Charting an earnings surge, powered by AI capex

Once again, U.S. companies came through. In the second quarter S&P 500 companies delivered their fastest earnings growth in five years.

As the earnings season began, expectations were already elevated, with bottom-up consensus forecasts calling for 10% sales growth and 22% earnings growth. S&P 500 companies easily surpassed those expectations across sales, margins, and earnings growth metrics, as they have done over the past few quarters. After excluding one-off items, the S&P 500 reported 14% sales growth and 29% earnings growth—the fastest pace since the economy reopened post-pandemic in 2021. Some 85% of companies beat consensus earnings expectations, which is also the highest share since 2021.

And once again a key market theme, capital investment in AI, explains much of the profit growth. Semiconductor companies (including market star Nvidia) make up 20% of S&P 500 market capitalization, but 40% of total S&P 500 earnings growth this quarter. The sectors that include the biggest beneficiaries of AI spending (information technology, communication services and consumer discretionary) were among those with the fastest year-over-year earnings growth. Energy and financial services companies also posted strong earnings growth due to higher oil prices and a pickup in capital markets activity, respectively.

Second quarter results from the U.S. hyperscalers—Microsoft, Alphabet’s Google, and Amazon—were a particular bright spot following a rough Q1. For the first half of 2026, U.S. hyperscalers underperformed significantly: the Bloomberg Magnificent 7 (“Mag 7”) index fell 2%, compared with a 10% return for the S&P 500 and a 113% return for the semiconductor index. That gap reflected investor concerns that hyperscalers may not be able to generate sufficient return on investment (ROI) on their significant AI capex, which is now increasingly funded by debt rather than free cash flow. But in the second quarter hyperscalers reported accelerating cloud revenues and rising profit margins, which reassured investors to some extent (Exhibit 1, Accelerating AI revenues. Hyperscalers also provided more clarity around the timing of ROI on these investments. So far in 2H, the Mag 7 has outperformed the S&P 500 and the semiconductor index by 3 percentage points and 24 percentage points, respectively.

U.S. earnings outlook: More upside than downside

Since the start of the year, consensus earnings expectations for S&P 500 earnings growth—which are typically too optimistic—have been revised higher (Exhibit 2 US earnings revisions). That is a notable shift. Over the past 30 years, bottom-up consensus annual EPS estimates for the S&P 500 have been revised lower by 4% per year, usually because analysts are overly bullish in their margin expectations.

Positive revisions have typically occurred only when the economy is emerging from a recession. But in 2026 analysts have consistently underestimated AI capex spending. Since Jan. 1, bottom-up consensus expectations for AI capex have risen from USD 650 billion to USD 800 billion for 2026 and from USD 800 billion to USD 1.1 trillion for 2027. As a result, both 2026 and 2027 EPS estimates were revised higher, by 17% and 15%, respectively.

We think it’s too early to call a peak in AI capex. Accelerating AI demand and cloud revenue, along with ongoing supply chain constraints, suggest that AI capex will likely continue to rise. We expect the trend of positive earnings revisions to persist in the coming months.

The potential productivity benefits from AI adoption represent an upside risk to earnings expectations. In the second quarter, 65% of S&P 500 companies mentioned AI in earnings calls, but only 2% quantified the impact of AI-driven productivity on their profits. Other measures we track, such as changes in operating leverage and revenue per employee, and shifts in earnings expectations in the sectors exposed to the AI adoption theme (including health care), also show limited signs of an AI-driven productivity boost so far. However, while the timing is unclear, the benefits of AI adoption help tilt the distribution of earnings outcomes to the upside over the next 12-18 months.

A risk-on bias—with diversification

Seeing a positive distribution of earnings outcomes, we remain overweight equities and the AI theme more broadly. Since the start of 2026, including during the peak of the U.S.-Iran conflict, we have had a positive earnings outlook and maintained an equity overweight. Second quarter earnings results, our expectations around AI capex and the recent resilience of the U.S. economy reaffirm that view.

The U.S. remains our most favored market. Over the next 12 months we estimate the S&P 500 could deliver mid-teens earnings growth. Our portfolios are also overweight EM equities in light of the fact that around 50% of the index is directly geared to the AI investment cycle.

How do we view the much discussed AI capex cycle? We note a healthy dose of investor skepticism around the AI market narrative. This suggests to us that the distribution of outcomes for multiples is also positively skewed. While earnings expectations have continued to rise this year, multiples have contracted. The S&P 500 trades at a P/E multiple of around 19.5x forward 12-month earnings, toward the lower end of its range over the past three years (Exhibit 3 S&P forward P/E). EM equities trade at a decade-low 10x multiple, both in absolute terms and relative to global equities.

Geopolitics is only part of the story. We think multiples reflect investor uncertainty on three main fronts. First, investors wonder whether the earnings of AI capex beneficiaries are sustainable. (Thus we see downward pressure on memory sector multiples.) Second, investors question the ability of hyperscalers to generate sufficient ROI (despite the recent rebound, hyperscaler multiples still trade near five-year lows). And finally, investors debate whether AI adoption in the corporate sector will bring meaningful and durable improvements in productivity. As a result, multiples for the equal weighted S&P 500 trade near 2–3 year lows.

While our return expectations are driven mostly by our earnings growth forecasts, a resolution to the U.S.-Iran conflict and/or a decline in investor skepticism about AI’s potential could also support some multiple expansion in the coming months.

However, risks remain elevated and so we continue to manage our exposure to the AI theme through diversification. Within equities, our overweight to Japanese equities is less tied to AI capex and more to ongoing corporate governance reform and domestic economic reflation. Across other asset classes, we overweight U.S. high yield and a G10 FX carry basket as expressions of a stable nominal growth theme. Our multi-asset portfolios also include overweights in EM debt (local and dollar-denominated), given its positive carry and, importantly, its exposure to regions such as Africa, Latin America and Southeast Asia, where returns are less closely linked to the AI theme.

4f3482bf-a84f-11f1-bee3-d15f8c4e9fb6