The FT article states that capital expenditure on AI data centers by hyperscalers is forecast to reach cumulative totals approaching $9 trillion within the decade, driven by GPU cluster buildouts from firms including Microsoft, Google and Meta. It cites primary data on current project pipelines showing more than 1,000 new facilities planned or under construction in the US alone. (https://www.ft.com/content/805f78f3-8da3-4fc0-b860-207a859ac723)

Power consumption projections in the piece indicate AI data centers could consume 8 percent of US electricity by 2030, with some regional grids already reporting interconnection queues exceeding 50 GW. The report references utility filings and IEA data on the scale of required new generation capacity. (https://www.ft.com/content/805f78f3-8da3-4fc0-b860-207a859ac723)

The article notes that current AI revenue streams remain concentrated in a narrow set of enterprise use cases, citing financial disclosures from cloud providers that show AI-related revenue still below 10 percent of total for most operators. It references comparable infrastructure oversupply cycles without drawing conclusions. (https://www.ft.com/content/805f78f3-8da3-4fc0-b860-207a859ac723)