2 articles
Jensen Huang's keynote revealed $1T in Blackwell and Vera Rubin orders, 10x efficiency gains, Kyber rack architecture, and a 2028 Feynman GPU preview.
AI inference costs are dropping at roughly 10x per year — faster than Moore's Law ever achieved — and this cost trajectory should be reshaping how you design systems, price products, and think about build-versus-buy decisions.