RESEARCH — SEPTEMBER 2026
The Price of Intelligence
Hyper 5 spent $1.1 trillion on capex in the past five years, with $5.3 trillion more expected by 2030
By Liam Hynes and Drew Bowers
The artificial intelligence (AI) build-out has reached historically extreme investment intensity and unprecedented concentration, but it has not yet developed the financing characteristics that made the dot-com build-out fragile. Measured by capital expenditure (capex) relative to depreciation (dep), investment intensity has exceeded prior cycle peaks, including those of the dot-com era and Great Recession (GR). This cycle is being driven disproportionately by hyperscalers, which have collectively spent $1.1 trillion in capex over the past five years. Visible Alpha sell-side consensus adds $5.3 trillion more by 2030. The key risk is the transition from cash- to debt-funded investment before returns validate the spending.
Key findings:
- Lessons learned. Capex cycles typically move through four phases: build, stretch, overshoot and digest. Investment intensity rises when expected earnings translate to higher returns on new capital (build). This has historically persisted despite tighter monetary policy (stretch). The cycle turns when the next investment return no longer exceeds the cost of financing (overshoot). Markets then undergo a period of lower investment (digest).
- Cash wedge. S&P 500 cash-to-capex reached an unprecedented 330% (3.3x) in 2021, three times the level at the dot-com peak, and has been unwinding since, partly to fund the build-out.
- Concentration. Visible Alpha sell-side consensus is for the Hyper 5 (Amazon, Alphabet, Microsoft, Meta, Oracle) to reach 52% of total S&P 500 capex by 2028, versus 13% and 30% in 2020 and 2025, respectively.
- Valuations. From an earnings multiple perspective, the rest of the S&P looks historically more expensive than the hyperscaler companies spending the capex as the market awaits increased earnings from AI efficiency gains.
Explore the data used to conduct this research:
Build your Fundamental Powerhouse
With financial history dating back to 1950, Compustat® has powered investment models, academic studies, and quantitative research for over 75 years. With more than 3,000 standardized fields and point-in-time snapshots beginning in 1987, Compustat helps teams analyze company performance across market cycles using consistent, backtest-ready data trusted by financial professionals worldwide.
Visible Alpha Estimates and S&P Capital IQ Estimates
The S&P Capital IQ and Visible Estimates datasets combine the forecasts, assumptions, and logic from full working sell-side models with an extensive global contributor network and deep history to provide the most comprehensive view of market expectations. Through the breadth of S&P Capital IQ Estimates and the granularity of Visible Alpha Estimates, the data provides both scale and precision, enabling better forecasting, benchmarking, and decision-making.
SNL Energy and SNL Sector Financials
The SNL Energy dataset provides access to financial and operational data for more than 450 power and midstream companies. Featuring detailed revenue, volume, and customer metrics by customer class, along with operational breakdowns across production, transmission, and distribution, this dataset enables analysis of industry performance and trends. It also provides historical insights into changes in U.S. power generation dating back to 1994.
The SNL Sector Financials delivers detailed, GAAP- and IFRS-based data for over 25,000 companies globally. With more than 6,000 sector-specific fields across Financial Institutions, Real Estate, Energy, Metals & Mining, and TMT, this dataset enables precision analysis tailored to each industry’s unique reporting structure.