As we near Halloween and the end of October, the U.S. third-quarter earnings season is now fully underway.
Today, we will take a look at the results.
We see that more companies across more areas of the economy are returning to earnings growth. Although we still see that AI dominates the level of earnings — and earnings growth — across the market.
Earnings beats keep getting better
Data shows that it’s typical for U.S. companies to manage analyst expectations down, and then beat those expectations when earnings are actually announced.
Recently, the proportion of companies beating has been growing. This quarter, it reached the highest level in at least 16 years, according to Barclays Research.
Chart 1: More companies are beating earnings than recent years
Gains are spreading to small-cap companies
Index-wide earnings data also shows that small-cap companies, which had been struggling as wages and interest rates climbed, are finally starting to see a recovery in profits.
So far, market-wide earnings don’t seem to have much tariff impact either. We’ve already seen GM gain +15% post-earnings, in part because it said it’s making faster-than-expected progress in reducing its (multibillion dollar) tariff bill.
Chart 2: Small-cap earnings expected to continue to recover
Tech driving earnings across all market caps
Although the breadth of earnings is improving, it’s still AI that has been driving the majority of earnings growth and U.S. stock-market outperformance. The chart below shows how the tech sector dominates earnings growth across all market caps. In the chart, the bar:
- Heights show the contribution to earnings growth.
- Width shows the combined net income of companies in each group.
Tech (blue bars) is still disproportionately driving earnings growth in the third quarter. More than half of earnings growth in all indexes, except the mid cap (S&P 400), comes from the tech sector alone. That’s despite the tech sector representing less than 10% of the net income in the S&P mid- and small-cap (S&P 600) indexes.
However, a consistent 3-4 percentage points of earnings growth can be attributed to other sectors in all S&P index caps too.
Chart 3: Earnings growth from tech vs. other sectors (by index market cap)
The Nasdaq-100® is a slightly different index. First, it only includes Nasdaq-listed companies, but also it excludes Financials (by rule). As a result, Tech in the Nasdaq-100® Index weighs in at 54% (using GICS sectors).
Tech earnings are catching up to market cap concentration
Some have worried that the large weight of mega-cap stocks in the U.S. market highlights risks of concentration and a bubble. Importantly, we have seen similar concentration before —notably in the 1930’s and 1950’s.
It’s interesting to look at the calculations from Goldman Sachs below. If we look closely, we see that, although stock prices of the 10 largest companies increased first (in expectation of earnings gains), the contribution from earnings is now catching up to the market cap in the group.
Chart 4: Top 10 companies are responsible for 30% of earnings in S&P 500
In fact, earnings in the five largest stocks are now rising faster than stock prices. As a result, price-earnings multiples (a simple valuation measure) are now falling for these mega-cap companies.
Chart 5: P/Es for the top 5 companies in the S&P 500 vs. remaining 495
AI is also contributing to the economy
Importantly, AI spending is not just helping U.S. indexes grow earnings — it is also adding to economic activity.
One estimate has AI investment accounting for 92% of GDP growth in the first half of 2025. Data center construction has grown to nearly match office construction spending.
Other estimates show AI spending adds more to GDP growth than the U.S. consumer this year.
Research from Citi shows that AI equipment investment has increased 0.9% (as a share of GDP) since 2023 (blue line). That’s a ~$270bn increase. Other data shows hyper-scalers spending $350bn a year on AI buildout.
These numbers are clearly very large. So, it’s interesting to compare the AI investment cycle to the scale of other historic, decade-long investment cycles – for things like Railroads, Electricity, Automobiles, and the Internet. As the chart below shows, the current AI cycle is just over half of what was spent on the internet, as a proportion of GDP. For comparison, other estimates show Railroad spending was ultimately around 5-times larger than internet spending.
Chart 6: Comparing AI to other large infrastructure spending cycles
One question is: Will all this investment pay off?
Clearly railroads and electricity changed productivity in major ways – decades ago. Even the internet revolution was followed by around a decade of above-average productivity growth, increasing GDP by more than 15% above prior trend rates.
To put that in context: The U.S. is a $30 trillion economy. Even if AI only adds 10% to productivity, it will add $3 trillion to output. By that measure, the economic benefits of AI seem likely to outweigh the costs of all the investments that are driving earnings that we are seeing today.