Latest news with #AmazonAWS

Forbes

Is Nvidia Competing With Its GPU Cloud Partners?

Nvidia Headquarters in Santa Clara, CA. Nvidia recently announced two new cloud initiatives. First, the company announced DGX Cloud Lepton, designed to connect artificial intelligence developers with Nvidia's wide network of cloud providers. Second, Nvidia announced a new cloud service, the Industrial AI Cloud, intended to provide AI services to manufacturing companies in Europe. While these moves pit Nvidia against its cloud partners, the larger cloud service providers (CSPs) chose to compete with Nvidia using their in-house developed GPU alternatives. Google has the TPU, Amazon has Trainium, Microsoft has Maia, etc. (Nvidia is a client of Cambrian-AI Research.) Turn about is fair play, and Nvidia is helping its cloud partners sell AI services that keep their GPUs running at high utilization, maximizing profit, while also helping developers access a broader inventory of rare and expensive GPUs. Much to the consternation of its cloud partners, Nvidia launched the new DGX Cloud Lepton service at Computex this year, and has already garnered a healthy suite of CSPs to agree to join the service. While Oracle and Google have yet to sign up publicly for Lepton, Amazon AWS and Microsoft Azure have done so. They see the benefits of having their clouds accessible and promoted by Nvidia. The smaller GPU cloud players have also joined the party, including CoreWeave, Crusoe, Firmus, Foxconn, GMI Cloud, Lambda, Yotta Data Services, Nebius, Nscale, Firebird, Fluidstack, Hydra Host, Scaleway, Together AI, Mistral AI, SoftBank Corp. These providers offer both on-demand and long-term GPU access, supporting a wide range of AI development and deployment needs. Other CSPs won't want to miss the train, and will likely join soon. At the Paris GTC, Nvidia CEO Jensen Huang announced that Nvidia and Deutsche Telekom were building an AI Cloud for European manufacturing companies. The Industrial Cloud will provide access to state-of-the-art AI infrastructure and Nvidia's rich portfolio of software. Support will be available for CAD, CAE, Omniverse, Robotics, and Autonomous Vehicles. The cloud is fully configured to support Nvidia's optimized enterprise AI software portfolio, and should be open for business in early 2026. Nvidia's Industrial Cloud for Europe represents a major step in building sovereign, AI-powered infrastructure for the continent's industrial sector. By providing secure, high-performance compute resources and a robust AI software ecosystem, the initiative aims to propel European manufacturing into the next era of digital innovation Nvidia is partnering with Deutche Telekom to build the first Industrial AI Cloud for European ... More manufacturing companies. The Industrial Cloud will be powered by 10,000 Nvidia GPUs, including the latest DGX B200 systems and RTX PRO servers, making it one of the largest industrial AI deployments in Germany. Think of this as a manufacturing-focussed sovereign data center managed and operated by Deutsche Telekom, ensuring data sovereignty and compliance with European regulations, addressing concerns about dependency on non-European cloud providers. The lack of NVL72 racks tells us that Nvidia expects customers to fine-tune and serve AI inferencing, not create new foundation models. Users will have access to Nvidia's CUDA-X libraries and workloads accelerated by Nvidia GPUs and Omniverse, supporting a wide range of industrial applications such as simulation, digital twins, robotics, design, engineering, and factory planning. The cloud will also support applications from leading industrial software providers including Siemens, Ansys, Cadence, and Rescale, enabling advanced manufacturing workflows for companies such as BMW, Maserati, Mercedes-Benz, and Schaeffle. First, it says that Nvidia isn't afraid to compete with its cloud partners in its quest to provide access to state-of-the-art AI infrastructure to its end users. As we noted, the larger CSPs chose to develop competing AI accelerators, so they should not be surprised. Second, in reality Lepton doesn't compete with CSPs; it provides aggregated access to their massive arrays of Nvidia GPUs, not a cloud that is owned and operated by Nvidia. And the Industrial Cloud is filling a gap left by the CSPs to provide focussed and sovereign resources for the European manufacturing base. Customers will love it, and so will the ISVs whose software has been optimized to run on Nvidia GPUs.