Deutsche Telekom partnered with Nvidia and vCluster Labs on 21 October to provide private AI clouds for German industrial customers, with the operator expecting to expand the platform from 500 vCluster-supported Nvidia GPUs to more than 10,000, SDxCentral reported. Unveiled at the European leg of Nvidia’s GTC event in Berlin, the arrangement adds isolated customer environments to Deutsche Telekom’s Industrial AI Cloud.
Key points
- Deutsche Telekom expects vCluster-supported Nvidia GPU capacity to rise from 500 to more than 10,000.
- Each customer environment has a dedicated control plane on a shared platform.
- A German automotive manufacturer is the first private sovereign AI cloud customer.
DSX OS underpins separate customer clouds
The architecture combines Nvidia’s open-source DSX OS, software for designing and operating AI infrastructure, with vCluster’s cloud management platform. VCluster supplies each customer with a dedicated control plane — the layer used to manage a computing environment — for Kubernetes, Slurm or inference clusters. The three companies designed the architecture for Deutsche Telekom’s Nvidia-backed Industrial AI Cloud, which came online in February.
Andreas Falkner, Deutsche Telekom’s head of industrial AI cloud, said the arrangement gives customers private environments on a single platform spanning the generations of Nvidia GPUs that the operator runs. That lets Deutsche Telekom offer separate customer clouds while operating the underlying platform itself. Nvidia launched DSX OS less than six months before the Berlin announcement, according to a vCluster release distributed by Business Wire.
Deutsche Telekom launched the Industrial AI Cloud with Nvidia in Munich. Its use of DSX OS follows Nvidia’s decision to make the full Run:ai platform open source within DSX AI Factory, as AI Affairs previously reported. VCluster Labs, formerly Loft Labs, was co-founded by Lukas Gentele and Fabian Kramm, SDxCentral reported.
vCluster targets customer provisioning time
VCluster says its automation covers the steps between the underlying hardware and a customer’s private cloud. It says work that previously took days across network, platform and infrastructure teams can now provision a new environment at the click of a button. The company also says new customers can be brought online in hours rather than days.
According to vCluster, each environment has its own URL, cloud console and monitoring tools, as well as an isolated network. The company says the architecture is designed to address Deutsche Telekom’s security and sovereignty requirements, including Germany’s C5 cloud security criteria.
Gentele said Deutsche Telekom’s Industrial AI Cloud was where Nvidia DSX OS first entered production at scale. That is vCluster’s account of the deployment. The operational change the company describes is repeatable provisioning for individual customers, rather than a separate manual build for each environment.
Deutsche Telekom’s first automotive customer
The customer proposition is aimed at industrial organisations seeking dedicated control over AI infrastructure and data. Deutsche Telekom is expanding the cloud for German industry, while the planned increase in vCluster-supported GPUs would give it capacity for more workloads on the same platform. The expansion remains an expectation rather than a reported completed deployment.
Germany and Nvidia signed a declaration to develop industrial AI models, AI Affairs reported. Deutsche Telekom’s arrangement concerns the cloud infrastructure and customer environments on which industrial workloads can run.
Deutsche Telekom already has its first private sovereign AI cloud customer, an unnamed German automotive manufacturer, SDxCentral reported.
