AI Sparks Capital Expenditure Super Cycle, Mid-Term Valuation Should Focus on EV/EBITDA

Overview

Barclays released an equity strategy report discussing the impact of the current AI-driven capital expenditure super cycle on market valuation and investment strategies. The report points out that as AI investment scales up to match the 1990s and 2000s, the US equity market is experiencing a surge in capital expenditure. In this context, traditional P/E valuation tends to become distorted, while enterprise value multiples (EV/EBITDA) show stronger signal value in assessing mid-term investment returns. Despite the surge in capital expenditure intensity, the report remains optimistic about the short-term (next 12 months) performance of the US equity market, but reminds long-term investors to improve their stock selection standards.

Core views

Capital expenditure is entering a super cycle and highly concentrated in the technology sector. The report estimates that the overall capital expenditure of the S&P 500 will increase by about 40% in 2026, reaching the fastest pace in at least 35 years, and consuming about 50% of operating cash flow. Unlike previous cycles, this growth is highly concentrated in communication services, technology, and non-essentials consumption sectors (i.e., large-scale cloud providers, chip manufacturers, and their supply chains). The proportion of TMT sector capital expenditure to the total of the S&P 500 is expected to approach 50% this year, which has raised concerns about overinvestment and declining marginal capital returns. In a high capital expenditure environment, EV/EBITDA is a better mid-term valuation anchor than P/E. The report uses historical data to show that during periods of high capital expenditure, P/E is easily distorted by leverage, tax shields, and depreciation policies; whereas EV/EBITDA directly reflects the price investors pay for core operational assets of the company. Data shows that since the mid-1990s, the three-year forward total return rate of the S&P 500 based on EV/EBITDA ranking has shown a clear monotonic decreasing relationship (the cheaper, the higher the return), and the return spread between "cheap" and "expensive" combinations is greater than P/E, proving the effectiveness of EV/EBITDA as a linear test tool for mid-term expected returns. Maintain a constructive outlook in the short term, but be cautious of valuation digestion pressure in the long term. Currently, the EV/EBITDA of the S&P 500 is relatively high compared to its long-term historical levels. This high valuation is driven by optimism about long-term growth and AI capital expenditure beneficiaries such as tech hardware, capital goods, and semiconductors. The report points out that although the market generally expects capital expenditure growth to slow down in 2027 or 2028, the current high enterprise value means that once capital expenditure normalizes, the space for market growth through valuation expansion will be compressed, and future gains will depend more on actual profit realization. However, given the still robust profit backdrop for the next 12 months (especially across the broad technology sector), the report remains positive on the US equity market in the short term.

Analysis framework

The report uses a combination of historical comparison and empirical backtesting for analysis. First, it compares the current AI investment wave with the telecommunications/Internet construction in the 1990s and the energy/real estate boom in the mid-2000s, defining the current period as a 'capital expenditure super cycle' based on two dimensions: capital expenditure growth rate and proportion of cash flow. Second, the report quantitatively tests the effectiveness of different valuation metrics under specific macroeconomic conditions. By comparing the correlation between P/E and EV/EBITDA quintile sorting (Quintile Sorting) during high capital expenditure periods and the three-year forward total return of the S&P 500, it argues that EV/EBITDA can more truthfully reflect asset pricing after eliminating the interference of financial leverage, tax shields, and accounting depreciation, thus deriving the conclusion that 'mid-term investment should focus more on EV/EBITDA'. Finally, combining the support from short-term profit expectations and the limited space for valuation expansion in the long term, it concludes an investment strategy of being optimistic in the short term and cautious in the long term about entry points.

Methodology notes

• Valuation MethodEV/EBITDA valuation

  In a high capital expenditure (Capex) cycle, EV/EBITDA has stronger signal value than P/E

  Enterprise value multiples (EV/EBITDA) exclude the interference of depreciation policies, capital structure, and tax differences. When companies are in a capital expenditure super cycle, massive depreciation severely distorts net income, leading to distortion in P/E. At this time, focusing on EV/EBITDA can more directly measure the price investors pay for core operational assets of the company, aligning more closely with the mid-term capital expenditure cycle.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Technology Hardware, Capital Goods, and Semiconductors

  As direct beneficiaries of AI capital expenditures, their valuations have pushed the current market's high EV/EBITDA levels

  Strengths

  Directly benefit from the massive capital expenditures of AI hyperscale cloud providers and chip manufacturers

  Risks

  If capital expenditure growth slows in 2027 or 2028, high valuations may face mean reversion pressure

• AI Hyperscale Cloud Providers

  Main spenders of the current capital expenditure super cycle

  Strengths

  Have strong long-term growth narratives and profitability

  Weaknesses

  High capital expenditure as a percentage of cash flow, which could lead to declining marginal capital returns if overinvested

  Risks

  History shows that after super cycles, there are often painful integration periods; valuation expansion space is limited after capital expenditure normalization

Key data

• S&P 500 Capital Expenditure YoY Growth Rate~40%Expected growth rate in 2026, the fastest in at least 35 years
• Capital Expenditure as a Percentage of Operating Cash Flow~50%Expected level in 2026, comparable to historical super cycles
• TMT Capital Expenditure Share~50%Proportion of total capital expenditure of the S&P 500, far exceeding the public cloud construction period at the end of the 2010s

Impact & implications

For the US equity market, the current AI capital expenditure craze means that the future market drivers will switch. In the short term, technology hardware, semiconductor, and capital goods — the so-called 'water providers' — will directly benefit from the surge in capital expenditures, supporting overall earnings expectations. However, in the medium to long term, when capital expenditures peak and decline, due to the current high valuations, the market will find it difficult to gain further gains through valuation expansion. Investors must pay more attention to whether companies can convert huge capital expenditures into actual free cash flows and profit growth. For long-term investors, the report suggests that they should raise their stock selection standards when valuations are high, avoiding paying for marginal return declines from overinvestment.

Risks

• Overinvestment leading to declining marginal capital returns (ROIC).
• After the end of the capital expenditure super cycle, historically, there are often painful industry integration periods.
• High initial valuations limit the space for gains through valuation expansion after capital expenditure normalization.

What to watch

• The slowdown in capital expenditure growth of AI hyperscale cloud providers in 2027 or 2028.
• The realization of S&P 500 earnings (NTM EPS) expectations over the next 12 months, especially in the broad technology sector.
• The efficiency of tech giants in converting massive capital expenditures into actual profits and free cash flows.