Deutsche Telekom's Industrial AI Cloud to Expand to 10,000 GPUs

en.Wedoany.com Reported - On October 21, German cloud computing platform developer vCluster Labs announced that Deutsche Telekom's industrial AI cloud platform in Munich has deployed its automated management system and integrated NVIDIA DSX OS software components to enable the automated creation and management of enterprise-dedicated AI cloud environments. Currently, the platform manages approximately 500 NVIDIA GPUs through vCluster, and Deutsche Telekom plans to expand the scale to more than 10,000, covering Blackwell and Rubin series accelerated computing devices.

Deutsche Telekom's previously built industrial AI cloud is already operational in Munich. This software deployment focuses on addressing the provisioning and management of customers' independent cloud environments. vCluster Labs, together with NVIDIA and Deutsche Telekom, has integrated underlying GPU servers, network configuration, computing resource allocation, and the customer management platform into a unified automated process. Deployment work that previously required network, platform, and infrastructure teams several days to complete can now be accomplished within hours.

Each onboarded enterprise receives an independent cloud management control platform, a dedicated access address, and an operations monitoring interface, with isolation at the network, storage, and computing resource levels. Through Deutsche Telekom's dedicated portal or API, customers can choose to deploy Kubernetes container clusters, Slurm job scheduling environments, GPU virtual machines, or directly use bare-metal servers. All related workloads and industrial data remain within Germany, and the platform operates in accordance with Deutsche Telekom's data security and sovereignty requirements.

The underlying infrastructure adopts multiple core software components of NVIDIA DSX OS. Among them, NVSentinel is responsible for GPU fault detection and recovery, GPU Operator manages GPU resources in container environments, and Network Operator handles network configuration management; vCluster coordinates these components to automatically complete resource delivery from physical computing devices to customer-dedicated cloud environments. Server resources are managed by the vMetal system, currently supporting NVIDIA RTX PRO servers and DGX B200, with DGX B300 to be included later and planned compatibility with the Vera Rubin NVL72 architecture.

Commercial deployment has already entered the customer usage phase. A major German automotive manufacturer has become the first industrial customer to run an independent private AI cloud on the platform. Deutsche Telekom also plans to gradually introduce the NVIDIA Omniverse software library and the open-source Run:ai computing resource management platform, providing corresponding software environments to enterprises through vCluster Stacks. As additional GPU resources are connected, existing customers can continue to use computing devices of different generations through the unified management interface and APIs.