The S&P 500 has returned 9% year-to-date and trades at 21x forward earnings — precisely where it stood a year ago. Beneath that surface stability, however, a structural tension is building. Goldman Sachs Portfolio Strategy Research, led by Ben Snider, argues that the durability of current valuations hinges almost entirely on one variable: corporate profitability.1
"Record profitability has been one of the factors supporting high S&P 500 valuations," the authors write. The S&P 500's trailing four-quarter ROE has reached a record 22% — and Goldman's macro model makes the stakes explicit: "every 1 percentage point change in S&P 500 ROE is associated with about a 1x turn in S&P 500 P/E." With the index trading in the 87th percentile of historical valuations since 1980, the margin for error is thin.
The Mega-Cap Mirage
The record ROE is not broadly distributed. Goldman Sachs is direct about the divergence: "ROE for the aggregate S&P 500 has benefited from the growing weight of the mega-cap tech stocks while ROE for the median S&P 500 company has declined." The seven largest technology stocks — NVDA, AAPL, GOOGL, MSFT, AMZN, AVGO, and META — collectively register a 44% ROE, expanded by 900 basis points over three years. Meanwhile, the median S&P 500 company's ROE sits at 17%, below prior highs, pressured by higher interest expense and modestly elevated taxes.
The AI capex boom is the engine behind mega-cap ROE expansion — but Goldman Sachs identifies it as the same force now turning against it. "The AI capex boom has lifted semiconductor profitability but will be an increasing headwind to mega-cap tech ROE going forward," the authors state. Consensus estimates imply ROE for the seven largest technology stocks will decline by an average of 700 basis points next year, with Apple facing a projected 26-point contraction and NVIDIA 12 points.
The Depreciation Reckoning
The mechanism is straightforward but underappreciated. Hyperscalers have been spending at a scale that is fundamentally reshaping their business models — and their income statements. "Consensus estimates suggest the hyperscalers will spend $770 billion on capex in 2026, equivalent to 100% of cash flows from operations." To fund continued growth, these companies have turned to debt and equity issuance while pulling back on buybacks. Net debt for the hyperscalers has increased by $170 billion since the start of 2025.
The asset-intensity shift is equally consequential. "For the decade prior to AI capex build-out, the asset-light business models of the hyperscalers were reflected in rising sales/asset ratios, supporting ROE expansion. However, the physical footprint required to support AI workloads has led to a steady decline in asset turnover." Analyst estimates project depreciation and amortization climbing from 7% of hyperscaler revenues in 2022 to 12% by 2027 — a five-percentage-point structural drag that will consume a meaningful portion of the semiconductor earnings boost AI capex has generated.
Semiconductors: Beneficiaries With an Expiry Date
The most direct earnings winners of the AI buildout have been semiconductor firms, where net profit margins have climbed to a record 41%. "Strong moats and a shortfall of supply relative to demand have given these semiconductor firms strong pricing power, boosting their margins and lifting ROE." But Goldman flags investor skepticism: the firm notes that analysts "fear the combination of slowing demand and increasing supply, including from new competitors, will eventually weigh on semiconductor profitability." Forecasts imply margins compress to 49% near-term before moderating — and ROE faces pressure as equity growth outpaces earnings growth.
The Long-Term Upside Case
Goldman Sachs is not entirely bearish. The report presents a credible bull case: "AI also creates a key long-term upside case for ROE, both for the mega-cap tech stocks and for the broad equity market." Q1 earnings offered early evidence — positive revenue revisions, large backlogs, and rising gross margins among hyperscalers. The firm's analysts also "expect an inflection in token economics ahead for hyperscalers and model providers, enabled by stabilizing token prices but falling compute cost per token."
For the broader index, productivity is the mechanism. "For the average company in the index, eventual AI productivity gains should boost sales and earnings per employee, in turn lifting profitability." The caveat: adoption remains nascent. Last quarter, only 2% of S&P 500 companies quantified AI productivity's impact on earnings.
The long-term distribution of those gains also faces a policy risk Goldman Sachs doesn't minimize: "The share of AI-related productivity gains that accrue to corporate profits over the long term will depend in part on the response of government policy." Corporate profits as a share of GDP are at their highest level since 1950.
Key Takeaways for Advisors and Investors
1. Valuations are profitability-dependent. At 21x — the 87th percentile historically — the S&P 500 leaves little room for ROE deterioration. Goldman's 1:1 model relationship between ROE and P/E is the number advisors should keep front of mind.
2. The mega-cap ROE peak may be in. Consensus already prices in a 700 bp average ROE decline for the seven largest tech stocks next year. That's not a tail risk — it's the base case.
3. Depreciation is the slow-moving headwind. The capex being deployed today converts into depreciation charges over 2025–2027, compressing hyperscaler free cash flow and ROE systematically. This is structural, not cyclical.
4. The median stock is a different story. Aggregate index ROE flatters the picture. At 17%, median company ROE is below prior cycle highs — and the AI productivity tailwind for the average business remains largely unrealized.
5. Goldman's ROE Growth basket (GSTHGROE) — rebalanced in this report — targets the 50 stocks with the highest consensus ROE growth over the next 12 months. The median basket constituent carries expected ROE growth of 18% versus 0% for the median S&P 500 stock, at a similar valuation.
The AI capex cycle is not a headwind to avoid — it is a transition to navigate. The question for markets is whether hyperscaler monetization accelerates fast enough to offset the depreciation wave now cresting. Goldman Sachs frames the answer as genuinely open. Advisors should treat it the same way.
Footnote:
1 Goldman Sachs Portfolio Strategy Research, "US Weekly Kickstart: The impact of the AI capex boom on S&P 500 return on equity," Ben Snider, Ryan Hammond, Jenny Ma, Daniel Chavez, Kartik Jayachandran, and Christophe Sung, June 12, 2026.
