Oracle and NVIDIA push AI on OCI with Blackwell GPUs and Italian Velvet models | Hardware computer list | Is gpu a graphics card | Hardware and software of computer | Turtles AI

Oracle and NVIDIA push AI on OCI with Blackwell GPUs and Italian Velvet models
Almawave Showcases Its Multilingual AI in Paris, While Oracle Integrates NVIDIA AI Enterprise and Enhances Cloud Infrastructure for Advanced Training and Inference
Editorial Team12 June 2025

 

Oracle and NVIDIA strengthen their collaboration, integrating NVIDIA AI Enterprise natively on OCI and introducing Blackwell GPUs for advanced AI training and inference. Almawave presents Velvet, multilingual generative models developed in Italy using OCI infrastructure and NVIDIA Hopper GPUs.

Key points:

  • NVIDIA AI Enterprise is now available directly in the OCI console, simplifying access to over 160 AI tools.
  • Up to 131,072 Blackwell GPUs available to support very intensive AI workloads.
  • The Italian company uses OCI and NVIDIA GPUs to develop multilingual LLMs focused on European languages, including Italian.
  • Oracle becomes one of the first hyperscalers to integrate DGX Cloud Lepton, offering flexible access to GPU resources worldwide.


Here is an opportunity that confirms how the synergy between advanced cloud infrastructures and next-generation AI leads to new frontiers of innovation: during the NVIDIAGTC conference underway in Paris–VivaTech on June 12, 2025, Oracle and NVIDIA announced a strengthening of their collaboration, extending the native availability of NVIDIAAAIEnterprise in the Oracle Cloud Infrastructure (OCI) Console, activating GB200 NVL72 liquid-cooled systems on OCI Supercluster with up to 131072 Blackwell GPUs and integrating OCI with NVIDIA’s DGX Cloud Lepton platform for global access to GPU resources; the Italian demonstration on stage came from Almawave, which has long been committed to building a multilingual generative AI in the Velvet family based on OCI infrastructure and NVIDIA Hopper GPUs, with a focus on European languages ​​including Italian, French, German, Portuguese and Spanish. Almawave, through its AIWave platform, uses these models with fine-tuning and inference workloads, confirming how a “sound and performing” AI design can also be born in national contexts integrated into European regulations – as Valeria Sandei, CEO of Almawave, underlined. To support this innovation, OCI offers a complete range of bare-metal and virtual GPU shapes (including Blackwell, Hopper, H100, L40S) that ensure high performance, ultra-low latency RDMA networking and distributed cloud options compliant with sovereign requirements – including no-code deployment with AI blueprints and integration with NVIDIA NIM containers. Furthermore, the transition through DGX Cloud Lepton guarantees developers modular and global access to reasoning training, AI agents and digital twins, a choice also confirmed by other cases such as the Danish startup Cerebriu in the field of medical diagnostics.