BofA believes the fragility of the AI capex boom has not dissipated and maintains its underweight stance on European equities

Overview

The report examines whether the AI capex boom can continue and how this issue affects European large caps and sector, style and country allocations. BofA believes the market has priced in excessive certainty regarding AI profitability without offering sufficient compensation for price competition, financing costs, power constraints, slowing US growth and energy supply risks. It therefore maintains its negative view on European equities and continues to favor defensive, quality and small-cap styles.

Core views

AI capex has become the most important driver of global equity markets. Consensus expectations for US hyperscaler capex over the next 12 months have risen from less than $300 billion at the beginning of 2025 to $940 billion, equivalent to 3% of US GDP. This spending has helped global growth remain resilient despite the energy shock caused by the trade war and the US-Iran war, while making corporate earnings momentum considerably stronger than macroeconomic momentum. The report emphasizes that large capex cycles recognize revenue for equipment and service suppliers first, while purchasers capitalize their costs and recognize them later in the income statement through depreciation, temporarily boosting aggregate market margins. Capital flows have simultaneously caused industries directly benefiting from AI investment to substantially outperform over the past three years, making market performance highly momentum-driven. The report views market pricing of this boom as exceptionally optimistic. Consensus expectations for global equity margins over the next 12 months have reached an all-time high of 14.5%. The technology sector is expected to contribute 25% of global earnings over the next 12 months, above its 15% average over the past decade and the 13% peak during the technology boom. Consensus forecasts also assume that high earnings growth can persist, with compound EPS growth of 18% over the next three years and 22% over the next five years, levels normally seen only during post-recession recoveries. Despite substantial uncertainty about AI profitability, the global equity risk premium has fallen to 3%, its lowest since 2002; the European risk premium has also declined to 4.5%, its lowest since the beginning of this century. The report argues that the coexistence of high earnings expectations and extremely low risk compensation leaves the market vulnerable to any negative surprise. Whether AI investment can ultimately be sustained depends on hyperscalers earning returns above their cost of capital, which in turn depends on competitive intensity and pricing power. The report compares three historical paths: the cloud computing market in the 2010s was dominated by a few companies whose pricing power and high margins were sufficient to absorb depreciation, preventing an investment collapse; fiber-optic investment in the 1990s created excess capacity, deteriorating pricing and a substantial equity valuation adjustment; and although the airline industry created enormous value for consumers, intense competition left companies with meager profits, converting most of the economic value into consumer surplus. The report believes AI could still follow a favorable path, but the history of technology capex booms frequently turning into busts means the market should not treat success as a certainty. Recent market movements have already demonstrated this fragility. The MSCI Europe AI Capex Select Index underperformed the broader market by 22% between June 22 and July 29. Second-quarter results from hyperscalers and improving AI revenue subsequently drove the index to outperform by 14% over two weeks, but AI capex beneficiaries have again fallen 6% relative to the market in recent days. The report identifies three structural pressures. First, Chinese high-performance open-weight models are approaching frontier capabilities at lower per-task costs, prompting customers to switch to lower-priced models. The Silicon Data Token Expenditure Index has fallen 50% since May, reflecting both price reductions by frontier labs such as Anthropic and OpenAI and increased use of low-cost Chinese models. Because frontier model providers contribute a substantial portion of hyperscaler AI revenue, weakening pricing power at the model layer could reduce returns across the entire AI ecosystem. Second, hyperscalers' previously abundant free cash flow is turning negative because of enormous investment commitments, making AI capex increasingly dependent on bond financing. Their credit spreads have more than doubled from last year's low of 55 basis points to 120 basis points, while risk-free rates have risen to a two-decade high. Based on past capital cycles, the report argues that capex expansion generally ends when expected investment returns fall below the cost of capital; increasing reliance on financing and higher debt costs are making this tipping point easier to reach. Third, power generation and transmission bottlenecks could constrain data-center construction. Even if AI projects themselves are profitable, they may not be able to obtain the electricity required to meet consensus forecasts. Public opposition to data-center construction is also increasing in the United States, with the report citing Texas's suspension of approvals for AI-related construction projects as an example of political resistance. In addition to AI risks, BofA is concerned about the US labor market, consumption and energy supply. Three-month US nonfarm payroll growth is again close to zero, historically a warning sign of macroeconomic weakness. Weak retail sales, declining consumer confidence and slowing inflation momentum have also prompted the firm's US rates strategists to turn bullish on US duration following the recent bond sell-off. Meanwhile, the US-Iran conflict remains unresolved, and energy shipments through the Strait of Hormuz are close to zero. The US Strategic Petroleum Reserve has declined from 415 million barrels to 296 million barrels, while China's crude oil imports have fallen from 12 million barrels per day a year earlier to approximately 7 million barrels per day, indicating that calm in the oil market partly depends on temporary measures. Refined-product crack spreads remain elevated, and US gasoline prices have rebounded to a two-month high of $4.8 per gallon, potentially placing further pressure on consumers. Prediction markets assign only a 19% probability that transportation through the Strait of Hormuz will normalize before early December. These risks, combined with optimistic earnings expectations, support the report's negative assessment of European equities. BofA expects the Stoxx 600 to fall to 610 points by end-2026 and to 580 points in early second quarter 2027. The main text describes the latter as approximately 10% downside, while the related chart indicates more than 10% downside. Its model shows that every 3-point decline in euro-area PMI reduces the Stoxx 600's macro-implied fair value by approximately 5%, while every 100-basis-point increase in real yields also reduces fair value by approximately 5%. Although the euro-area macro surprise index has improved and the global composite PMI new orders index stood at 52.4 in July, the report expects tight credit conditions, the fading World Cup effect and rising natural gas prices to push euro-area PMI down to 49 by year-end. BofA also expects US high-yield credit spreads and the European equity risk premium to rise while earnings expectations weaken. Although falling real yields can support valuations, they would weigh on the value-oriented structure of European indices. Europe's valuation discount to the United States remains substantial but has narrowed to its smallest since 2022, while Europe still faces a structural disadvantage in earnings growth. The report therefore maintains an underweight position in Europe relative to global equities. In style allocation, cyclicals have underperformed defensives by 6% over the past two months, giving back one-third of their gains since March, but their relative level had previously risen to a 30-year high. If AI momentum continues to reverse, risk premia rise and bond yields fall because of weakening employment and inflation, the report expects cyclicals to have another 6% downside relative to defensives by early 2027. BofA also underweights value relative to growth and overweights quality stocks and small caps relative to large caps; small caps are viewed as pricing in an excessively pessimistic outlook. At the sector level, the report overweights food and beverages, pharmaceuticals, telecommunications, luxury goods, software, real estate and chemicals. Food and beverages and telecommunications have relatively strong negative correlations with AI proxy indicators and can provide defense when risk premia rise; pharmaceuticals may benefit from declining bond yields. Software combines defensive and growth characteristics. It has already outperformed by 30% over the past month, but the report still expects approximately 5% relative upside. Luxury goods have underperformed by 50% over the past three years, and the report believes they price in an excessively pessimistic global growth scenario and could benefit from Chinese fiscal stimulus. The report has neutral allocations to airlines, automobiles, utilities, insurance, energy, mining and construction materials. It downgrades financial services from neutral to underweight and remains underweight banks, semiconductors and capital goods. Semiconductors have underperformed by 15% since late June, but the report remains concerned that AI capex will fall short of expectations. Banks are at a 15-year relative high versus the market and are vulnerable to rising risk premia, declining yields and slowing euro-area growth. In country allocation, the report overweights Switzerland to capture defensive performance when risk premia widen. It overweights Germany because growth is pessimistically priced while fiscal stimulus has arrived, and it recently upgraded the United Kingdom from neutral to overweight. France remains neutral because of policy and macroeconomic uncertainty, while Spain and Italy are underweighted because their banking sectors may underperform. German equities and small caps are used as low-valuation cyclical hedges within the overall defensive portfolio, preserving participation in periodic improvements in euro-area macroeconomic data.

Analysis framework

The report first explains the current equity rally through the scale of AI capex, the timing of profit recognition and market momentum, and then uses margins, earnings growth and risk premia to assess the degree of optimism already priced into the market. It subsequently compares scenarios using the historical capital cycles of cloud computing, fiber optics and airlines, and tests the sustainability of AI investment returns through model price competition, financing costs, power supply and political resistance. Finally, the research team incorporates US employment and consumption, energy supply, PMI, real yields, exchange rates, oil prices, credit spreads and risk premia into long-term historical relationships and regression frameworks to generate forecasts for the Stoxx 600 and style, sector and country allocations.

Methodology notes

• Quantitative/Factor/Portfolio TheoryMulti-factor model

  Macroeconomic driver regression model

  The research team uses variables such as euro-area or global PMI, real bond yields, policy uncertainty, the euro, the US dollar and oil prices to explain the historical performance of equity indices and sectors, weighting each driver according to its historical regression relationship with the PMI benchmark. Forecasts combine published data with analysts' judgments about underlying variables and adjust for outliers. The report explicitly notes that historical correlations and forecasts are not guaranteed to remain effective.

• Cycle and Business Conditions Framework

  Historical capex-cycle analogies

  The report compares cloud computing, fiber-optic investment in the 1990s and the airline industry to assess whether AI could create a high-return oligopolistic market or instead lead to low profitability or even an investment reversal because of excess supply and intense competition. The core criterion is whether investment returns can remain above the cost of capital.

• Industry/Sector Analysis FrameworkUpstream, Midstream and Downstream Industry-Chain Transmission

  AI capex industry-chain transmission

  The report tracks how hyperscaler capex first translates into revenue for semiconductor, equipment and service suppliers, and then affects earnings and equity performance across the AI ecosystem through hyperscalers' future depreciation, free cash flow and financing needs.

• Fixed Income and Credit AnalysisSpread analysis

  Linkage between credit spreads and equity risk premia

  The report uses hyperscaler credit spreads to measure AI financing costs and US high-yield spreads as an indicator of global risk appetite. Wider spreads indicate higher capital costs and greater default concerns and generally correspond to a higher European equity risk premium and declining equity prices.

• Valuation MethodPE/PEG valuation

  Forward P/E, risk-premium and real-yield valuation

  The report analyzes the Stoxx 600's valuation using 12-month forward earnings expectations, the equity risk premium and real bond yields: a higher risk premium reduces the P/E multiple, while lower real yields can provide some valuation support.

• Event-Driven Strategy and Behavioral FinanceEvent-driven analysis

  Analysis of the US-Iran conflict and energy supply shock

  The report assesses whether the energy shock has genuinely eased by examining transportation through the Strait of Hormuz, strategic petroleum reserves, Chinese imports, crack spreads and gasoline prices, and uses these indicators to evaluate pressure on consumers, inflation and European cyclical sectors.

Asset mapping & comparison

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

• Stoxx 600

  The report expects rising risk premia and weakening earnings to drive the index lower.

  Strengths

  Euro-area macroeconomic surprise indicators have improved, while declining real yields could provide some valuation support.

  Weaknesses

  Market earnings expectations are optimistic, risk premia are low, and European earnings growth faces a structural disadvantage relative to global markets.

  Comparison

  Expected to reach 610 points at end-2026 and 580 points in early second quarter 2027.

  Risks

  Disappointment in AI commercialization, weakening US growth, energy supply shocks and widening credit spreads.

• European cyclicals relative to defensives

  Underweight cyclicals and favor defensives, with cyclicals expected to decline another 6% in relative terms by early 2027.

  Strengths

  Low-valuation cyclical assets such as German equities and small caps can participate in periodic improvements in euro-area data.

  Weaknesses

  Cyclicals had previously risen to a 30-year relative high and are vulnerable to a reversal in AI momentum, rising risk premia and declining yields.

  Comparison

  Cyclicals have already underperformed defensives by 6% over the past two months.

  Risks

  If AI investment remains strong and macroeconomic growth continues to exceed expectations, the relative advantage of defensive positioning may weaken.

• Food and beverages, pharmaceuticals and telecommunications

  The report maintains overweight positions and views them as its primary defensive allocations.

  Strengths

  Food and beverages and telecommunications are negatively correlated with AI proxy indicators and generally benefit from rising risk premia; pharmaceuticals may benefit from declining bond yields.

  Weaknesses

  Pharmaceuticals are sensitive to a weaker US dollar because of their high exposure to US revenue.

  Comparison

  Food and beverages and pharmaceuticals are listed as the most favored defensive overweight sectors.

  Risks

  Yields or risk premia may move contrary to the report's expectations.

• European software

  The report maintains an overweight position and expects approximately another 5% upside relative to the market.

  Strengths

  It has defensive growth characteristics and has demonstrated better-than-expected resilience to AI competition.

  Weaknesses

  It has already risen 30% relative to the market over the past month.

  Comparison

  Compared with semiconductors and capital goods, software is viewed as a more defensive technology allocation.

  Risks

  Intensifying AI competition could still affect sector earnings.

• European banks and financial services

  Banks remain underweight, while financial services are downgraded from neutral to underweight.

  Weaknesses

  Banks are at a 15-year high relative to the market and face pressure from rising risk premia, declining bond yields and slowing euro-area growth.

  Comparison

  Banks and capital goods are listed as the most favored cyclical underweight sectors.

  Risks

  If bond yields rise or euro-area growth exceeds expectations, the underweight view could come under pressure.

• European semiconductors and capital goods

  The report maintains underweight positions, believing both are vulnerable to AI capex falling short of expectations.

  Strengths

  They continue to benefit directly from large-scale spending on AI equipment.

  Weaknesses

  Earnings depend on continued AI investment, and semiconductors have underperformed by 15% since late June.

  Comparison

  Compared with defensive software, these two sectors are more sensitive to the AI capex cycle.

  Risks

  Model-layer price competition, hyperscaler financing costs, power bottlenecks and political resistance could reduce capex.

• European small caps and quality stocks

  The report overweights quality stocks and small caps relative to large caps.

  Strengths

  Quality stocks benefit from the interest-rate and risk-premium environment expected by the report, while small-cap pricing is considered excessively pessimistic.

  Weaknesses

  Small caps remain cyclically sensitive.

  Comparison

  Small caps and German equities are used as low-valuation cyclical hedges within the overall defensive portfolio.

  Risks

  If the euro-area macroeconomic improvement fades, the catch-up potential of small caps may be constrained.

• Swiss, German, UK, French, Spanish and Italian equities

  Overweight Switzerland, Germany and the United Kingdom; neutral France; underweight Spain and Italy.

  Strengths

  Switzerland has defensive characteristics; German growth is pessimistically priced and fiscal stimulus has arrived; the United Kingdom was recently upgraded from neutral to overweight.

  Weaknesses

  France faces policy and macroeconomic uncertainty, while Spain and Italy are weighed down by the potential underperformance of their banking sectors.

  Comparison

  Country allocation balances defensive markets with low-valuation cyclical opportunities.

  Risks

  Changes in risk premia, bank performance, fiscal stimulus and country-specific macroeconomic data.

Key data

• US hyperscaler capex over the next 12 months$940bnEquivalent to 3% of US GDP, up from less than $300bn at the beginning of 2025
• Consensus expectations for global equity margins over the next 12 months14.5%All-time high
• Technology sector's contribution to global earnings over the next 12 months25%Above the 15% average over the past decade and the 13% peak during the technology boom
• Consensus compound EPS growth18% over the next three years; 22% over the next five yearsThe report believes such levels normally occur only during post-recession recoveries
• Global equity risk premium3%Lowest since 2002
• European equity risk premium4.5%Lowest since the beginning of this century
• European AI capex index volatility-22%, +14%, -6%Underperformed the broader market by 22% from June 22 to July 29, subsequently outperformed by 14% over two weeks, and then fell another 6% in relative terms over recent days
• Silicon Data Token Expenditure Index-50%Decline since May, reflecting model price reductions and customers switching to lower-cost models
• Hyperscaler credit spreads120bpsMore than doubled from last year's low of 55bps
• Three-month US nonfarm payroll growthClose to zeroThe report views this as a historical warning sign of macroeconomic weakness
• US Strategic Petroleum ReserveDown from 415mb to 296mbContinued depletion indicates that calm in energy supply may depend on temporary measures
• China's crude oil importsDown from 12mb/d to approximately 7mb/dReached 7.2mb/d in June, the lowest since late 2016, before beginning to recover in July
• US gasoline price$4.8/gallonTwo-month high
• Probability of transportation through the Strait of Hormuz normalizing before early December19%Prediction-market data
• Stoxx 600 targets610 points; 580 pointsCorresponding to end-2026 and early second quarter 2027, respectively
• Performance of cyclicals relative to defensives-6% over the past two months; another 6% decline expectedThe forecast for further downside corresponds to early 2027
• Stoxx 600 sensitivity to PMIEvery 3-point decline in PMI reduces fair value by approximately 5%BofA macro-implied value model
• Stoxx 600 sensitivity to real yieldsEvery 100bps increase in real yields reduces fair value by approximately 5%BofA macro-implied value model
• Year-end euro-area PMI forecast49Tight credit conditions, the fading World Cup effect and rising natural gas prices could cause PMI to retreat from its July recovery high
• Software sector relative performance+30% over the past month, with another approximately 5% increase expectedThe report emphasizes its defensive growth characteristics and better-than-expected resilience to AI competition
• Luxury goods relative performance-50% over the past three yearsThe report believes current pricing reflects an excessively pessimistic global growth scenario
• Semiconductor relative performance-15% since late JuneThe report remains concerned that AI capex will fall short of expectations

Impact & implications

The report believes that if AI investment falls short of expectations in any one of pricing power, returns on capital, financing or infrastructure, the current combination of extremely low risk premia and extremely high earnings forecasts could undergo a simultaneous correction. For European markets, this means pressure on the Stoxx 600, cyclicals, value stocks, banks, capital goods and semiconductors, while food and beverages, pharmaceuticals, telecommunications, quality stocks and other defensive assets should outperform on a relative basis. Improving European macroeconomic data and declining real yields could provide localized buffers, but are insufficient to alter the report's cautious assessment of Europe's earnings structure and performance relative to global equities.

Risks

• Price competition at the AI model layer could weaken pricing power, causing the market's high expectations for AI earnings and margins to disappoint.
• Negative hyperscaler free cash flow, widening credit spreads and rising risk-free rates could push investment returns below the cost of capital.
• Power generation, transmission and data-center electricity bottlenecks could prevent AI capex from expanding in line with consensus expectations.
• Social and political opposition to data-center construction in the United States could delay or suspend approvals for related projects.
• Near-stagnant US employment growth, weak retail sales and declining consumer confidence could signal weakening global growth.
• The US-Iran conflict, disruptions to transportation through the Strait of Hormuz, declining strategic petroleum reserves and high refined-product prices could further damage consumers and market risk appetite.
• European equity earnings expectations are elevated while risk premia are low, meaning any macroeconomic or earnings disappointment could trigger a valuation downgrade.

What to watch

• Track per-task AI model prices, the Token Expenditure Index and the usage share of low-cost Chinese open-weight models.
• Monitor hyperscaler AI revenue, free cash flow, debt issuance, credit spreads and capex plans.
• Watch data-center power supply, transmission bottlenecks and changes in US local-government approvals for AI construction projects.
• Track US nonfarm payrolls, retail sales, consumer confidence, inflation momentum and expectations for Federal Reserve easing.
• Monitor transportation through the Strait of Hormuz, the US Strategic Petroleum Reserve, Chinese crude oil imports, refined-product crack spreads and gasoline prices.
• Track euro-area and global PMI, US high-yield credit spreads, the European equity risk premium and real bond yields.
• Watch whether the Stoxx 600 moves toward the forecast path of 610 points at end-2026 and 580 points in the second quarter of 2027, and whether cyclicals continue to weaken relative to defensives.