According to @AveryChing, several trillions of dollars are expected to be spent on AI hardware over the next 2-3 years. Source: @AveryChing on X, Nov 17, 2025. He estimates more than 500 exabytes of storage will be needed to feed the compute workload, approximately 62,500,000 8-TB drives. Source: @AveryChing on X, Nov 17, 2025. He states that global neo-clouds will stream data over dedicated fiber and pay internet users for data used in training, inference, and preprocessing. Source: @AveryChing on X, Nov 17, 2025. He emphasizes the infrastructure cannot pause for cloud failures, high fees, lengthy contracts, or slow access, underscoring a need for always-on, low-fee data rails. Source: @AveryChing on X, Nov 17, 2025. He adds he is working on these solutions with @jump_ on @shelbyserves. Source: @AveryChing on X, Nov 17, 2025. For crypto traders, these requirements imply rising demand for scalable, low-fee, continuous decentralized storage, bandwidth, and data-monetization primitives aligned with AI data pipelines. Source: @AveryChing on X, Nov 17, 2025.