MAY 25·benzinga.com▼
AI’s $8 Trillion Buildout Is Not A Bubble – It’s A Bottleneck - Micron Technology (NASDAQ:MU), NVIDIA (NA
22V Research’s Jordi Visser says the AI buildout is constrained by physical infrastructure rather than software, citing only 12%–18% of an estimated $8 trillion capex cycle completed. He flags bottlenecks in memory, cooling, copper/fiber, substations and power. He warns of revenue-recognition risk from large backlogs and says investors should focus on platform/connectivity and power-related names over memory.
MAY 25·finance.yahoo.com→
IREN’s CEO Just Said 11 Words that Delivered the AI Boom a Brutal Reality Check
Iren CEO Daniel Roberts told Bloomberg Tech that building a gigawatt AI data center could take until 2030 to get compute online. He said utilities may require 18–24 months just to confirm power capacity, making grid interconnection a key bottleneck. The article notes Nvidia’s fiscal 2026 revenue rose to $215.9B amid GPU demand.