Fastweb + Vodafone is expanding the computing capacity of its NVIDIA-powered infrastructure and strengthening its AI offering with NVIDIA AI software to further drive the digital transformation of Italian businesses and the Public Administration through new technologies and reinforce the country's digital sovereignty.
Thanks to the collaboration with NVIDIA, the company further consolidates its positioning as a leader in AI in Italy, offering a sovereign and secure digital ecosystem, with increasingly advanced models, high-performing and reliable network infrastructures, and Cloud and cyber security service platforms, all localized on national territory.
Fastweb + Vodafone is expanding in the next weeks the computing capacity of its Supercomputer through NVL8 B300 devices, giving business customers and public administrations greater processing power and is also strengthening its ability to develop, optimize and operate AI services. The company has integrated NVIDIA software - NVIDIA NIM, NVIDIA GenAI‑Perf, NVIDIA NeMo RL, NVIDIA NeMo Gym - into its technology stack to support the deployment, evaluation, training and post‑training of AI models and agents.
Open models are central to Fastweb + Vodafone’s sovereign AI strategy. Both MIIA, the company’s Italian-native language model, and ROSS, its agentic AI assistant developed by NextMindLab, spin‑off of the company dedicated to research and development of innovative artificial‑intelligence‑based solutions, are built on open models. This enables Fastweb + Vodafone to adapt and optimize AI capabilities for Italian-language and enterprise requirements, while retaining control over infrastructure and data located in Italy. NVIDIA full-stack AI infrastructure help the company move these models and agentic applications from development into production.
To bring AI models from the development phase to production use, Fastweb + Vodafone has integrated NVIDIA NIM into its technology stack, the platform of microservices optimized for inference that enables efficient and scalable deployment of open AI models within enterprise environments. This technology accelerates model deployment and supports the delivery of generative AI capabilities at high levels of performance and reliability, also supporting agentic applications like ROSS, while ensuring control over infrastructure and data. To complete the stack, NVIDIA GenAI-Perf is used to measure and optimize model performance, facilitating their transition from experimentation to production use.
For the fine-tuning and post-training phases, Fastweb + Vodafone has implemented NVIDIA NeMo RL and NVIDIA NeMo Gym, open libraries that enhance model capabilities in areas such as function calling - the ability of a model to interact with external tools and systems through structured calls - and programming, through the creation of dedicated agentic environments. The company is also experimenting with advanced distillation techniques, which enable the transfer of capabilities and knowledge from larger models to more compact ones, optimizing their efficiency and production readiness.
Thanks to the use of these technologies, Fastweb + Vodafone is able to accelerate model development and their adaptability to tasks performed by agents, strengthening its ability to build, optimize, and deploy new agentic AI systems. This approach, based on continuous innovation, further consolidates Fastweb + Vodafone's integrated AI ecosystem developed in collaboration with NVIDIA and contributes to positioning Italy as a European hub for sovereign artificial intelligence.