises. In late May, Dell announced two AI-optimized servers and a collaboration with Nvidia to help enterprises adopt generative AI applications. Lenovo last week also announced new AI-ready servers for the data center and the edge.

Analysts say HPE's launch of a new AI supercomputing cloud service is a smart move because it makes the combination of AI and high-performance computing (HPC) accessible to everyone.

Alexander Harrowell, Omdia's principal analyst of advanced computing for AI, said there is a convergence happening between AI and HPC. The process of training some AI models can be so computational demanding that they need HPC-class hardware. But not every organization can afford to buy HPC systems for their large-scale AI needs, he said.

Nvidia in May launched a new AI supercomputer for generative AI workloads, and prior to that in March, the company partnered with Oracle Cloud, Microsoft Azure, Google Cloud, and others to make its AI supercomputers available over the cloud. Cerebras and SambaNova Systems have also been selling cloud access of their fully integrated computer cluster systems for AI applications, according to Harrowell.

Related:AI and the Data Center: Challenges and Investment Strategies

"The high ASPs [average selling prices] and high minimum buy-in are a big barrier to customer adoption, so offering this technology as a cloud service is one way around that," Harrowell said.

HPE GreenLake for LLMs

HPE's Hotard said the company hopes to attract enterprises of all sizes — from startups to Fortune 500 companies and even public sector organizations — to use HPE GreenLake for LLMs. Some customers may run AI workloads on-premises, but they may want to turn to the AI supercomputing cloud service for bursting capability, he said.

The company expects HPE GreenLake for LLMs to be available by the end of 2023 in North America and early 2024 in Europe. Users will be able to support AI and HPC jobs on hundreds or thousands of CPUs or GPUs at once, he said.

"We've got a lot of customers and partners already interested," Hotard told the media in a news briefing last week. "We've been doing a … cloud testbed for months now, and we've gotten a lot of positive feedback."

According to HPE, the cloud service will include access to Luminous, a pre-trained LLM from Aleph Alpha, which will allow organizations to leverage their own data and train a customized model to gain real-time insights on their own proprietary knowledge.  

The AI cloud service will also leverage the HPE Cray Programming Environment, a fully integrated software suite of HPC and AI applications that includes tools for developing, porting, debugging, and tuning code, the company said.

Other HPE GreenLake Announcements 

HPE today also announced a set of new services to expand its GreenLake platform, the company's on-premises solutions that are offered through a cloud-like, subscription-based, as-a-service model.

HPE in 2019 announced plans to make every product it sold available as a service — and its success prompted other hardware makers, including Dell and Cisco, to pursue similar strategies. Over time, HPE has added more products and services to GreenLake, including introducing SaaS offerings.

The company also unveiled its enhancements to HPE GreenLake for Private Cloud Enterprise, a managed service introduced last year in which HPE designs, installs, and manages private clouds for customers in their data centers, edge locations, and colocation sites.

The service includes installation, provisioning, firmware updates, maintenance operations, hardware replacement, growth planning, and support. And through a unified management portal, customers can easily track capacity, usage, and costs.

HPE executives today said they are adding support for Amazon Web Services (AWS), Azure, and Google Cloud, allowing customers using the private cloud to also self-provision workloads on the public clouds. In addition, the company added support for two container platforms: Red Hat OpenShift and AWS' EKS managed Kubernetes Service.

In addition, HPE at Discover 2023 rolled out a new version of GreenLake for Private Cloud Enterprise, called the Business Edition, that allows customers to manage the infrastructure themselves.

Some companies don't want a fully managed solution and prefer to manage the infrastructure themselves — and the Business Edition enables that, said Bryan Thompson, vice president of HPE GreenLake Cloud Services Solutions, in a briefing with news media.

The Business Edition will also support public clouds, allowing customers to spin up VMs on their private clouds and on public clouds, he added.

HPE announced that Equinix will offer both versions of HPE GreenLake for Private Cloud Enterprise as a pre-provisioned offering in its data centers worldwide. Equinix will begin offering the service In August, HPE said.

In other HPE Discover news, the company announced: