Conference Presentation, Product Demonstration
Powering Europe's Industrial Renaissance | Paolo Guglielmini, NVIDIA | RAISE Summit 2026
- A surge in agentic AI adoption is anticipated within the final weeks of the year, particularly driven by urgent initiatives in Europe to close the compute gap, with an inflection point described as occurring immediately alongside the current discourse.
- Industry-wide adoption patterns are expected to mirror software development trends observed on GitHub, where output growth significantly outpaces population growth, signaling that AI is becoming a force multiplier across sectors.
- Return on investment is projected to materialize specifically at the agentic layer of a five-layer industry structure, contingent on open models serving as the primary strategy for Europe to build ecosystems and deploy specialized, domain-specific, and culturally relevant AI.
- Europe currently accounts for only 5% of global AI compute capabilities, a deficiency viewed as insufficient, prompting a call for policymakers and capital allocators to accelerate compute deployment to avoid becoming mere spectators in the revolution.
- From GTC Paris to the present, 800 exaflops of incremental capacity have been deployed, with expectations for a 10X increase in total compute capacity by 2027 to meet the evolving demands of agentic AI workloads.
- NVIDIA remains committed to the Nemotron roadmap, featuring open models scaled from 5 billion to 550 billion parameters, with innovators in France, Germany, Israel, Poland, and via the Euro LLM already deploying technologies on this stack.
- The Vera Rubin architecture is scheduled for release between the end of this year and the next, introducing seven new chips and five systems designed to deliver a 35X improvement in tokens creation per megawatt and a 2X productivity gain per chip.
- Approximately 90% of AI factories currently operate on NVIDIA accelerated compute platforms, and future hardware iterations like Vera Rubin are explicitly engineered to support the high-performance requirements of agentic AI.
- NVIDIA intends to continue expanding its partner ecosystem and open strategy regarding weights, data, and libraries, with specific plans to attend GTC Berlin in October to foster further development and inspiration.
- Specific model portfolios are positioned for distinct physical and biological applications, with Cosmos leading open physical AI, Isaac Groot targeting robotics and humanoids, and Bionemo specializing in digital biology and health.