AI Earnings Set to Drive Two-Thirds of S&P 500 Q3 Profit Growth

Expectations rise again, and AI is behind most of it

Wall Street begins another earnings season expecting strong numbers, and the forecasts depend heavily on artificial intelligence. Analysts on average expect S&P 500 earnings to rise about 31% from a year earlier in the third quarter. According to LSEG's head of earnings and equity research, Tajinder Dhillon, about two-thirds of that increase will come from the technology sector and from Alphabet, Amazon.com and Meta Platforms.2 The forecast lands as tech gains have pushed the S&P 500 to a record high. Reporting season starts unofficially with JPMorgan Chase and Goldman Sachs.3

Some strategists put AI's share even higher. Sameer Samana, head of global equities and real assets at Wells Fargo Investment Institute, said he would not be surprised if tech and AI turned out to explain 70% to 80% of the growth. In his view, the rest comes mostly from energy and materials, which are benefiting from geopolitics.2 He also pointed to the sectors with no AI exposure. Consumer staples and real estate have some of the weakest year-over-year growth estimates for the quarter.2

The big takeaway is that the headline number makes earnings look broad when they are not. A 31% index-wide gain mostly reflects a handful of companies building or buying AI infrastructure, plus energy producers riding an oil shock.

Different forecasters, different numbers

Everyone agrees growth is strong. They disagree on how strong. LSEG's October 9 scorecard put blended third-quarter growth at 30.6%, down from 53.7% in the second quarter. That blended estimate stood at just 15.1% on January 1.19 FactSet's figure was 29.6% as of the same date, up from 26.7% when the quarter began on June 30.13 Goldman Sachs data show a lower consensus of roughly 27%, and Deutsche Bank's Bankim Chadha expects 34%.10 Zacks is the lowest, with an estimate of about 24.1% as of October 1.23

Part of the gap comes from how each firm measures earnings. The direction, though, is the same across all of them, and it is unusual. Analysts usually cut quarterly estimates as the quarter goes on, by an average of 2.2% over the past five years. This time FactSet found they raised third-quarter per-share estimates by 1.4% between June 30 and September 30.12 Company guidance points the same way. Of 116 S&P 500 companies that issued third-quarter guidance, 72 guided above expectations. That is the highest count since FactSet started tracking it in 2006, and 44 of those companies are in information technology.13

Revenue forecasts are rising too, though more slowly. FactSet expects 12.4% sales growth, which would make three straight quarters of double-digit revenue gains.13 Earnings growing more than twice as fast as revenue means much of the forecast depends on wider margins.20 FactSet puts the expected net profit margin at 15.0%. That is below the previous quarter's 17.0% but well above the 13.0% of a year earlier.13

Beats are likely. They may not be enough

Earnings seasons tend to end higher than they start because most companies beat estimates. FactSet has calculated how much this usually adds. Using the most conservative of its historical averages, third-quarter growth would end near 35.6%. Using the past four quarters, when beats were unusually large, it would reach 43.3%.13 The early reports fit that pattern. Of the first 19 S&P 500 companies to report, 84% beat the average estimate, and in total they came in 4.9% above forecasts.13

That is why the 31% figure is best read as a minimum. Strategists still doubt the third quarter can top the second.2 Second-quarter earnings rose nearly 54%, the fastest pace since 2021, though that figure was inflated. Without mark-to-market gains on AI-related investments at Alphabet and Amazon.com, growth was closer to 35%, which was still a post-2021 high.2 So the real benchmark is the 35% underlying figure, not 54%. A third quarter in the mid-30s would basically hold steady, not slow down.

For investors, the question is no longer whether companies will beat but by how much. One preview argued that a beat is now only the starting point. Hyperscalers such as Microsoft, Alphabet, Amazon and Meta face growing pressure to show AI revenue catching up with their spending.

Who is actually doing the work

The coverage agrees that growth is concentrated but names different companies at the center. The Reuters account credits Alphabet, Amazon and Meta, alongside the broader tech sector.2 Another preview says ten companies should produce more than two-thirds of index earnings growth, with Micron and Nvidia alone accounting for more than a third. It adds that AI infrastructure suppliers as a group, meaning makers of chips, memory, networking gear and power equipment, should deliver more than half of the index's per-share growth.10

Both accounts can be true at once. The hyperscalers spend the money and the chipmakers collect it, so both show up in the numbers. LSEG data show US semiconductor earnings expected to rise about 136% in the third quarter, down from about 158% in the second.3 Micron shows how large the windfall has been. When it reported on September 30, its net income had risen to $37.7 billion from $3.2 billion a year earlier.10 The company also forecast revenue above estimates and said customer commitments under long-term supply agreements had grown to $32 billion.5 FactSet's own numbers show how much depends on chips. Without the semiconductor industry, expected growth for the technology sector falls from 63.5% to 24.2%.

The timing of reports matters as well. Only about 7% of S&P 500 market value reports in the first week. The busiest stretch runs from October 26 to 30, when 44% of the index by value reports, including Alphabet, Meta and Amazon. Nvidia does not report until November 19.10 Bank results will shape the mood early on, but the AI picture will not be clear until early November.

Energy is the second engine

The other major source of growth comes from the war, not from AI. Energy earnings are expected to rise about 115% after US oil prices climbed roughly 30% during the quarter amid the US-Israeli war with Iran.2 FactSet's estimate is higher, at 121.1%, up from 79.3% at the end of June. It notes that West Texas Intermediate crude averaged $85.68 in the quarter, 32% above a year earlier.22 Supply problems through the Strait of Hormuz have been a key factor in those prices.16

Energy profits like these depend on oil prices staying high. That leaves AI as the only large source of growth that analysts treat as structural, which increases how much the market relies on it.

Peak growth, cooling revisions and higher rates

The main concern this season is whether growth can keep accelerating. Ameriprise Financial's chief market strategist, Anthony Saglimbene, warned that the cycle may be nearing peak earnings growth. He argued that each quarter of continued AI capital spending raises the return investors will demand and invites closer scrutiny.2 Nick Raich of the Earnings Scout said estimate revisions are starting to lose momentum. In his words, the AI buildout is still moving at 100 miles an hour, down from 150 three months ago.3

Interest rates add pressure. US bond yields have climbed on inflation worries, higher oil prices and debt problems in France and elsewhere, and Samana singled out heavy borrowers such as utilities as vulnerable.2 One preview noted that Treasury yields have topped 5% and the Federal Reserve is leaning toward raising rates again.10 A stronger dollar is a smaller headwind, since FactSet estimates that 42% of S&P 500 sales come from abroad.16

Valuations: cheaper than they look, but fragile

One point in the market's favor is that valuations have not risen along with prices. FactSet put the forward 12-month price-to-earnings ratio at 19.1, equal to the 10-year average, below the five-year average of 19.8 and down from 20.4 at the end of June.13 Reuters had earlier reported that LSEG Datastream showed the multiple falling to 19.2 from 22 at the start of the year.17 In plain terms, earnings estimates have risen faster than stock prices. That makes the record highs easier to justify than the headlines suggest.

The weakness is market breadth. One analyst described the market's narrowness as about the weakest since the dot-com bubble.10 Concentration on that scale was dangerous in 2000. What is different now is that today's leaders earn large amounts of cash, though that offers no protection if AI spending slows.9

The likely outcome is another quarter of earnings beats that pushes reported growth into the mid-30s. The market's reaction, however, will depend less on those beats than on what Alphabet, Meta, Amazon and later Nvidia say about spending and returns in 2027. Analysts already expect full-year 2027 growth of about 15.8%, roughly half of 2026's pace.16 The real test this season is whether companies can make that slowdown look gradual.