LiveRamp integrates NVIDIA GPU infrastructure into clean rooms to deliver up to 15x faster model training and inference for AI partners and brands

LiveRamp announced native support for NVIDIA AI infrastructure on April 27, 2026, upgrading its clean room architecture to enable GPU-optimized model training and inference. AI partners and brands can now train and deploy models at up to 15x speed compared with CPU-based environments without exposing data or model weights.

The GPU-optimized infrastructure eliminates the need for AI partners to rearchitect models for CPU environments; they can now run existing code directly within LiveRamp's clean rooms. Model providers can execute proprietary models securely to deliver campaign optimization, audience selection, and performance prediction services to brands while keeping raw data within the clean room.

Marketers gain faster performance optimization timelines with preliminary tests showing 15x speed improvement over CPU-based environments, and easier access to advanced AI models for data collaboration. AI partners can distribute scaled models and protect intellectual property by training and executing on brand first-party data without exposing proprietary code and model weights to partners.