The trajectory is real but the economics are not yet closed. In an interview with Construction Dive, Bain & Company adviser Peter Hanbury called a cumulative $5 trillion to $6.5 trillion of data center investment through 2030 “certainly conceivable,” pointing to roughly $780 billion of 2026 capital expenditure from the five largest hyperscalers, nearly five times what they spent three years earlier. His caveat is the ratio that has to hold behind it: Bain estimates annual AI infrastructure spending could reach about $1.5 trillion by 2031, and if capital expenditure runs at roughly 25 percent of industry revenue, sustaining that spend would require an AI market approaching $6 trillion in annual revenue.
That is the question the headline names. For the investment to pay, two things have to happen at once: AI must move beyond efficiency and productivity use cases into new sources of economic value such as autonomous systems, physical AI and new AI-enabled products, and the physical bottlenecks have to ease. Power is the tightest one, Hanbury said, calling for system-level solutions: more generation, faster interconnection, behind-the-meter capacity, storage, and closer coordination between utilities and technology companies.
The figures vary widely by scope. A Brookings Institution study by economist Stijn van Nieuwerburgh puts total US data center and AI infrastructure investment at $10.3 trillion from 2025 to 2032, about 3.6 percent of GDP a year; Goldman Sachs estimates AI investment alone at 1.9 percent of GDP in 2026; S&P Global Ratings projects combined 2027 capex by the big six tech firms above $1.3 trillion. US Census data shows private data center construction at a $75.2 billion annualized rate in July 2026, up 57 percent year over year, already exceeding general office construction. The near-term issue, as Hanbury framed it, is not whether enough capital exists, but whether the revenue arrives to keep it deployed.