New: our quarterly report on the S&P 500, read from 500 companies’ own filings and calls. Read the Q3 2026 report

The S&P 500 in its own words — Q3 2026

Built from the companies’ filings and earnings calls, analysts’ consensus and the news, to 18 November 2026. Nearly all 500 companies have reported the quarter; 448 have filed their quarterly or annual report, and the rest file through November. How the figures were made is in the appendix.

The quarter in brief

  • Profit grew much faster than sales. Revenue rose 14% on the year and net income 35%; 85% of companies beat analysts’ earnings forecasts, but only 66% beat on revenue, close to the fewest in five seasons.
  • The AI build-out now leans on borrowed money. Alphabet, Amazon, Meta and Microsoft spent twice as much as a year ago on data centres and went from $63 billion of free cash flow to a deficit of $25 billion. They borrowed the difference.
  • Chips and oil did most of the growing. Technology grew revenue 40% and energy 36%; Micron alone added $34.5 billion of quarterly profit.
  • Households are spending, but trading down. Consumer companies’ revenue growth slowed from 10% to 6% on the year. The strain shows in what people chose: fewer new homes, cheaper cigarettes, and spending kept up by saving less.
  • Two windfalls flattered the quarter. Tariff refunds after a Supreme Court ruling, and refining margins of up to three times normal after the war with Iran. Neither is likely to repeat on the same scale.

The S&P 500 sold $5.42 trillion of goods and services in its latest quarter and kept $768 billion of it as profit, which on the face of it makes this an exceptional quarter, until you ask where the growth came from. Technology and energy grew revenue 38%. The other 387 companies grew 7.7%, a little faster than the economy in dollars, while the real economy grew about 2%. The headline describes a handful of industries, not the economy they sit in, and this issue is about the distance between the two.

Underneath, two things changed the shape of the index. The companies building artificial intelligence are spending more than their businesses bring in, and a war with Iran that began in February reached company results in full: refiners earned record margins, oil producers lost output in the Gulf, and airlines paid billions more for fuel. A third change was a windfall. In February the Supreme Court ruled that tariffs imposed under the International Emergency Economic Powers Act were unauthorized, and the refunds arrived this quarter. More than 100 companies wrote about them, and for retailers they were the difference between a good quarter and a flat one.

The scorecard

Each line is a measure from this issue, with our assessment of it and how it moved since the second quarter.

Measure This quarter Assessment Since last quarter
Earnings against forecasts 85% beat on EPS Strong Continuing: 85% in Q2 too
Revenue against forecasts 66% beat on revenue Weaker Weakening: 71% in Q2, near the lowest in five seasons
Revenue growth +13.8% on the year; +7.7% outside technology and energy Strong but narrow Slowing: +15.4% in Q2
The index against the economy Revenue +13.8% against GDP in dollars +5.9%; real GDP +2.1% Wide gap, set by two sectors Narrowing slightly: 15.4% against 6.3% in Q2
Profit growth Net income +35% on the year Very strong Partly one-off: energy +151%
Free cash flow Up 6% to $621bn; the four AI spenders −$25bn Weakening New: three of the four in deficit
Debt +10% to $10.2tn; the four AI spenders nearly tripled theirs Rising Accelerating
Household demand Consumer cyclical revenue +5.9% on the year; saving rate 4.5% Softening Weakening: +9.7% in Q2
Capital allocation Capex 48% of operating cash, the most in ten years Shifting to building Accelerating
Company guidance 70 of 106 raised full-year EPS guidance, 12 cut Confident Continuing
Valuation 24 times earnings, from 29 a year ago Lower Continuing: 25.3 at 30 June

Assessments are ours. “Since last quarter” says whether a trend is continuing, accelerating, weakening, new, or likely to be one-off.

Sectors at a glance

Sector Revenue y/y, Q3 Net income y/y, Q3 Revenue y/y, Q2 The quarter’s story
Technology +40% +81% +35% AI hardware and memory sold out
Healthcare +6% +28% +7% Insurers losing government-funded members
Financial +4% −6% +11% Investment gains swung; rates rising again
Consumer Cyclical +6% +9% +10% Shoppers trading down; tariff refunds
Consumer Defensive +7% +14% +9% Costs passed on, volumes under strain
Industrials +14% +43% +12% Freight to rail; power equipment backlogs
Energy +36% +151% +43% The war’s windfall, mostly for refiners
Communication Services +14% +13% +14% Alphabet and Meta borrowing to build
Utilities +11% +33% +5% Data-centre demand, paid for with debt
Real Estate +12% +2% +12% Refinancing at higher rates
Basic Materials +10% +42% +11% Steel and aggregates busy; less gold mined

Amazon is in consumer cyclical and Alphabet and Meta in communication services, which is why those two sectors carry most of the AI spending outside technology.


Part one: the quarter

What happened in corporate America between July and September: the results, the stories behind them, what companies made and moved, and the state of their finances.

Results against expectations

Profit beat, sales less so

Earnings beat expectations almost everywhere. Of 482 companies whose quarter can be compared with its consensus, 409 beat on earnings per share, 20 matched it and 53 missed. The comparison is with the adjusted figure analysts forecast, which leaves out one-off items. That 85% beat rate is the highest of the last five seasons, and the typical beat was about 5%.

Share of companies beating the consensus S&P 500, by earnings season 60%68%75%82%90%Q3 25Q4 25Q1 26Q2 26Q3 26EPS 85%Revenue 66%

Revenue was a different matter. Only 66% of companies beat on sales, level with the lowest share in the five seasons, and 18% fell short, so the extra profit came less from new customers than from what companies kept of each sale. Where sales did surprise, the war usually explains it: analysts expected $33.6 billion from Marathon Petroleum, and it reported $56.5 billion. Technology beat almost without exception, 81 of its 87 companies on earnings with a single miss, while consumer discretionary missed 13 times out of 52.

Speeding up or slowing down

Comparing one quarter with the one before mostly measures the calendar: utilities sell the most power in summer, retailers before Christmas. Growth on the same quarter a year earlier takes the season out, so the useful question is whether that growth is rising or falling from one quarter to the next. Here it is for the last five quarters, over the same 494 companies. That is more than the roughly 445 that have filed for the third quarter: a company is counted here once it has nine quarters of results, and tables built on this season alone use the companies that have reported it.

Sector Q3 2025 Q4 2025 Q1 2026 Q2 2026 Q3 2026 Direction
Technology +16.2% +19.7% +27.1% +35.0% +39.9% Faster for a year
Energy +0.6% −2.2% +4.8% +42.5% +36.4% One step, mostly held
Communication Services +9.2% +11.6% +14.3% +13.7% +14.3% Steady
Industrials +5.2% +9.8% +9.1% +12.4% +14.4% Faster
Real Estate +9.1% +10.0% +11.9% +12.1% +11.7% Steady
Basic Materials +4.9% +3.1% +9.6% +10.7% +9.5% Steady
Utilities +7.1% +11.5% +14.1% +5.4% +10.5% Uneven
Consumer Defensive +5.6% +5.6% +7.3% +8.9% +6.7% Slower
Consumer Cyclical +8.5% +7.3% +9.2% +9.7% +5.9% Slower
Healthcare +10.4% +10.5% +7.1% +7.5% +6.4% Slower
Financial +5.8% +6.6% +7.8% +11.5% +3.9% Much slower
Index +8.2% +9.1% +10.9% +15.4% +13.8% Slower, from a high

Revenue growth on the same quarter a year earlier, by each company’s own fiscal quarter, over companies with nine quarters of results.

Revenue growth on a year earlier, by quarter Technology against the slowing sectors and the S&P 500 0%10%20%30%40%Q3 25Q4 25Q1 26Q2 26Q3 26Technology 40%Index 14%Consumer Cyclical 6%Financial 4%

The index grew a little less quickly than in the spring, but the average hides two economies moving apart.

  • AI hardware is still accelerating. Semiconductor revenue grew 33% a year ago and 94% this quarter; computer hardware went from 12% to 80%, chip equipment from 14% to 37%. Industrials sped up behind them, led by makers of electrical and building equipment.
  • Energy took one step and stayed there. The sector grew 5% in the first quarter, 43% in the second and 36% in the third. That growth will fall out of the comparison a year from now.
  • Everything the household touches slowed. Restaurants went from 9% growth in the first quarter to 1%, travel companies from 12% to 6%, and consumer cyclicals as a whole had their slowest quarter of the five.
  • Financials fell back hardest, from 11.5% to 3.9%. Asset managers’ revenue, which moves with markets and fees on gains, fell 4% after growing 23% in the spring, and life insurers’ fell 4%. Banks slowed from about 9% to 3%.

Winners and losers

A quarter has two scorecards: what the business did, and what investors paid for it. This season they often disagreed, so we keep them apart.

On the results. The operating winners came from two places: the companies that sell the parts of AI data centres, and the companies that sell oil.

Operating winner Net income, a year ago → now Revenue growth
Micron $3.2bn → $37.7bn +379%
Nvidia $31.9bn → $65.3bn +103%
Broadcom $4.1bn → $13.1bn +86%
SanDisk $0.1bn → $8.1bn +378%
Exxon Mobil $7.8bn → $16.0bn +33%
Marathon Petroleum $1.9bn → $7.9bn +61%

Micron and Nvidia alone added $68 billion of quarterly profit. Memory makers did best of all, because the shortage described under the big stories lets them charge more for every chip: SanDisk earned seventy times what it did a year ago. The refiners’ revenue ran far beyond forecasts; analysts expected $33.6 billion from Marathon Petroleum and it reported $56.5 billion.

Operating loser Net income, a year ago → now Revenue growth
Alphabet $35.0bn → $26.5bn +16%
AT&T $9.7bn → $1.3bn +5%
Fiserv $0.8bn → −$5.9bn −1%
Uber $6.7bn → $0.6bn +10%
Johnson & Johnson $5.2bn → $0.2bn +6%
Carrier $0.5bn → −$2.5bn +12%

The operating losers are a different kind of story. Most of them grew: AT&T’s, Uber’s and Johnson & Johnson’s revenue rose while their profit fell by more than four-fifths, so the fall happened below the revenue line, in charges or in a year-ago quarter that carried gains.

In the market. Set the results beside the share price and the companies sort differently.

Company The quarter’s results Market value, 30 June → 4 November Market and results
Marathon Petroleum Profit four times a year ago +59% Agreed
Exxon Mobil Profit doubled +29% Agreed
Alphabet Profit down a quarter; missed forecasts by 28% −20% Agreed
Carrier Swung to a loss −26% Agreed
SanDisk Profit seventy times a year ago −47% Parted
Lam Research Revenue up 55%, in line with forecasts −44% Parted
Applied Materials Revenue up 25% −45% Parted
Micron Profit twelve times a year ago −25% Parted
Fiserv Swung to a loss of $5.9bn +19% Parted
Air Products Swung to a loss of $1.4bn +15% Parted

Where the market agreed, it was rewarding cash it could see: oil sold at war prices, and punishing Alphabet for spending that ran ahead of earnings. Where it parted company with the results, it was looking past the quarter in both directions. The AI suppliers posted the best results in the index and lost between a quarter and half of their value, on the doubt, set out under the big stories, about how long borrowed money keeps paying for the build-out. A few companies with poor quarters gained, as investors treated their losses as one-off charges rather than a weaker business.

A strong quarter, then, was not enough. Investors paid for profits they expected to last, and marked down the ones they thought the cycle had only lent.

Net income and revenue for the latest quarter against the same quarter a year earlier. Market value from 30 June to 4 November 2026.

The big stories

Every quarter companies rewrite a little of their filings: a new risk, a sentence added to the outlook, a cost they now have to explain. We read what changed in 428 companies’ filings this season and sorted the new text by subject. The three largest new subjects are all things that happened in the past nine months: tariffs rebuilt after a court ruling, an oil shock from a war, and a memory shortage created by AI spending. Each has its own story below.

Subject Companies affected Change on the summer Change on a year ago Who writes about it, or what changed
Rising: companies adding text
Tariffs and trade policy 192 +14 +48 Utilities (65%), consumer staples (62%), industrials (53%)
Rising input costs 188 +34 +86 Basic materials (80%), energy (78%), consumer staples (54%)
Interest rates and financing 150 +45 +52 Financials (58%), real estate (50%)
Cyberattacks 107 +36 +49 Technology (44%), communication services (45%)
China and export controls 95 +31 +16 Consumer staples (38%), healthcare (30%), technology (29%)
Fuel and energy costs 84 +11 +63 Industrials (29%), consumer discretionary (31%)
The Middle East conflict and shipping 67 −11 +58 Energy (44%), basic materials (33%)
Weather and catastrophes 62 +25 +16 Insurers
Labour and immigration 52 +25 +22 Real estate, services
Weaker consumers 50 +7 +16 Consumer discretionary (33%), consumer staples (35%)
Memory and component shortages 44 +11 +30 Technology (23%)
Government health funding 41 +9 −8 Healthcare (35%)
AI taking business away 32 +16 +20 Media, software
Power and grid capacity 17 +6 +10 Utilities (55%)
Fading: companies cutting text
The pandemic 49 +13 +19 50 more paragraphs cut than written
Russia and Ukraine 42 +12 +26 61 more cut than written, the most of any subject
Elections 40 +7 +27 Written in a year ago, now written out
Recession 19 −4 +12 18 more cut than written
Government shutdown 19 −2 +17 Written in a year ago, now written out
Strikes 17 +3 +10 4 more cut than written
End of the de minimis exemption 9 0 +7 Written in a year ago, now written out
Red Sea shipping 3 −1 0 Replaced by the Strait of Hormuz
Banking turmoil 3 +2 +3 No company wrote anything new

Companies writing new text on a subject (rising) or deleting text about it (fading) in their 10-Q and 10-K filings this season (428 companies read), against the summer season (filings 15 April – 14 July, 402 read) and the same season a year ago (377).

More companies wrote about almost every worry than in the summer, and far more than a year ago. The largest rises on the summer were interest rates (45 more companies), cyberattacks (36), input costs (34) and China (31); on the year, input costs (86 more), fuel (63) and the Middle East (58), which a year ago only nine companies mentioned. Only the Middle East fell against the summer, and only government health funding against a year ago. At the same time older subjects are being edited out. Russia and Ukraine, the pandemic and recession lose text season after season. Last autumn’s new subjects — elections, the government shutdown and the end of the de minimis exemption — reversed within a year. No company added a word about bank failures, and Red Sea shipping, the worry of 2024, gave way to the Strait of Hormuz. Interest rates swung most: in the summer companies cut 221 more paragraphs about rates than they wrote, and this season the balance turned positive again, as analysts on calls asked about rates more than three times as often.

AI: the build-out turns to borrowing

For two years the largest technology companies paid for their AI build-out from the cash their businesses produced. This quarter that stopped being true.

Company Capex, quarter A year ago Free cash flow, quarter A year ago Debt now A year ago
Amazon $75.9bn $34.2bn −$33.5bn $1.3bn $173.1bn $50.7bn
Alphabet $51.4bn $24.0bn −$9.1bn $24.5bn $125.1bn $21.6bn
Microsoft $39.4bn $19.4bn $18.6bn $25.7bn $35.5bn $43.2bn
Meta $33.9bn $18.8bn −$1.2bn $11.2bn $80.9bn $28.8bn
Together $200.6bn $96.4bn −$25.2bn $62.7bn $414.6bn $144.3bn
The four biggest AI spenders, one quarter Amazon, Alphabet, Microsoft and Meta combined, latest quarter against a year earlier A year agoThis quarterCapital spending$96bn$201bnFree cash flow$63bn−$25bnDebt$144bn$415bnCash$339bn$526bn

Amazon alone ran a deficit of $33.5 billion in the quarter, more than the free cash flow of any single company in the index, and Alphabet and Meta also spent more on capital projects than their operations brought in. Between them the four nearly tripled their debt in a year, though not all of it has been spent yet: their cash rose from $339 billion to $526 billion, held for building still to come.

On the calls, nobody sounded worried. Amazon’s Andy Jassy said he had “even more conviction” that the AWS investment would pay off, and that Amazon would spend “meaningfully more” in 2027 than the $220 billion planned for this year. Alphabet’s finance chief said the spending “will put pressure on free cash flows”; in July she had said Google was renting other companies’ computing capacity as a bridge, because its own was short. Analysts asked about the money rather than the demand. Asked how Microsoft would split its funding between debt and equity, Amy Hood said it had used leases and had not yet tapped the bond market, which is why Microsoft is the one of the four whose debt fell. Meta said it did “not have any specific financing plans to announce at this time”.

Against what? Mostly a backlog of signed contracts, which companies report as remaining performance obligations: revenue already contracted but not yet earned. Across the 121 companies that report it, the backlog reached $5.45 trillion, up 47% in a year, and three companies account for half of the increase. Alphabet’s rose from $158 billion to $568 billion, Microsoft’s from $398 billion to $683 billion, and Oracle’s from $455 billion to $664 billion.

The news filled in who signed them. Microsoft said OpenAI drove most of its increase, and OpenAI committed to buy a further $250 billion of Azure services. A large part of Oracle’s backlog is tied to OpenAI as well. Anthropic committed to spend roughly $450 billion on cloud computing, with agreements at both Alphabet and Amazon. Much of the index’s contracted growth now rests on two private companies that are themselves spending well ahead of their income.

Through September the argument filled the news: Michael Burry compared Nvidia to the dot-com era, Fitch warned about circular financing between chipmakers and their customers, and a Bank of America survey named an AI bubble as one of investors’ two biggest risks. Alphabet then missed earnings expectations by 28%, and Amazon by 6%. Between 30 September and 4 November Alphabet lost 18% of its value, Meta 16%, Nvidia 10%, Micron and AMD 20% each, Lam Research 29%, and Intel almost half. Berkshire Hathaway, Eli Lilly, Walmart and Visa each gained between 17% and 23% over the same weeks.

The money flows outward to the suppliers who sell into the build-out. Micron, which sells the memory these systems need, turned $0.1 billion of free cash flow a year ago into $32.9 billion this quarter, on revenue of $54.2 billion. Its market value rose from $183 billion to $1.2 trillion in twelve months, taking it into the ten largest companies in the index. Semiconductor companies as a group made $99.8 billion of free cash flow in the quarter, against $38.2 billion a year earlier.

Everyone else who buys memory pays for the shortage. AI data centres are absorbing the world’s supply of memory chips. Apple wrote that it is “experiencing a period of supply constraints and increasing costs” for advanced chips and memory, and that it “expects these trends to intensify”. Cisco said it expects to keep being hurt by rising memory costs, and its inventory rose 80% to $5.7 billion partly because of them. Microsoft warned that component shortages “have and may increase” its cost of revenue. Dell, Hewlett Packard Enterprise and Flex wrote the same. Flex passes the cost on where its contracts allow, which raises its sales and lowers its margins.

Our reading: the spending is real and so is the demand behind it. But the index’s largest companies have moved from funding AI out of profit to funding it with debt, and their contracted revenue depends heavily on customers who are borrowing too. That makes the next few quarters’ cash flow, not their earnings, the number to watch.

Tariffs: the bill keeps changing

Tariffs remain the most common new worry, but the reason changed this year. After the Supreme Court ruling in February, companies began booking refunds of the emergency tariffs. At the same time the government turned to other powers, chiefly Section 232 investigations into specific industries. The tariffs that were refunded are being replaced by new ones aimed more narrowly.

The result is a cost companies cannot plan around. Pfizer described the April 2026 order imposing Section 232 tariffs of up to 100% on imported patented medicines. Amgen wrote that it had won relief from them for about three years because of its US manufacturing investment. Apple described the investigation into semiconductors and the products that contain them. Nvidia wrote that it had been “unable to pass along any of the tariff” on the chips it is licensed to sell in China. Target summed up the problem for retailers: the mix of tariffs, refunds, sourcing, pricing and consumer response “could materially impact” its results.

Utilities wrote about tariffs most often, because their building programmes depend on imported transformers, switchgear and solar panels. Consumer staples came next: they sell imported or ingredient-heavy goods at thin margins.

The war with Iran: a windfall for some, a bill for others

The conflict that began in February between the United States, Israel and Iran is written into the filings of companies far from the oil business.

Energy companies are the ones it helps. Chevron wrote that each dollar on the Brent price moves its annual earnings and cash flow by about $600 million. Diamondback raised its production guidance because of “global oil supply constraints that began in the first quarter of 2026”. Valero said demand for transport fuel stayed resilient “amid constrained worldwide supply”. Their worry is the reverse: Valero listed routes that bypass the Strait of Hormuz, releases from strategic reserves and more production elsewhere as things that could ease prices.

Everyone else describes a bill. Lennar said a cost increase was “driven in significant part by inflation associated with energy prices tied to the ongoing geopolitical tension with Iran”. Lowe’s named “elevated fuel prices” alongside weak demand for do-it-yourself projects. Carnival said further escalation could raise its fuel costs again. Mosaic, which ships fertiliser through the Gulf, listed disruption in the Strait of Hormuz as a risk to its trade. Sysco said it may not be able to pass on higher fuel costs in full.

What the real economy is doing

The economy as a whole

The official figures for the third quarter describe an economy growing at an ordinary pace, with prices doing more of the work than they did a year ago. We set them beside the index’s own revenue, split between the two businesses running hot, technology and energy, and the companies that are neither.

Growth on a year earlier Q3 2025 Q4 2025 Q1 2026 Q2 2026 Q3 2026
Real GDP +2.6% +2.1% +2.6% +2.2% +2.1%
Real consumer spending +2.7% +2.1% +2.1% +2.5% +2.5%
Real retail sales +1.0% 0.0% +0.4% +1.7% +1.9%
Industrial production +1.7% +1.6% +0.8% +1.5% +1.7%
Jobs +0.5% +0.2% +0.2% +0.2% +0.3%
Consumer prices +2.9% +2.8% +2.7% +3.8% +3.5%
GDP in dollars +5.5% +5.1% +5.8% +6.3% +5.9%
S&P 500 revenue, technology and energy +10.4% +12.0% +19.1% +37.6% +38.7%
S&P 500 revenue, everyone else +7.6% +8.4% +8.8% +10.0% +7.7%

Official figures from the Bureau of Economic Analysis, the Bureau of Labor Statistics, the Census Bureau and the Federal Reserve, as quarterly averages. Company revenue by each company’s own fiscal quarter.

The index against the economy Growth on a year earlier, by quarter 0%10%20%30%40%Q3 25Q4 25Q1 26Q2 26Q3 26Tech and energy 39%Everyone else 8%GDP in dollars 6%Real GDP 2%

Three things stand out.

  • The real economy grew about 2%, as it has all year. Output, consumer spending and factory production all grew at much the same pace as a year ago. Nothing in the official figures looks like the boom in the index’s headline numbers.
  • The index outside AI and oil tracks the economy. The 385 companies outside technology and energy grew revenue 7.7%, under two points faster than the economy in dollars, about the gap of a year ago. They slowed from the spring with it. The 38.7% at the top of the index is technology and energy alone.
  • Prices rose faster than a year ago, and quantities did not. Consumer prices rose 3.5% against 2.9% a year ago, in the year of the oil shock and the new tariffs. The volumes companies report say the same: across about 930 measures of things sold, shipped or served at 202 companies, the typical company’s volumes grew 2.8%, close to real GDP. The rest of their revenue growth was price.

Jobs barely grew, 0.3% in a year, while unemployment fell to 4.2%. The parts of the economy that borrow to buy were the weak spot: car sales were 1.1% lower than a year ago and housing starts 4.1% lower.

What companies sold

The volumes themselves (vehicles, barrels, kilowatt-hours, members, parcels) show where that 2.8% came from and where it did not. Grouped in four:

Energy and industry

The typical energy company’s volumes rose 4.8% and the typical industrial’s 1.7%, while materials companies shipped slightly less. The growth sits where the AI build-out and the war send it.

More electricity, less gas. Utilities sold more power almost everywhere. Southern Company’s retail sales rose 7.8% on the year, Entergy’s 7.4% and Ameren Missouri’s 6.8%. Gas went the other way: Alliant delivered 12.4% less and Sempra 7.9% less. Much of the extra electricity is going to data centres, which is what the utilities wrote about too.

Less fuel at home, more for export. Kinder Morgan’s pipelines moved 3.9% less petrol and 2.7% less diesel and jet fuel, as US pump prices rose. Refiners still ran harder, because the war cut supply elsewhere: Marathon Petroleum processed 4.8% more and Phillips 66 22.9% more.

Busy factories, less gold. Nucor shipped 12.5% more steel and Martin Marietta 15.9% more aggregates. Gold producers mined less even with gold near record prices: Freeport’s output fell 18.5% and Newmont’s 8.4%.

For these industries, demand now follows data centres and the war more than the business cycle. For the wider economy, it means the strongest physical growth is in power and building materials, and the weakest in the fuel households buy for themselves.

The consumer

The typical consumer company sold 4.3% more and the typical staples company 2.3% more, but the volumes split by how much a purchase costs and whether it can wait.

People travel, if they can afford the ticket. Airbnb booked 16% more nights, Booking 6.8% more room nights, and Norwegian Cruise Line carried 22% more passengers. Airlines flew almost exactly as many seats as a year earlier but charged 16% more for each seat mile.

Fewer petrol cars, and slower electric ones outside Tesla. General Motors sold 11% fewer vehicles and Ford’s shipments fell 9%; its electric Model e fell 32%. Tesla delivered 16% more. Used cars did well: Carvana sold 39% more.

Fewer homes sold, more built for rent. Every large homebuilder took fewer orders: Lennar 9.2% fewer, D.R. Horton and NVR 7.3%, PulteGroup 6.5%. D.R. Horton sold 20% more rental apartments. Its buyers were still there, its chief executive said, “just having a hard time moving forward”, and it paid for mortgage rate cuts on close to nine in ten of the homes its own lender financed.

Smokers trade down. Altria shipped 6.4% fewer Marlboro cigarettes but 40.5% more discount ones, and 13.3% more nicotine pouches. Philip Morris shipped 56% more e-vapour products. Altria put it plainly: prices have risen about 30% since 2021, petrol has taken more of what was left, and discount brands now hold 34.1% of the market, after gaining two points in the quarter.

For retailers and carmakers, volume now has to be bought with discounts, which is why their margins fell. For the economy, households are protecting trips and everyday spending and postponing the largest purchases: the pattern of a squeeze on real income, set out under the household in part two, not of collapsing demand.

Services and health cover

Service volumes grew about 3%. What changed is less how much was delivered than how, and by whom.

Freight shifts to rail. Union Pacific moved 4.2% more and CSX 5% more; J.B. Hunt’s rail-and-truck loads rose 8.5%. Trucking was flat, and UPS handled 3.3% fewer US parcels a day while FedEx handled 3.2% more.

Fibre wins over cable. AT&T’s fibre customers rose 53.5% and Verizon’s more than doubled, after it bought Frontier. Comcast lost 9.8% of its broadband customers, and Charter lost 156,000 internet customers in the quarter, more than the 109,000 it lost a year earlier, with Cox losing another 77,000.

Cards replace cash, and trading booms. Visa processed 9.6% more transactions and Mastercard 9.7% more purchases, while Mastercard’s cash withdrawals fell 5.1%. Interactive Brokers opened 35% more accounts and Charles Schwab 8.7% more new brokerage accounts.

Health insurers are shrinking. The large insurers covered fewer people, mostly in government programmes. Elevance lost 15.5% of its Medicare Advantage members and UnitedHealth 9.8%. Centene’s marketplace membership fell 41.6%, the largest drop in any measure we read, after enhanced tax credits for plans bought on the exchanges expired and it raised prices on 95% of those members.

For these industries, the winners are taking share from older networks (rail from trucks, fibre from cable, cards from cash) rather than growing the market. For the economy, the clearest change is the government’s retreat from health cover, which shows in membership long before it shows in spending.

Employment, the link between these volumes and the household, is in part two. Together, the volumes show an economy spending on power, travel and AI, and cutting back on cars, homes and fuel at the pump. The government’s step back from health insurance shows up here more clearly than in any revenue line.

Financial warning signs across the index

We flag filings that show a pattern worth a second look. Some are routine and some are serious:

  • Usually a question, not a problem: receivables or inventory growing faster than sales, and new shares diluting holders. Both are common in a company that is growing or buying another.
  • More serious: a material weakness, meaning the company or its auditor found a flaw in the controls that make its accounts reliable; a goodwill write-down, which admits an acquisition was worth less than was paid; and language about breaching the terms of its loans.
Sector Companies Any flag Receivables Covenant Dilution Audit
Utilities 31 52% 35% 16% 13% 3%
Technology 89 47% 36% 4% 12% 7%
Communication Services 22 45% 18% 14% 9% 5%
Basic Materials 19 42% 16% 11% 5% 0%
Energy 21 38% 24% 5% 14% 0%
Real Estate 30 30% 0% 23% 10% 0%
Consumer Cyclical 54 31% 9% 13% 4% 4%
Industrials 75 25% 9% 11% 1% 4%
Consumer Defensive 31 23% 13% 6% 0% 0%
Healthcare 60 20% 10% 3% 2% 2%
Financial 71 15% 3% 8% 3% 1%

Utilities and technology carry the most flags, and in both the reason is the build-out. Utilities’ cash goes into projects before it comes back as revenue. In technology, 36% of companies have receivables or inventory growing faster than sales. Amazon, Nvidia, Micron, Broadcom and Intel lead on receivables, and Apple, Dell, Broadcom, Cisco and Hewlett Packard Enterprise on inventory: the same memory squeeze that lifted Micron, seen from the buyers’ side.

The serious flags are rarer. Fourteen companies reported a material weakness this season, among them Microsoft, ADP, Copart, FactSet, Monolithic Power, Super Micro Computer, Starbucks and TransDigm. Kraft Heinz wrote down $2.4 billion of goodwill and International Paper $1.4 billion. Two companies stand out for carrying several flags at once. Super Micro Computer has five: a material weakness, language in its auditor’s opinion, rising receivables and inventory, and new shares. International Paper has a material weakness, a goodwill write-down and covenant language in the same quarter.

A flag matters more when it is new or getting worse than when a company has carried it for a year. Set against the same companies’ filings in the summer and a year ago:

Flag New this season Worsening Persistent Cleared since the summer
Receivables ahead of sales 29 8 16 9
Inventory ahead of sales 19 10 7 9
Dilution 8 6 16 11
Goodwill write-down 7 1 4 5
Material weakness 1 — 13 2
Covenant language 7 — 40 1

New: flagged this season and in neither earlier season. Worsening: flagged in the summer too, with a gap at least a tenth wider. Persistent: flagged in the summer or a year ago as well. Companies, by flag.

The working-capital flags are the new ones, and they follow the AI build-out. Inventory ahead of sales is new at Apple, Dell, Hewlett Packard Enterprise and NetApp, buyers stocking memory and components, and worsening at Cisco, Qualcomm and Super Micro. Receivables ahead of sales are worsening at Amazon, Micron and Intel, and newly flagged at five utilities (Dominion, DTE, Exelon, PSEG and PPL). The serious flags are mostly old: thirteen of the fourteen material weaknesses were already there in an earlier season, and the covenant flag catches the same standard loan language quarter after quarter. Seven companies took a first goodwill write-down, among them UPS, Cummins and Cardinal Health.

The index

Six companies joined the S&P 500 in the quarter, among them Bloom Energy, Ferguson, Illumina and Reddit, while AvalonBay, Builders FirstSource, Electronic Arts, Equity Residential, Molson Coors and The Trade Desk left. AppLovin moved from technology to communication services and DuPont from materials to industrials. Since the end of September, Okta and Twilio have joined, and Corteva, Kenvue, Paramount Skydance and Warner Bros. Discovery have left. None of this did much to the index’s make-up; prices and results did.

Sector Members Share of value, Q2 Q3 4 Nov Revenue, quarter Revenue growth Net income Free cash flow Cash Debt
Technology 88 36.6% 38.9% 36.3% $866bn +40% +81% +41% +24% +13%
Financial 70 12.0% 11.4% 12.3% $743bn +4% −6% — +9% +10%
Communication Services 19 10.7% 10.7% 9.3% $403bn +14% +13% −64% +90% +25%
Consumer Cyclical 54 10.2% 9.4% 10.1% $643bn +6% +9% −87% +2% +16%
Healthcare 60 9.0% 9.2% 10.1% $996bn +6% +28% −6% +1% +5%
Industrials 75 8.1% 7.4% 7.5% $459bn +14% +43% +6% +11% +1%
Consumer Defensive 31 5.1% 4.8% 5.4% $584bn +7% +14% +20% +18% 0%
Energy 21 2.9% 3.3% 3.6% $438bn +36% +151% +140% +56% −4%
Utilities 31 2.1% 1.8% 2.0% $138bn +11% +33% — +3% +14%
Real Estate 30 1.8% 1.7% 1.8% $49bn +12% +2% −13% 0% +2%
Basic Materials 19 1.5% 1.4% 1.5% $107bn +11% +42% +30% +4% +6%
Index 498 $67.8tn $69.9tn $69.0tn $5,423bn +14% +35% +6% +13% +10%

Share of the index’s value at the end of the second quarter (30 June), the third (30 September) and on 4 November; the index row gives its total value. Revenue is the latest quarter; every growth figure is against the same quarter a year earlier.

Technology reached 38.9% of the index’s value at the end of September, earning a third of its profit and 16% of its revenue, and gave back two and a half points in five weeks. Health care produces the most revenue of any sector, $996 billion a quarter, but is a tenth of the index’s value; it was the sector investors moved towards in October. Energy had the best quarter on every line, earning two and a half times what it did a year ago while paying debt down.

The ten largest companies in the index now account for half of all its capital spending, up from 34% a year ago, but for only 16% of its free cash flow, down from 22%. They earn 32% of the index’s profit and are 40% of its value.

The ten largest companies' share of the S&P 500 Share of the index total, latest quarter against a year earlier A year agoThis quarterCapital spending34%48%Profit28%32%Operating cash flow26%30%Free cash flow22%16%Revenue14%16%
Net income growth by sector, a year on Latest quarter against the same quarter a year earlier Energy+151%Technology+81%Industrials+43%Basic Materials+42%Utilities+33%Healthcare+28%Consumer Defensive+14%Communication Services+13%Consumer Cyclical+8%Real Estate+2%Financial−6%

Our reading: the index ended the quarter more concentrated in a handful of companies whose cash flow is going the wrong way, and more of its contracted growth depends on customers outside it. Energy’s windfall and the tariff refunds flattered the quarter; neither is likely to repeat on the same scale.


Part two: what it means

Part one showed a very good quarter that was built narrowly. This part makes an argument about what it means: corporate America is investing for a very different future, but the returns, the financing and the economy behind that investment are becoming more uneven. The sections follow the argument in order. Companies are changing how they use their money. The growth that pays for it is concentrated, and more of it is price than volume. Management is confident, but analysts are asking harder questions. Investors have started to separate the profits they trust from the ones they do not. And the next phase depends on whether this year’s spending earns a return.

We set this quarter against ten years of the same companies’ results, so that what is unusual can be told from what is merely large.

How corporate America is paying for its future

The largest change in corporate behaviour this year is not how much companies earn but what they do with it. A growing share of their cash goes into building rather than back to shareholders, and the building is paid for in three very different ways.

Building replaces buying back

Companies outside banking and insurance now put almost half their operating cash into capital projects, the most in ten years, and less into buybacks and dividends than at any time in that period.

Sector Operating cash, 12 months Capex as share of it Net buybacks as share of it Dividends as share of it
Technology $1,012bn 32% (was 28%) 29% (was 39%) 13%
Communication Services $513bn 69% (was 50%) 7% (was 27%) 9%
Consumer Cyclical $383bn 76% (was 56%) 19% 12%
Healthcare $347bn 20% 23% (was 26%) 28%
Energy $302bn 37% (was 49%) 21% (was 28%) 21%
Industrials $260bn 27% (was 32%) 29% (was 37%) 22%
Consumer Defensive $200bn 35% 12% (was 21%) 40%
Utilities $152bn 154% Net issuers 27%
Index, excluding financials $3,305bn 48% (was 45%) 19% (was 27%) 19% (was 23%)

Twelve months to each company’s latest quarter, against the twelve months to a year earlier. Net buybacks are shares bought back less shares issued, so a company issuing more than it buys counts against the total.

Where operating cash goes Share of twelve months' operating cash to June each year, S&P 500 excluding financials 10%25%40%55%70%'17'18'19'20'21'22'23'24'25'26Capex 47%Buybacks 19%Dividends 18%Free cash flow 53%

Capital spending rose from $1.15 trillion to $1.60 trillion; net buybacks fell from $687 billion to $635 billion and the share of cash paid as dividends fell too. The shift is not a reallocation between buybacks and dividends. Both are losing ground to investment at once, and both are at their smallest share of cash in the last ten years.

Three questions decide what it means.

  • Structural or a cycle? Over ten years capex took between 36% and 44% of operating cash. It is now 48%, and technology and communication companies account for 42% of all capital spending, against about a third in 2017. A shift that large in who spends is closer to a change in the economy’s structure than a cycle in one industry.
  • Opportunity or obligation? Part of the spending is chosen: Amazon, Alphabet, Meta and Microsoft are building against signed contracts. Part is forced. Utilities spend 154% of their operating cash because data centres have asked for power they are obliged to supply, and retailers and carmakers keep investing through weaker volumes to hold their place. Energy is the one large sector doing the opposite, spending less on new wells and more on dividends and debt.
  • What shareholders give up. Alphabet went from buying back a net $55 billion of its shares to issuing a net $25 billion; Intel and Oracle issued $21 billion each. Fewer buybacks mean slower growth in earnings per share for the same profit, and the payoff for owners is deferred to whenever the new capacity earns.

Three kinds of borrower

Debt across the index rose 10% to $10.2 trillion, and cash 13% to $9.4 trillion. Measured against the cash companies generate, debt is at its lowest in ten years: 2.1 times a year’s operating cash, against 2.5 a year ago and around 3 in 2016–2020. If there is a problem, it is not the total but where the borrowing sits, and it comes in three kinds that should not be read together.

  • Borrowing to grow, from strength. Amazon, Alphabet and Meta borrowed heavily while holding more cash than before (part one). Their debt is a choice to build faster than profits allow.
  • Borrowing because the cash is not enough. Regulated utilities added $91 billion of debt, to $911 billion, and utilities as a whole spent $60 billion in the quarter against $50 billion of operating cash. Oracle’s debt rose from $82 billion to $118 billion while its free cash flow fell to minus $5.4 billion. For these companies borrowing is the only way to keep building.
  • Building and repaying at once. Energy’s free cash flow rose from $30 billion to $72 billion a quarter and its cash by half, while it paid down $13 billion of debt. Apple cut its debt by $18.5 billion and health insurers by $15 billion.

Watch the second group. Small as it is, its spending depends on lenders’ terms rather than on its own cash, and it is the group most exposed to the higher rates analysts started asking about this autumn. None of the three is borrowing to keep up its payouts: the debt is paying for assets, which is a different risk from debt that props up dividends or covers a business in decline.

Debt added in a year, by industry Change in total debt, latest quarter against a year earlier Internet content (Alphabet, Meta)$156bnCapital markets$142bnInternet retail (Amazon)$122bnRegulated electric utilities$91bnLarge banks$77bnApplication software$46bnInfrastructure software$32bnIndependent power producers$28bnHealth insurers−$15bnOil and gas producers−$16bnApple−$18bn

The deals: capacity, consolidation and break-ups

The season’s deals had three motives, and only one of them is about growth.

  • Securing capacity. The largest deals bought access to AI hardware rather than companies. Broadcom is backing about $43.5 billion of debt in a $63 billion package to buy chips and networking equipment for Anthropic to lease, and Alphabet holds an option to buy up to $10 billion more of Anthropic’s shares. These are suppliers financing their own customers.
  • Buying scale in slow markets. Where volumes are flat, companies bought each other. Sysco agreed to buy Jetro Restaurant Depot with about $21 billion of new debt; Hormel is buying Brakebush, Paychex borrowed $5 billion for Paycor, and Fifth Third is merging with Comerica.
  • Simplifying. Abbott is separating its nutrition business, Genuine Parts its automotive business, Keurig Dr Pepper its coffee business, and Flex plans to spin off Axiom.

The pattern matches the rest of the quarter: where demand is growing, companies spend to secure supply; where it is not, they consolidate to cut cost, or split up so that investors can value the growing part on its own.

Who depends on whom

Reported growth can look broad while the customers behind it are few. Cloud customers were 89% of Western Digital’s revenue and data-centre customers 81% of Seagate’s. Microsoft and Meta together were 42% of Arista Networks’ revenue, one data-centre customer 44% of Super Micro Computer’s, and technology customers 59% of Comfort Systems’, a building contractor. Concentration is not new elsewhere: the US government is 72% of Lockheed Martin’s sales, advertising 98% of Meta’s revenue, and households earning $35,000 a year or less 60% of Dollar General’s.

What is new is that the two kinds of concentration reinforce each other. The suppliers growing fastest sell to a handful of customers, and those customers are themselves funding their purchases with borrowed money and with contracts from two private AI companies. The index’s growth is concentrated in the same few places on both sides of the sale.

What the money says

Three findings. The building is still paid for mostly from cash, but the borrowing is concentrated in two groups whose debt is growing faster than their cash. The spending has spread along the AI supply chain, not across the economy. And only part of it is investment that can earn; the rest is obliged or defensive, so the return on the first part is what the rest of this issue turns on.

The economics of growth

The question underneath the quarter’s growth is where it came from: companies selling more, charging more, or keeping more of each sale. The answer this year is mostly charging more and keeping more, and outside technology and energy, much less than the headline suggests.

Prices did more of the work than volumes

Over ten years the index’s companies outside technology, energy and finance have grown revenue a little over a point a year faster than the economy in dollars. This year they grew 8.1% against the economy’s 5.7%, a wider gap than usual, while the real economy grew 2.4%. Most of that difference is price, and the table shows where.

Company revenue against the economy Growth over twelve months to June, on the twelve months before 0%6%12%19%25%'17'18'19'20'21'22'23'24'25'26Technology 22%Everyone else 8%GDP in dollars 6%Real GDP 2%
Sector Volumes, a year on Revenue, a year on Gap
Energy +4.8% +38.4% 33.6
Basic Materials −0.9% +12.1% 13.0
Industrials +1.7% +13.5% 11.8
Technology +5.5% +12.5% 7.0
Financial +2.9% +9.6% 6.7
Communication Services +7.7% +13.2% 5.5
Utilities +1.1% +6.1% 5.0
Real Estate +0.3% +4.3% 4.0
Consumer Defensive +2.3% +5.7% 3.4
Consumer Cyclical +4.3% +6.6% 2.3
Healthcare +2.8% +5.0% 2.2
Index +2.8% +8.3% 5.5

Median company in each sector, over the 202 companies that report operating volumes (about 930 measures of things sold, shipped or served). The gap is price, mix and acquisitions together.

The table divides the economy in two. Energy, materials and industrials grew almost entirely through price: oil at war prices, and equipment and materials whose costs were passed on. The consumer sectors and health care grew mostly through volume, and kept little extra from price. Within them, two industries sold for less: travel companies booked 8% more for 6% more revenue, and homebuilders took in 8% less on 4% fewer homes. That is where higher prices stopped working. O’Reilly, the car parts retailer, shows the mechanism within one company: do-it-yourself transactions were under pressure, higher tickets more than made up for it, and like-for-like prices were up slightly over 4%. Its management warned that “the pressures consumers are facing could intensify”.

Margins: pricing power or a windfall

Gross margin, what companies keep of each dollar of sales after the direct cost of what they sold, has risen steadily for ten years, from 36.6% in 2017 to 40.7%, for the index outside financials. Almost all of this year’s rise came from technology. For everyone else outside technology, energy and finance, gross margin was 35.7%, the same as a year ago.

Gross margin, ten years Twelve months to June each year, S&P 500 excluding financials 30%34%38%41%45%'17'18'19'20'21'22'23'24'25'26Index 41%Outside tech and energy 36%

Two different stories sit inside the rise, and they should not be confused.

  • Windfalls from scarcity. Semiconductors’ margin went from 58.0% to 68.6% on memory and AI chip prices, and refiners’ from 7.9% to 14.5% on war-time fuel margins. Both last only as long as the shortage behind them. New memory capacity or a ceasefire would take them back. Micron sees it otherwise for memory: it expects supply to be “much tighter” in fiscal 2027 and 2028 than this year, and has signed multi-year take-or-pay contracts with floor and ceiling prices. If those hold, the windfall lasts longer than a normal cycle; if customers’ AI spending slows, the floors are tested.
  • Costs that could not be passed on. Margins fell for homebuilders (20.1% to 17.4%), carmakers (12.5% to 10.2%) and pipeline companies (47.8% to 44.4%), industries paying for fuel, incentives or materials their customers would not cover.

Durable pricing power, the kind that survives a normal year, is rarer than the index margin implies. Most of the improvement this year depends on shortages that the spending in the first section is designed to end.

Working capital: two different signals

Receivables, the money customers owe, reached 12.4% of a year’s revenue for companies outside financials, the highest share in ten years. Inventory, at 7.9% of revenue, is no higher than usual. The two say different things.

Technology accounts for most of both. Its inventory rose 57% against revenue growth of 31%, as companies secured memory and components ahead of demand; in a shortage, that is a sign of confidence, not of goods left unsold. Its receivables rose 43%, as companies extended credit to the cloud companies buying their equipment. That is the one worth watching: revenue booked but not yet collected, from customers whose own cash flow has turned negative.

Productivity, not hiring

The index has learned to grow without hiring. S&P 500 companies employed 29.7 million people at the end of their last fiscal year, 0.9% more than a year before, while their revenue grew 7.1%. Since 2023 their headcount has barely moved and revenue per employee has risen about 6% a year; before 2020 their workforce grew by 1% to 4% a year. Across the economy, payrolls grew 0.4% in the year to September, matching last year as the slowest outside 2020 in ten years, while unemployment stayed low at 4.2%.

Growing without hiring S&P 500 companies by fiscal year, and US payrolls in the year to September −8%−2%+3%+8%+14%'16'17'18'19'20'21'22'23'24'25Revenue per head +6%Headcount +1%US payrolls +0%

The cuts are concentrated in a few large employers: Meta has 14.9% fewer staff than a year ago, Centene 11.5%, C.H. Robinson 9.7% and Intel 8.3%, while Alphabet has 7.7% more and Micron 15% more. Only nine companies filed a formal restructuring notice in the season, against eight in the summer and eight a year ago, so the reductions are coming through attrition, not announced closures.

Pay has stopped keeping up. Average hourly earnings rose 3.2% in the year to September while consumer prices rose 3.7%, the first time since 2022 that wages have lost ground to prices. That is the link between the quarter’s profits and the household’s thinner savings: companies are earning more from each worker, each worker’s pay buys a little less, and spending is being kept up by saving less.

The household: spending out of savings

Part one showed households still spending, but choosing: more travel and card payments, fewer new homes and cars, cheaper cigarettes. The household’s finances explain why.

Household measure A year ago Now Ten-year range
Real disposable income, growth +1.6% +1.5%
Saving rate 5.3% 4.5% 2.7% (2022) to 8.4% before the pandemic
Debt payments, share of income 11.2% 11.1% 9.7% to 11.9%
Credit-card delinquency 3.0% 2.9% 1.5% (2021) to 3.2% (2024)
Unemployment 4.4% 4.2% 3.4% to 4.7%, outside 2020–2021

Federal Reserve and Bureau of Economic Analysis, latest available quarter or month.

Incomes after inflation grew more slowly than spending, and households made up the difference by saving less: the saving rate is the lowest since 2022. They are not in distress. Debt payments take the same share of income as a decade ago, card delinquencies have eased from their 2024 peak, and unemployment is low. Households are choosing because prices rose faster than their incomes, not because they are overstretched. That makes the next leg of consumer spending depend on income growth, because the cushion of savings is already thinner.

Not everyone sees strain. Bank of America, which sees more American household spending than almost any company, said “the consumer is so strong”. McDonald’s described a “challenging consumer environment” but traffic improving in September and October. Both readings fit the figures: spending is holding up, paid for partly by saving less.

In order of size, then, the quarter’s growth came from a few industries expanding exceptionally fast, from higher prices, from productivity, and last and least from an economy producing more. The growth is real, but it is the growth of a few industries and of higher prices, not of a broad expansion, and the household paying those prices is doing it out of savings.

What management expects

The question for this section is how far to believe management. On this year, quite far: their forecasts rose on revenue as well as earnings, and they beat them. On next year, less: the questions analysts are asking have moved faster than the answers executives are giving.

Guidance: raised, and beaten

Between the summer and the autumn, 70 of 106 companies that gave comparable full-year EPS guidance raised it, 24 held it and 12 cut it; on revenue, 36 of 57 raised and 3 cut. Raising guidance as the year goes on is normal, so the cuts say more, and health care made four of the twelve, more than any other sector: Bristol Myers Squibb, Boston Scientific, Eli Lilly and Stryker, the sector where government funding and drug pricing changed during the year.

The raises were earned. Of 29 companies that gave a third-quarter revenue range in July and August and have filed since, 22 came in above the top of it, 6 within it and only Howmet below it. Revenue guidance rose as well as EPS guidance, so the optimism rests on demand, not only on cost cutting.

Against their own range Companies By how much, against the middle of the range
Well above Vertiv, Arista, Amphenol, Palantir, West Pharmaceutical +15%, +11%, +9%, +7%, +6%
Above 17 more, among them AppLovin, Expedia, Datadog, Meta, AMD and Airbnb +0.2% to +5.5%
Within Amazon, Intel, Fortinet, Equinix, Cognizant, ON Semiconductor
Below Howmet −1.7%

The largest beats came from AI hardware and software, the same narrow base as the quarter’s growth, and Amazon, the largest AI spender, landed inside its range rather than above it. One season cannot say which management teams habitually promise low and deliver high: our record of company guidance begins in April 2026. This table is the first entry of a track record that the next issues will extend, company by company.

What executives stress, and what analysts challenge

We read 419 earnings calls from the summer and 431 from the autumn, and compared what executives talked about with what analysts asked. The interesting findings are the gaps between the two.

Subject How often executives raised it, change Calls where analysts asked, summer → autumn
Interest rates +151% 5% → 14%
AI agents +22% 7% → 15%
Backlog +15% 13% → 14%
Pricing +8% 47% → 50%
Consumers +3% 17% → 20%
Demand −1% 40% → 43%
AI in general −10% 36% → 39%
Iran and the Middle East −19% 8% → 7%
Memory −23% 8% → 6%
Tariffs −50% 14% → 11%

Our transcripts begin in April 2026, so these changes cannot yet be checked against a year ago.

  • Tariffs left the script, prices did not. Executives raised tariffs half as often once the refunds settled this year’s bill, but pricing was what analysts asked about most, on half of all calls and two-thirds with consumer staples and materials companies. The cost is resolved; whether customers will keep paying it is not.
  • AI talk fell, AI questions got sharper. Executives spent less of their time on AI, while analysts asked about it more often and about agents, the products meant to earn money from it, twice as often. The question has moved from what AI can do to what it will earn.
  • Rates arrived as a question before an answer. Analysts asked about interest rates on 14% of calls, up from 5%, and on 38% of real estate calls, up from 8%: landlords live by refinancing.
  • Confidence held, with more caveats. Executives still described record results and strong demand as often as before, but they hedged more: analysts asked about obstacles on a third of calls, up from a quarter.

The widest gaps between executive confidence and analyst doubt are in the industries that borrow to build, real estate and utilities, and in AI, where the question has moved to returns. Those are the same places the first two sections found the most exposed financing and the narrowest growth.

The tone of the calls

Executives sound upbeat on almost every call, and each industry has its own way of talking: an insurer discusses losses in a good quarter. So we compared each company with itself. For the 382 companies with a call in both seasons, we asked whether the executives sounded warmer, the same or cooler in the autumn than in the summer.

Sector Calls Warmer Same Cooler Net, warmer less cooler
Energy 17 47% 29% 24% +23
Healthcare 46 48% 20% 33% +15
Technology 60 28% 38% 33% −5
Industrials 59 27% 36% 37% −10
Utilities 27 26% 33% 41% −15
Basic Materials 19 26% 26% 47% −21
Financial 56 27% 16% 57% −30
Real Estate 26 23% 23% 54% −31
Consumer Defensive 18 17% 33% 50% −33
Consumer Cyclical 38 13% 39% 47% −34
Communication Services 16 12% 38% 50% −38
S&P 500 382 28% 30% 42% −14

Executives’ words only, from a finance word list (Loughran–McDonald). A call counts as warmer or cooler when its share of positive words moved by more than five points. Sectors with few calls, such as energy and communication services, move on a handful of companies.

The autumn calls were cooler than the summer ones in nine sectors of eleven, at the same time as guidance went up. The cooling was sharpest where the consumer and borrowing meet: the homebuilders PulteGroup and D.R. Horton had two of the largest drops in the index, regional banks four in six cooler calls, real estate more than half. Energy warmed on the war’s prices. Health care, the sector that cut the most guidance, warmed too, led by diagnostics and research, where six of ten calls were warmer: a sector that had already said its bad news.

Most of the cooling was a loss of conviction, not a run of bad news. Executives described their businesses in the same terms as in the summer, with less certainty about where things were heading. The new worry was the cost of borrowing. Interest rates came up on 147 of the 382 calls, against 110 in the summer, and the talk had turned from rates that might come down to rates that had gone up. That worry ran straight into housing and the shopper. PulteGroup said rising rates had “forced more consumers to the sidelines”. Weyerhaeuser described prices sliding “as housing activity softened”. Equifax cut its mortgage revenue by $60 million from what it expected in July. Ralph Lauren saw the strain in its outlet stores, the ones serving the shoppers most stretched by prices. The job market cooled alongside: Willis Towers Watson said the labour market “has continued to soften”, and Equifax reported slower hiring.

Other themes warmed. Tariffs came up on 100 calls, against 128 in the summer, and the Middle East on 152, against 174: a worry that has been lived with is mentioned less. A quiet hurricane season took the weather off the list of complaints. Banks and insurers talked about credit more often and more calmly. Artificial intelligence stayed the most confident subject of all, raised on 300 calls and nearly always as good news. Fuel and energy costs went the other way, raised on 79 calls against 69.

How investors are responding

Investors drew a line through the quarter’s profits: they paid less for the earnings that come with heavy spending and more for those that come with cash.

Valuation in its history

Sector P/E, Q3 ’25 Q2 ’26 Q3 ’26 Now Earnings, 12 months, y/y
Technology 40.4 34.2 32.5 29.9 +63%
Healthcare 25.2 32.7 32.4 34.9 −3%
Consumer Cyclical 35.5 28.7 26.8 28.5 +28%
Consumer Defensive 29.1 28.6 26.8 29.5 +14%
Industrials 32.0 31.8 27.5 27.7 +32%
Utilities 21.7 21.2 16.9 18.4 +21%
Communication Services 24.2 17.9 18.2 15.6 +45%
Financial 17.8 14.7 14.7 15.7 +16%
Energy 17.1 14.6 13.0 14.0 +77%
Index 28.6 25.3 24.3 24.0 +35%

Price to earnings at the end of each quarter; now is 4 November 2026.

Twelve-month earnings rose 35% while the index’s value rose 15%, so the price paid for each dollar of earnings fell from 29 times to 24. That is lower than a year ago but not low. On the ten-year record below, the index is still above every year before 2020, and below only the peaks of 2020 and 2024–25. All of the return this year came from earnings; none from investors paying more for them.

Price to earnings, ten years Q3 each year, and now (4 November 2026); companies with positive earnings 1018263442'17'18'19'20'21'22'23'24'25'26Technology 29Index 23

Set against its own history, each sector tells a different story.

Sector P/E now Ten-year low Median High Where it sits
Communication Services 15.7 16.7 23.3 28.3 Below its low
Energy 13.7 7.5 15.1 47.0 Below its median
Financial 16.8 12.6 15.1 20.9 Near its median
Utilities 18.2 12.3 18.7 25.0 Near its median
Consumer Cyclical 27.1 25.2 27.8 45.8 At its median
Technology 28.9 21.1 28.3 40.2 Near its median
Industrials 26.5 17.7 20.7 27.9 Near its high
Healthcare 33.0 20.2 24.5 41.0 Well above its median
Consumer Defensive 29.5 19.6 25.2 28.8 Above its high

End of Q3 each year, 2017–2025, against now (4 November 2026); companies with positive earnings.

The sectors priced lowest against their history are there for two different reasons. In energy and communication services, investors are paying less because they doubt the earnings will last: energy’s depend on a war, and Alphabet’s and Meta’s come with spending that has turned their free cash flow negative. Health care and consumer staples are the opposite: priced above their own history, for earnings that fell 3% in health care and grew 14% in staples, because investors moved into them in October as a refuge. The index as a whole is not low against its history. Prices are low where investors doubt the earnings and high where they are sheltering, and that is investors’ judgement about next year, not this one.

What analysts did after results

Analysts revised their views roughly evenly: 61 companies were upgraded and 67 downgraded; 148 had price targets raised and 149 cut. That balance is the finding. Earnings rose 35% and the share of companies beating estimates was the highest in five seasons, yet analysts raised targets no more often than they cut them, and technology had more downgrades than upgrades, 34 against 22. Analysts, like investors, did not carry the quarter’s earnings through to higher prices. Energy and health care had the most one-sided raises.

Where the big investors and insiders moved

Two different records describe what investors did, and they should not be read as one decision. Large managers’ 13F filings show what they held at the end of June, a quarter behind the results and before the autumn sell-off. Insiders’ trades are reported within days and cover July to September. One is a snapshot of portfolios, the other a flow of individual sales, much of it planned in advance.

At the end of June institutions were adding: they opened 23,745 new positions in technology, up from 15,998, a rise of 48% against 19% at the same point a year earlier, and 17,127 in industrials. They were leaving energy just before its best quarter in years. The share of the index held by institutions rose from 83.6% to 86.0%.

New institutional positions, by sector Positions opened by 13F filers, by quarter; the third quarter's filings are due on 14 November Q2 '25Q1 '26Q2 '26Technology15,27215,99823,745Industrials10,67211,23217,127Financial10,2939,00910,662Healthcare7,8298,2689,296Consumer Cyclical6,8966,5047,176Consumer Defensive4,3445,1094,704Energy2,4364,9473,408Utilities3,9154,4934,067

Between July and September executives and directors sold $12.7 billion of their companies’ shares and bought $1.9 billion, $6.1 billion of the sales in technology. Insiders sell more than they buy in most quarters, for pay and diversification, so the sales are a weak signal on their own; what is worth noting is that they were heaviest in the sector where institutions had been adding most.

The longer-term outlook

The central question for the next year is whether this year’s spending, borrowing and new capacity earn enough to justify them. Three kinds of evidence bear on it: what governments are changing, what each industry needs to go right, and how this year’s forecasts are faring.

Policy that changes the economics

Companies’ filings name the government decisions that changed their businesses. Counting the companies that added text naming each this season:

Event Companies Sectors citing it most
Supreme Court ruling on IEEPA tariffs (20 Feb 2026) 107 Industrials, consumer discretionary, technology
The July 2025 budget law 66 Technology, industrials, health care
The global minimum tax (Pillar Two) 53 Technology, health care, industrials
Government shutdown 23 Industrials, technology
Grid rules for large power users (PJM, ERCOT, states) 21 Utilities
Federal Reserve rate decisions 19 Financials
China’s rare-earth export controls 19 Technology, industrials
Section 232 tariffs 17 Technology, health care
End of the de minimis exemption 11 Industrials
US export controls and the Affiliates Rule 10 Technology
Medicare drug price negotiation 7 Health care
Stablecoin law 7 Financials, technology
Drug-pricing orders 5 Health care

Four of these change the economics rather than the paperwork.

  • Tariffs moved from broad to targeted. The IEEPA refunds were a one-time gain; the Section 232 tariffs replacing them fall on particular industries, medicines and semiconductors first, and favour companies that build in the US.
  • Government is paying less for health cover. The budget law’s Medicaid changes and the end of enhanced exchange subsidies are shrinking insurers’ government business; health care made four of the twelve guidance cuts, more than any other sector.
  • Who pays for the grid. The new rules for large power users decide whether data centres or households fund the utilities’ building, and so how much of the second kind of borrowing above is eventually repaid by AI.
  • Export controls cap the China market for the chipmakers whose margins lead the index. Nvidia sold $2.4 billion of licensed Hopper chips to China in the quarter, has stopped shipping them, and assumes no data-centre revenue from China in its outlook. Its board set up a national security and export control oversight committee the day it reported.

Legal costs were large but scattered: Meta’s $9.9 billion settlement and 3M’s agreement to pay up to $7.35 billion over nine years were the biggest of 184 companies with a lawsuit, settlement, fine or order in the season.

The bull and bear case, industry by industry

For each industry that carries the argument, the case for, the case against, and the figure that would show first which is right.

Industry Bull case Bear case What would change the view
Cloud and internet Backlogs of $500–680 billion turn into revenue faster than spending grows Capex keeps rising with no cloud growth to match, or OpenAI’s or Anthropic’s funding falters Free cash flow at Amazon, Alphabet and Meta
Semiconductors and memory AI demand keeps supply short into 2027 New capacity catches up and margins return to normal Memory prices; customers’ inventory
Utilities Data centres pay for the grid they need Rates rise while regulators push costs onto households Rate-case outcomes; borrowing costs
Energy Supply stays constrained by the war A ceasefire or reserve releases bring prices back Refining margins; spare capacity
Consumer Real incomes recover and carry spending Savings run out before prices settle Holiday volumes; the saving rate; delinquencies
Financials Lending and capital markets stay strong Credit losses rise as rates climb Delinquencies and loan-loss provisions
Health insurers Repricing restores margins after 2025 Government membership keeps falling faster than prices rise Medicaid and exchange enrolment

The predictions scorecard

Forecasts we will follow from issue to issue: the companies’ own on AI spending and demand, and the most quoted from outside.

Prediction Who, when Deadline Status so far
AWS revenue will grow more than 40% in 2027 and 2028, on capex of about $220 billion this year and meaningfully more in 2027 Amazon, 29 October End of 2028 Unresolved: the test of whether spending turns into cloud revenue
Capital spending of $195–205 billion this year, rising significantly in 2027 Alphabet, 28 October End of 2027 Unresolved
Azure revenue growth of 49–50% in the December quarter, with capex above $50 billion in it Microsoft, 28 October January 2027 Unresolved
Fourth-quarter revenue of $73–76 billion, on 2026 capex of $135–145 billion Meta, 21 October January 2027 Unresolved
An AI bubble could burst soon Michael Burry, late September Open Unresolved: chip and AI stocks fell 10% to 48% in October and early November, though Nvidia and Micron beat estimates
There is no AI bubble BlackRock’s Larry Fink, Blackstone’s Jon Gray, mid-October Open Unresolved: Meta raised its spending plan from $130 billion to $135 billion
Customers see returns, so AI spending concerns are overstated Amazon’s AWS chief, October Open Unresolved: AWS backlog of $578 billion, but Amazon’s free cash flow went negative
Even a quick end to the Iran conflict would not bring oil down quickly Chevron’s Mike Wirth, late September Open Correct so far: WTI above $90 in mid-October
The Fed will raise rates again in October and December Bank of America, October December 2026 Unresolved: market odds of an October rise were about 20% in mid-October
One more rate rise in December, not a long tightening cycle JPMorgan, October December 2026 Unresolved

Our reading

What is genuinely strong. Demand for AI hardware and electricity is real, contracted and growing faster than a year ago. Companies are profitable, beat their own forecasts, and carry less debt against their cash than at any point in ten years. Households are employed and not overstretched.

What is weaker than the headline. Outside technology and energy, revenue grew about 8%, and most of the growth above the real economy’s 2% was price. Margins outside technology and energy did not improve. Households are keeping up by saving less. The quarter’s 35% profit growth describes a handful of industries, not the economy.

What is becoming riskier. The financing of the build-out. Investment now takes a larger share of cash than at any time in ten years, the payout to shareholders is shrinking, and a small group of borrowers, utilities and the most aggressive builders, depends on lenders while rates rise. Their suppliers’ growth depends on the same few customers, who depend in turn on two private companies.

What could change the picture in the next two or three quarters. Whether the cloud companies’ revenue starts to cover their spending; whether memory and fuel prices hold, since much of the year’s margin depends on them; and whether incomes grow fast enough to carry consumer spending once savings stop falling. Investors have already started to price the risk: they pay less for earnings that come with heavy spending and more for those that come with cash. The next two or three quarters will show whether the capacity being built can turn into cash before the financing, the customers or the household savings that support it run short.

What to watch next quarter

Six developments that could change this issue’s conclusions, each with what would be good news and what would be bad.

What to watch Why it matters Good sign Bad sign
The AI builders’ free cash flow It decides whether the build-out pays for itself or keeps needing debt Cloud revenue grows faster than capex and free cash flow turns up Deficits widen and borrowing grows again
The backlog’s customers OpenAI and Anthropic stand behind much of the index’s contracted growth New funding for both at stable terms A delayed round, or contracts cut back
Memory prices They set chipmakers’ record margins and hardware makers’ falling ones Prices hold while new capacity is absorbed Prices fall as capacity arrives, and chip margins with them
Holiday-quarter volumes The household is spending out of savings Volumes hold while the saving rate stops falling Volumes fall or delinquencies rise
Oil and the end of the windfalls Energy’s profits and the tariff refunds flatter this year’s comparison Prices settle without a shock to supply A sharp fall that takes energy’s earnings with it, or a spike that hits everyone else
Interest rates Utilities and property companies are the borrowers building on credit Rates stop rising and borrowing costs ease Further rises that squeeze the borrowers’ building plans

Governance in brief

The season brought an unusual number of planned successions at very large companies: John Ternus at Apple, Greg Abel at Berkshire Hathaway, Henrique Braun at Coca-Cola, Andy O’Brien at ConocoPhillips. One was not planned: Nvidia reported its quarter on 17 November with Colette Kress, its finance chief, as interim chief executive, a search for a new one under way, an investigation by an independent committee of its board, and its quarterly 10-Q not yet filed. In all, 270 companies filed an executive or board change, against 210 a year ago, and six changed auditors, against one. For the 121 companies whose 2025 proxy pay is tagged so far, the median chief executive’s pay barely moved, from $18.9 million to $19.2 million. The awards that moved the total are share awards whose realised value follows the stock: Palantir’s chief executive’s came to $11.1 billion in the proxies’ own measure. Pay is aligned with shareholders in the narrow sense that it rises and falls with the share price, which in a year like this one rewards the market’s view of AI more than the business’s own results.


Appendix: method and data

Sources. Financial statements, segment figures and the operating measures companies report, as filed with the SEC; analysts’ consensus and reported adjusted EPS; market values; changes between successive filings; earnings-call transcripts; 13F and insider filings; and news coverage of the companies. Index membership is as of 30 June and 30 September 2026.

Coverage. By 18 November, 448 of the 500 companies had filed their report for the quarter and nearly all had announced results; Nvidia’s figures are its quarter to 25 October, from its results release of 17 November, because its 10-Q is delayed; a company with no results for the quarter is left out of this issue’s figures. Six members are missing from the sector figures because we have no record for them under their ticker: AvalonBay, Electronic Arts, Equity Residential, Kenvue, Paramount Skydance and Warner Bros. Discovery. Alphabet, Fox and News Corp are counted once. Sector names are our own classification, not GICS; the membership changes are GICS.

Definitions.

  • Latest quarter is each company’s most recent quarter ending 25 June – 5 October 2026, against the quarter before it and the same quarter a year earlier. A season is the quarters ending in a calendar window (Q3 2026: 6 July – 5 October).
  • Free cash flow is operating cash flow less capital spending. For banks and insurers it is not meaningful and is left out of sector comments; so is gross margin.
  • A beat or miss is more than 0.5% from the consensus, measured on the adjusted figure the consensus was made for. Results more than 50% from the consensus are left out as not comparable.
  • Subjects in filings are counted when a paragraph was added, or a changed paragraph newly names the subject, in a filing dated 15 July – 18 November.

Growth through the year. Each company’s revenue is set against its own fiscal quarter a year earlier, which takes the seasons out, and summed by sector over the 494 companies with nine quarters of results. Profit is too lumpy to read this way, because one-off charges and gains move a quarter’s net income by more than any season. Where we compare this season with the summer (guidance, the language of earnings calls, filings of each kind) we check it against the same season a year ago wherever the data goes back that far, and say so where it does not.

Warning signs. Working-capital flags compare the growth of receivables or inventory with the growth of revenue over the year. Dilution compares the share count with a year earlier. Material weaknesses, auditors’ language and covenant language are read from the filing’s text; goodwill write-downs from the accounts. The covenant flag is a text match and catches standard language as well as real breaches.

Comparability. When a company spins off or sells a business, its past results are restated without it. Comparisons with a year ago use the restated figures, which can differ a lot from what was first reported:

Company Q2 2025 revenue, first reported → now Why
DuPont $3.26bn → $1.75bn Spun off Qnity Electronics on 1 November 2025; moved to industrials
Becton Dickinson $5.51bn → $4.73bn Combined its biosciences and diagnostics business with Waters on 9 February 2026
Occidental down 18% Sold OxyChem, closed 2 January 2026
S&P Global down 11% Separated Mobility
International Paper down 9% Sold its global cellulose fibres business on 23 January 2026
Aptiv down 39% Separated a business in the second quarter

Not investment advice.