Nvidia Projects $1 Trillion in Revenue Through 2027, Declares Inference Boom
Jensen Huang’s GTC keynote forecast $1 trillion in Nvidia revenue from 2025 through 2027 and outlined expansion into networking, storage and orbital data centers. He declared the inference inflection has arrived, assigning Vera Rubin chips to prefill and Groq-derived silicon to decode for real-time, cost-efficient AI inference.
1. Keynote Vision and Ecosystem Expansion
Jensen Huang opened GTC by emphasizing Nvidia’s ambition to define the entire AI infrastructure stack, from compute and networking to storage, software and even orbital data centers. He positioned the company as more than a chip supplier, signaling a broader strategy to integrate AI solutions across industrial, data center and emerging orbital applications.
2. $1 Trillion Revenue Opportunity
Marking the 20th anniversary of CUDA, Huang revealed he sees at least $1 trillion in cumulative revenue between 2025 and 2027, driven by accelerating AI demand. This projection underscores Nvidia’s confidence in sustained industry spending and its role as the primary engine of the ongoing AI buildout.
3. Inference Inflection and Chip Architecture
Huang declared that the industry has reached an “inference inflection,” shifting the focus from model training to real-time model serving. He detailed a two-chip approach: Vera Rubin processors handling prefill workloads and Groq-derived silicon managing decode tasks, aiming to optimize cost, latency and scalability for live AI applications.