NVIDIA CEO Jensen Huang, at the GTC Taipei 2026 conference on June 1, laid out a new logic for the AI industry centered on “token economics”: tokens have become a quantifiable profit unit, AI factories are essentially token factories, and AI is being viewed as a “generator” of GDP. Citing data, he stated that the number of AI programming calls made by approximately 30 to 40 million professional software developers worldwide has risen from 300 million in 2023 to 400 million in 2024, 500 million in 2025, and has surged to 1.4 billion in the first few months of 2026 alone — nearly triple the total for 2025. In response to concerns about AI taking jobs, Huang pushed back, asserting that the actual number of employed software engineers is still increasing, and calling such claims “complete nonsense.”
On the hardware front, Huang announced that the flagship AI chip platform, Vera Rubin, has officially entered full production, describing it as NVIDIA’s “most ambitious product.” It involved 40,000 engineers in R&D, with a supply chain twice the size of its predecessor, Grace Blackwell, and featuring over 150 Taiwanese partners. The flagship Vera Rubin NVL72 cabinet integrates 72 Rubin GPUs and 36 Vera CPUs, utilizing a fully liquid-cooled design; NVIDIA estimates that, when combined with decoupled inference using Groq LPUs, the annual revenue potential of a 1 GW AI factory could leap from $30 billion to $300 billion. Huang also unveiled the next-generation Feynman architecture roadmap for the first time and announced the official opening of NVIDIA’s Taipei Constellation headquarters, stating that AI computing demand in Taiwan has “rocketed skyward.”