DDN Partners With Supermicro and Solidigm on Enterprise AI Storage Platform

DataDirect Networks Inc. (DDN), working alongside hardware manufacturer Super Micro Computer Inc. and storage supplier Solidigm, has unveiled the DDN Enterprise AI HyperPOD. Designed around Nvidia Corp.’s AI Data Platform reference architecture, the unified system is engineered to reduce the friction of configuring, deploying, and scaling artificial intelligence inference across enterprise computing environments.

The joint initiative comes as enterprise IT departments struggle with the operational overhead of assembling custom hardware stacks. Historically, enterprises have purchased compute nodes, networking gear, and storage arrays as individual components and spent months integrating them. Speaking in an interview series broadcast on SiliconANGLE Media’s theCUBE during the Supermicro Open Storage Summit, Supermicro Director of Storage Solutions Michael Ang noted that this piecemeal approach significantly delays how quickly companies can monetize internal data assets.

The HyperPOD framework addresses these integration bottlenecks by offering an all-in-one platform built for local datacenter installations. By basing the hardware bundle on Nvidia’s standardized blueprint, the platform allows enterprises to deploy AI workloads on-premises while keeping tight controls over data residency, compliance, and governance. The system also incorporates native multi-tenancy capabilities, allowing multiple business units or project teams to share central compute and storage pools while keeping their underlying datasets completely segregated.

A key focal point of the platform's architecture is optimizing hardware utilization, particularly for high-cost graphics processing units. In traditional setups, expensive GPUs frequently experience idle cycles while waiting for storage subsystems to transfer necessary datasets. Andrew Murphy, senior director of product management at DDN, explained during the broadcast that the platform aims to eliminate these performance bottlenecks by providing a high-speed pipeline that ensures GPUs remain continuously supplied with data, thereby maximizing return on enterprise IT expenditures.

The shifting nature of artificial intelligence workloads has also altered the balance between compute and storage requirements. As large language models become standard across commercial applications, storage performance has emerged as a primary constraint. Pompey Nagra, who leads products and ecosystems at Solidigm, explained that large models depend heavily on key-value cache data. Storing this information on solid-state drives allows systems to reuse cached AI workloads without triggering expensive recomputation on high-bandwidth memory or compute cores.

The collaboration also targets growing enterprise and public-sector demand for private and sovereign AI infrastructure. Organizations are increasingly reluctant to send proprietary data or intellectual property into shared cloud environments due to regulatory requirements and unpredictable operational costs. According to Nagra, hosting AI workloads on dedicated hardware—either on-premises or within hybrid environments—gives managers granular authority over operational spending, regulatory compliance, and proprietary data assets.

To lower financial barriers to entry, the Enterprise AI HyperPOD supports a modular scale-out strategy. Rather than forcing organizations to purchase extensive, high-cost GPU clusters during initial setup, the architecture enables deployments to start with a single rack configuration. Ang detailed how businesses can test and deploy workloads at a modest scale and subsequently expand infrastructure by simply connecting additional hardware racks to the network fabric as processing and storage demands grow, avoiding operational downtime.