Scaling Physical AI and Physics-Informed Surrogates with NVIDIA and Google Cloud

About the Session

AI-physics surrogate modeling is emerging as a critical, high-growth cloud workload across engineering, scientific simulation, and industrial design. However, Physical AI behaves very differently from traditional LLM pipelines: instead of linear token streams, these models compute on massive 3D meshes, point clouds, continuous fields, and complex geometries. As a result, they demand extreme memory bandwidth, low-latency data movement, and tight integration with live simulation, design, and optimization loops.

Join Todd McDevitt (NVIDIA) as we dive into the unique architectural realities of physical AI training and inference. We will explore the opportunity for NVIDIA and Google Cloud to define purpose-built deployment patterns, optimize data pipelines, and deliver the compute fabric required to scale surrogate models seamlessly in the cloud.