AI data center capex is no longer neutral: GPU capacity now defines who can train and deploy models at scale. Procurement and architecture must adjust, because reliance on a single AI stack increases cost volatility and long-term lock-in risk. Source @VisualCap via @antgrasso

→ View original post on X — @antgrasso, 2026-03-24 14:02 UTC