Jensen Huang Says Hyperscalers Get This ‘Almost Always Wrong' — Nvidia Has a Workaround

Nvidia CEO Jensen Huang highlighted a structural challenge facing major hyperscalers, stating they are "almost always wrong" when forecasting artificial intelligence capacity requirements due to relying on rigid annual planning cycles. Given that AI demand and technology evolve much faster than traditional annual schedules, hyperscaler forecasts quickly become outdated, creating operational friction in scaling infrastructure. To address this limitation, Nvidia is pursuing a strategy focused on fostering an expansive distributed network of regional cloud providers, also referred to as neoclouds. Huang emphasized that the broader industry ultimately needs between 50 to 1,000 regional operators rather than relying solely on a small group of hyperscalers. These localized providers can move more rapidly to secure vital resources such as power, land, and datacenter space. This decentralized strategy aims to drive robust, long-term hardware demand across global markets including Australia and Southeast Asia. By diversifying its customer ecosystem beyond mega-cap tech giants, Nvidia is securing multiple avenues for hardware consumption while mitigating infrastructure bottlenecks.