Goldman Sachs has argued that Wall Street is significantly underestimating the scale of AI infrastructure spending, projecting that hyperscaler capital expenditures could reach $1.1 trillion in 2027, well above consensus expectations of around $920 billion. In an even more aggressive upside scenario, the bank estimated spending could hit $1.4 trillion by 2027.
The portfolio strategy research team, led by Ryan Hammond, drew on historical precedents from previous technology build-outs including railroads and the automobile industry to support its view that AI investment will follow a steeper and longer trajectory than markets are currently pricing in. The bank projects hyperscaler capex will reach $757 billion in 2026, representing an 84% increase from 2025 levels.
What drives Goldman’s bullish outlook is its forecast that global token consumption will multiply 24 times by 2030, reaching approximately 120 quadrillion tokens per month. This explosion is primarily driven by enterprise AI agents rather than traditional human-prompted chatbot interactions. The bank expects daily AI queries to climb from 5 billion in 2025 to 23 billion by 2030, with roughly 30% coming from agentic use cases.
By 2040, if enterprise agents achieve full-scale adoption, Goldman estimates token consumption could reach 55 times current levels. A typical chatbot consumes about 1,000 tokens per session, while continuously active intelligent agents could exceed 100,000 tokens per day.
The bank has previously emphasized that physical constraints—not capital availability—are the real bottlenecks limiting AI infrastructure expansion. Goldman estimates global data center electricity demand will increase 50% by 2027, with 60% of that growth requiring entirely new power capacity. The firm expects supply and demand for AI compute to reach equilibrium no earlier than the second half of 2027, with datacenter occupancy peaking at approximately 93% before constraints begin easing.
Morgan Stanley has echoed similar views, estimating U.S. hyperscalers alone will spend more than $800 billion on capex in 2026, roughly matching what the entire non-tech group in the S&P 500 spent in the prior year.
Read Article: Google Moves to Dismiss Indie Artists’ Copyright Suit, Citing YouTube License for AI Training
