Jensen Huang Unveils ‘Token Economics’ at GTC 2026 as NVIDIA Starts Full-Scale Vera Rubin Production

The cost and value of artificial intelligence services are increasingly measured by how many tokens can be generated per unit of electricity and per data center, rather than by the number of GPUs purchased. NVIDIA CEO Jensen Huang called this “Token Economics” at GTC Taipei 2026, arguing that companies should gauge returns on computing investments by token output and embrace “computing is revenue” as the new business logic for AI infrastructure.

During GTC Taipei 2026, NVIDIA announced that its next-generation Vera Rubin architecture had entered full-scale production. The company aims to expand supply amid a global shortage of AI computing capacity and improve token-generation efficiency. Huang said the production ramp-up depends on Taiwan’s ecosystem, shifting market attention to the server, chip, cooling and power-supply chains. Publicly available information has not disclosed the value of orders or exact shipment dates.