Goldman Sachs: AI Capex Boom Puts S&P 500 ROE to the Test

Goldman Sachs' recent analysis indicates that the massive capital expenditures (capex) directed towards Artificial Intelligence (AI) technologies are under scrutiny for their impact on the S&P 500 index's return on equity (ROE). The S&P 500, which has gained 9% year-to-date, has seen this rise primarily driven by stronger earnings expectations rather than multiple expansion, as its forward price-to-earnings (P/E) ratio compressed from 22x to 21x. This suggests that the market is now demanding sustainable profits, making ROE a critical metric for evaluation.

According to the bank's report, the S&P 500's trailing four-quarter ROE reached a record 22% in the first quarter of 2026, surpassing the previous peak set in 2021. Rising profit margins have been identified as a key driver of this increase. However, this profitability boom is not evenly distributed across the market; instead, it is increasingly concentrated in mega-cap technology companies. The seven largest tech stocks, including Nvidia (NVDA), Apple (AAPL), Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), Meta Platforms (META), and Broadcom (AVGO), collectively generate an ROE of 44%, having expanded their combined ROE by nine percentage points over the past three years. This concentration highlights the index's growing dependence on a smaller group of high-performing companies.

A significant shift in market dynamics is underway, moving from a focus on "chip scarcity" to "capex reality." While semiconductor companies initially captured substantial margin prizes, hyperscalers are now bearing the depreciation costs associated with extensive AI infrastructure buildouts. Goldman Sachs analysts anticipate that ROE across the largest technology companies could decline by an average of seven percentage points next year as spending on data centers, computing infrastructure, and AI-related assets continues to accelerate. This is because hyperscale technology companies are becoming increasingly asset-intensive as they expand capacity to support burgeoning AI workloads. The market is now seeking tangible proof that AI spending translates into revenue, productivity gains, and durable earnings.

In a broader economic context, Goldman Sachs had previously raised its S&P 500 year-end 2026 target to 8,000 from 7,600, citing optimism around AI-driven earnings. The bank had projected S&P 500 earnings per share (EPS) to reach $340 in 2026, representing a 24% growth, and $385 for 2027, an additional 13% increase. However, Jim Cavell, Goldman's head of global equity research, issued a warning that the business case for AI remains largely unproven, and the industry is moving further away from justifying its investments. Cavell noted that most of the economic value has accrued to semiconductor companies, while companies actually building and deploying AI products have yet to show comparable returns.

Looking at analyst expectations and market outlooks, AI capital expenditure estimates have consistently proven to be too low. Hyperscalers spent $106 billion in capex in the third quarter of this year, marking a 75% year-over-year growth rate. While consensus estimates project a slowdown in this growth, historical trends suggest that upward revisions are likely. The market hopes that AI investments will lead to productivity boosts in a broader range of companies beyond just the infrastructure layer. However, the concentration of the S&P 500, with the top 10 companies accounting for 41% of its market capitalization and 34% of its earnings, raises concerns about diversification and potential risks for passive investors if this narrow leadership falters. Furthermore, excluding AI infrastructure and energy firms, S&P 500 2027 EPS estimates have remained flat year-to-date, underscoring the risks beneath this concentration.