With HPE Private Cloud AI you can run all your AI workloads on-premises

With HPE Private Cloud AI you can run all your AI workloads on-premises

Hewlett Packard Enterprise (HPE) is hosting its HPE Discover 2024 conference in Las Vegas this week. Again, this conference is mostly about AI; however, with its private cloud AI, HPE offers a solution where organizations can keep their data within their own data center. It has teamed up with Nvidia to develop a complete hardware and software stack.

If you want to use or develop LLMs, you will have to rely on hyperscalers. They currently have most of the infrastructure to do this quickly and efficiently. HPE wants to change that with its new Private Cloud AI solution, which it developed with Nvidia. All data can remain on-premises, whether you want to build LLMs, apply inferencing, or deploy RAG. HPE has a variety of configurations that allow organizations to use AI very efficiently.

Also read: HPE updates GreenLake For Private Cloud Enterprise

HPE has teamed up with Nvidia to perfect its stack

Combining innovations and solutions from both organizations makes the HPE stack much more efficient. As a result, LLMs could be built much faster and handle more AI processes than competing infrastructure. Private Cloud AI is also more energy efficient, according to HPE.

It combines Nvidia’s AI computing, networking, and software with HPE’s AI storage and compute technology. In addition, the new AI solutions are fully integrated into HPE Greenlake cloud. This means that while all infrastructure and data are running on-premises, organizations get a simple cloud interface for management. This allows organizations to configure and run new AI workloads quickly.

HPE President and CEO, Antonio Neri, said, “Generative AI holds immense potential for enterprise transformation, but the complexities of fragmented AI technology contain too many risks and barriers that hamper large-scale enterprise adoption and can jeopardize a company’s most valuable asset – its proprietary data.”

“To unleash the immense potential of generative AI in the enterprise, HPE and NVIDIA co-developed a turnkey private cloud for AI that will enable enterprises to focus their resources on developing new AI use cases that can boost productivity and unlock new revenue streams”

Brightly colored digital graphics with logos of Hewlett Packard Enterprise and NVIDIA in the center, illustrating a collaboration or partnership between the two technology companies.

Advanced software stack

The foundation of the software stack consists of Nvidia’s AI Enterprise software platform. In this platform, Nvidia NIM plays a major role, inferencing microservices. Nvidia AI Enterprise accelerates the development of data science pipelines, the development and deployment of copilots, and other GenAI applications.

HPE has developed the HPE AI Essentials software to complement Nvidia AI Enterprise and Nvidia NIM. These HPE AI Essentials provide several composite AI and data foundation tools with the GreenLake control panel. With these, data pipelines can be modified and kept compliant. Perhaps more importantly, all AI pipelines are explainable and reproducible, providing high reliability.

AI software stack cannot exist without an AI hardware stack

Complementing the AI software stack is the hardware stack, which, too, consists of technology from both organizations. The HPE Private Cloud AI infrastructure consists of Nvidia Spectrum-X Ethernet networking, HPE GreenLake for File Storage, HPE Proliant servers, and support for the Nvidia L40S, Nvidia H100 NVL Tensor Core GPU, and Nvidia GH200 NVL2 in various configurations.

For example, HPE states that the HPE Cray XD670 will support eight Nvidia H200 NVL Tensor Core GPUs, which is ideal for LLM developers.

The HPE ProLiant DL384 Gen12 server with the Nvidia GH200 NVL2 is ideal for organizations consuming large LLMs and for RAG.

The HPE ProLiant DL380a Gen12 offers support for as many as eight Nvidia H200 NVL Tensor Core GPUs, making it ideal for LLM users who want more flexibility to scale their genAI workloads.

Prices and ROI are not for every organisation

Looking at the configurations that HPE lists, it is safe to say that this will not be feasible for every organization. HPE has not announced prices, but this is only for very large organizations. If, as a large organization, you have the workloads and likely additional compliance requirements to keep data on-premises, then the price tag won’t be the issue. SMBs will still have to rely on hyperscalers for now.


HPE expects to deliver customers the first HPE Private Cloud AI systems this fall. Considering how much technology they have had to put together, this is still pretty fast.

Right now, HPE and Nvidia are busy training all sales teams and channel partners. Parties like Deloitte, HCLTech, Infosys, TCS, and Wipro should soon be able to quickly get large organizations up and running with these Private Cloud AI systems.

Fulfilling the potential of genAI quickly simply requires support from partners. In addition, genAI innovation does not stand still, and the world may look very different in a few months. That is what HPE is trying to anticipate on with this new infrastructure, which, in theory, should be able to handle all current and future AI workloads.