Lenovo adds GPUaaS for on-premises AI workloads

GPUaaS

Lenovo TruScale GPUaaS is a new feature for Lenovo. The service uses Nvidia GPUs, including Nvidia H100 and Nvidia L40S, and offers metered resources through Lenovo Intelligent Computer Orchestration, which is software for AI model development and training. With the software, GPU workloads and costs can be managed, Lenovo said.

Dell and HPE both have provided similar services for years. Initially, Dell and HPE focused on high-performance computing (HPC) customers, but that has changed with the onset of generative AI. In March, Dell expanded its partnership with Nvidia and introduced its AI Factory, a combination of infrastructure, partner ecosystem and professional services aimed at helping customers with AI workloads. In June, HPE introduced HPE Private Cloud AI, which also partners with Nvidia to combine GPUs with HPE hardware and software into a private cloud for customers.

A lot of people are going to use private AI use cases where they don’t want all their data, their IP, going to the public cloud.

But Lenovo offers a level of specificity beyond the Dell and HPE offerings, according to Matt Kimball, an analyst at Moor Insights & Strategy.

“Lenovo is more surgical in their approach of their GPU-as-a-service offering,” he said.

For AI and GPU usage, Dell sells infrastructure and partner building blocks that customers need to assemble, while HPE sells preconfigured offerings in various sizes. However, Lenovo is offering a more fine-grained approach, Kimball said.

Heavily regulated industries such as healthcare, financial services and insurance will use AI models on premises and will use TruScale GPUaaS to lower costs, Dickens said.

“Instead of paying by the sip, Lenovo enables five or six business units to use the GPUs,” he said.

Adding AI to management

Lenovo isn’t just thinking about AI use; it’s also using AI in its XClarity One management-as-a-service offering. It added predictive analytics to identify infrastructure issues before they fail in an effort to reduce downtime. Lenovo is also embedding generative AI into some of its security capabilities to conduct continuous learning for risk prevention, according to the vendor.

Adding AI to software for efficiency isn’t new in the enterprise infrastructure market, Kimball said. For example, HPE ramped up adding AI and automation through its acquisition of Morpheus, a hybrid cloud management platform with automation capabilities.

“They’re driving a more self-driving, self-healing, greater-protected infrastructure through AI,” Kimball said.

With fewer people and lower budgets, complexity becomes the enemy, while automation becomes a benefit to customers, he said.

Liquid cooling

AI and GPU usage also means increased energy consumption, as GPUs take a lot of energy to run. Lenovo’s Neptune liquid cooling technology has been in use for over a decade and is currently in its sixth generation. This week, Lenovo introduced its Power and Cooling Services program that supports energy and thermal needs through the use of Neptune. The new services aid in hardware and facility integration for liquid cooling, as well as data-driven insights to better determine the data center’s power needs.

Liquid cooling has moved from more niche HPC and edge use cases to the mainstream now that power-hungry GPUs are in the mix, according to Dickens. Other vendors are looking to catch up, but Lenovo has a significant lead in development.

“In its sixth generation, there’s probably good battle scars from the previous five generations [of Neptune liquid cooling] that are valuable to clients,” Dickens said.

Adam Armstrong is a TechTarget Editorial news writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.