NVIDIA DSX Platform enables AI factories with token-as-revenue economy

NVIDIA has announced the NVIDIA DSX Platform, a comprehensive suite aimed at infrastructure builders looking to establish AI factories. The platform combines open-source modular software libraries, APIs, reference designs, NVIDIA's accelerated computing hardware, and partner technologies into a co-designed stack. This integration is intended to streamline every phase of creating an AI factory—from initial design to ongoing operation. NVIDIA CEO Jensen Huang emphasized a shift in the industry, stating that tokens have become the primary revenue units, meaning the value generated by AI systems is increasingly tied to the volume and quality of tokens (outputs) they produce. The DSX Platform directly addresses this by providing builders with pre-validated building blocks and best practices, reducing time-to-deployment and optimizing for token-generating workloads.

By offering a unified ecosystem, NVIDIA aims to lower the barriers for enterprises and cloud providers to construct their own AI factories—large-scale data centers optimized for generative AI inference and training. The platform leverages NVIDIA's existing GPU and networking technology (e.g., NVLink, InfiniBand) while adding new orchestration and management layers. Key components include NVIDIA AI Enterprise software, Base Command Manager for cluster management, and partnerships with system integrators. For professionals in the field, this means access to a turnkey solution that can handle the entire lifecycle from model training to serving, with built-in scalability and efficiency. The DSX Platform is available now to select partners, with broader rollout expected in Q2 2025.