US Market Recap: Eleven Earnings Reports, Three Blowups, and Every One Broke on Guidance

I. Earnings scorecard: eleven reports, three blowups

Same-day performance for all eleven, measured against the October 28 close; Amazon and Intel also show their after-hours moves.

The three blowups sit in search, machinery and chips, and all three broke on guidance. Alphabet's problem is the money it intends to spend next year, Caterpillar's is a full-year framework it refused to raise, Intel's is a fourth quarter $2 billion short of expectations. Each posted numbers worth reading: Google Cloud grew 103%, Caterpillar's backlog of $78.1 billion is a record, Intel's adjusted EPS of 0.52 topped the 0.41 consensus. The market priced the slope of the next quarter; the current quarter only set how deep the cut would be.

Eli Lilly was the other end of the tape. Revenue of $24.4 billion, up 39%, and full-year revenue guidance lifted from $85–87 billion to $91–92 billion made it the only report of the day that both beat and raised. The stock gained 6.63% and touched an all-time high of 1,408.38 intraday. Merck, Norwegian, Chipotle and DoorDash beat as well, with gains from 3.83% to 13.45%. Microsoft was the counterexample: revenue of $92.3 billion, adjusted EPS of 4.89 and Azure growth of 48% all cleared the bar, yet the stock closed down 0.16%. The beat was cancelled out by capex coming in below plan, which the market read as hyperscalers slowing their orders — a thread that ran straight into the compute lessors (see section VII).

II. The three indexes: the damage sat in one sector

All three indexes fell for the first time after three straight up sessions, but the size of the declines differed fourfold. The Nasdaq fell hardest because its heaviest platform names were the ones reporting that day. The Dow lost only 0.25% because Caterpillar at -9.25% was offset by Amgen, Merck and Apple. The S&P's 0.30% sits between them, and a sector map of eight up and three down says the selling never spread beyond the industry it started in.

One intraday detail is worth keeping: the Dow ranged from 52,405.88 to 52,086.13, an amplitude of just 0.61%, while the Nasdaq's was 0.88%. That is not the volatility of a panic session. It reads like position adjustment — sell one sector, buy the other eight.

III. Seven numbers: only the Russell 2000 closed higher

The Nasdaq 100 and the composite fell by almost exactly the same amount, 1.38% each, and that is more informative than the size of the drop. If the selling had been a large-cap versus small-cap split, the gap between the two would be obvious. Instead the small-cap bid showed up in the Russell 2000 while the Nasdaq 100 tracked the composite, which means specific companies were sold, not an entire market cap tier. The VIX rising only 1.71% to 17.20 says the same thing.

The SOX ended a four-session winning streak with a 1.42% decline, while the Golden Dragon closed up 0.20%. Put those three directions together and the day priced one question: who pays for AI capex. It did not price whether the US economy is healthy.

IV. Eleven sectors: eight up, three down

The best and worst sectors were 9.1 percentage points apart, and communication services supplied 6.60 of them on its own. That is more than seven times the second-worst sector, information technology at -0.88%, an extreme reading by the history of the index: one sector falling more than the other ten combined. Information technology losing only 0.88% also tells you this was not read as a technology problem. The pricing was surgical, confined to one box.

The eight winners say more about the day's logic than the losers did. Health care 2.50%, real estate 2.25%, staples 1.64%, utilities 1.56%, financials 1.52%. What these five share has nothing to do with defense and everything to do with capex: they are businesses whose cash flow is visible today and which do not need to keep writing large cheques. The two sectors sold, communication services and information technology, are precisely the two heaviest spenders.

V. Communication services -6.60%: search growth slips to 12%

Cloud revenue doubled, contracted backlog stands at $561 billion with management saying more than 55% converts within 24 months, and the Gemini app reached a billion users. None of it held the stock up. What the market sold was the advertising line: search and other ads at $63.4 billion, up 12%, ran 2.7 points behind the 14.7% the stock was priced for, a gap of $1.6 billion. Advertising is more than sixty percent of Alphabet's revenue, and it is the entrance to the old route of "how does the AI spending get paid back". Narrow the entrance and the cloud numbers, however good, only push the question out a quarter.

The cash flow arithmetic is harder still. Operating cash flow failed to cover $51.4 billion of capex, leaving free cash flow at -$9.1 billion for a second consecutive quarter. Capex more than doubled year over year, the full-year guide held at $195–205 billion, and management put 2027 above $300 billion. Net income of $26.2 billion was 25% below last year's $35.0 billion, and an $8.2 billion "other" loss included roughly $11 billion of pre-tax mark-to-market damage on the SpaceX holding. That is not an operating problem, but it pushed GAAP EPS to 2.12, a $0.9 gap to the 3.02 consensus, and the market did not separate the two that day.

The stock closed at 283.82, down 12.64%, wiping out $479 billion of market value: the largest one-day decline in Alphabet's history as a public company and the second-largest single-session loss of value in US market history after Nvidia's record. There was no second heavy weight in communication services to offset it. Meta rose 1.96%, but it had already been sold on capex the previous day, and two sessions together recovered only a fraction of that drop.

VI. Where the money went: health care, real estate, staples

Eli Lilly was the day's most straightforward long. Third-quarter revenue of $24.4 billion rose 39%, net income of $6.5 billion rose 16%, and full-year revenue guidance went from $85–87 billion to $91–92 billion, the third raise this year. Demand for the obesity and diabetes franchise is still accelerating and the company is pushing an oral formulation. The stock closed up 6.63% after touching 1,408.38, and the market capitalisation crossed $1.3 trillion, the first drugmaker to reach that size.

Real estate and utilities rising together points at the long end. The 10-year yield fell 7 basis points back below 5%, and high-dividend, long-duration assets collected the benefit directly. Financials gained 1.52% the same day, with JPMorgan +1.26%, Wells Fargo +1.36% and Citigroup +1.33%, which says banks were not bought for higher rates but because money leaving technology weights needed somewhere to sit.

Staples ended a four-day losing streak: Procter & Gamble +1.47%, Coca-Cola +1.19%, PepsiCo +0.87%, Walmart +2.09%, Costco +1.36%. None of them had news. What rose was allocation demand. Moving money from communication services into staples was the least troublesome way to cut exposure on the day.

VII. The AI buildout splits: lessors sold, power equipment bid

The split starts inside Microsoft's report. Cash capex of $39.4 billion more than doubled year over year but came in $6.6 billion below the $45.9 billion analysts expected, while Azure guidance for the current quarter still points up, at 49% to 50%. For Microsoft that is spending less than planned. For the lessors who live off its orders it is the queue moving back: CoreWeave, Nebius, IREN and Applied Digital all fell at least 4.8%, in a neat line from 4.81% to 8.98%.

What rose the same day was the other line: Quanta Services +4.66% for grid construction, Constellation +2.52% for nuclear and retail power, Vertiv +1.23% for power and cooling, GE Vernova +1.12% for turbines and grid gear. The two moves are not in conflict. As hyperscalers build more of their own capacity, procurement shifts from renting compute to buying electricity, transformers and cooling. The money is still being spent, only the destination changed.

Oracle gave back 3.45% and is the cleanest example on the chain. It had risen 10.73% the day before on the payback schedule from its investor day and fell back to 139.56. The same buyers who happily paid a premium for "fiscal 2030 revenue above $225 billion" on Wednesday raised their discount rate on Thursday after seeing Alphabet's negative free cash flow. The story did not change, the numerator did not change, the denominator did.

VIII. Marvell buys Lumentum: the 12% after-hours pop gone in a day

Bloomberg reported after the October 28 close that Marvell is in advanced talks to buy Lumentum, with an announcement possible as soon as this week, no definitive agreement signed and both sides declining to comment. Standard merger pricing moves the target up and the acquirer down. In Thursday's regular session the target's premium was erased and the acquirer's discount became fact.

Marvell closed at 218.34, down 13.24%, one of the worst performers in the S&P 500. Lumentum opened at 947.88, ran to 978.40, a gain of 7.05% over the prior close, then slid all the way back to 919.53 for a gain of 0.61%. The entire 11% to 12% pre-announcement pop was returned. In all-stock terms the two moves are one event: a 13% decline in the buyer's stock shrinks the currency it is paying with by 13%, so target shareholders end up holding discounted paper and the original premium disappears on its own.

The rest of optical fell with it: Coherent -0.99%, Ciena -3.79%, Applied Optoelectronics -3.69%, Arista -1.34%. The chain has been pulled both ways all month, between orders from expanding AI capacity and multiple compression as capex expectations shift. On the day, Lumentum held up because it may be bought and Marvell was sold because it has to pay. Both point at the same conclusion: optical components are the hardest part of this buildout to expand, so whoever owns them is expensive and whoever buys them gets cheaper.

IX. Caterpillar -9.25%: unchanged guidance was the bad news

An EPS beat, revenue $73.5 million light, a record backlog and margins up 3 points, and the stock fell 9.25% to 647.39, its worst session since March 2020. The problem was not the quarter but the framing: keeping the full year at "mid-to-high teens" and calling the fourth quarter merely "modest growth" marks December growth down a notch.

Inside the segments, growth is still there but tiered. Energy and transportation revenue of $8.8 billion rose 18%, with power generation up 30% — the piece AI data centres buy directly. Construction industries at $8.0 billion rose 18%, half the 35% pace of the second quarter. Resource industries at $4.6 billion rose 15% with margins slipping to 13.5%. Global machine retail sales rose 22%, but power generation retail rose 75%, mining and aggregates slowed from 15% to 8%, and rail retail fell 70%. The data-centre and non-data-centre halves are now moving in opposite directions.

The Dow impact is direct. Caterpillar's price-weighted drag was about 64 points, and together with Alphabet the two erased a broad advance led by Amgen, leaving the index down 132.54 points. Without those two, the Dow would have closed higher.

X. Semiconductors: equipment up, the priciest memory down

Rank the SOX's 30 members by return and the shape is obvious: all four equipment names closed higher (KLA +5.45%, Applied Materials +1.36%, ASML +0.77%, Lam +0.22%), the design names were all lower except Qualcomm, and memory fell hardest (SanDisk -5.32%, Micron -4.98%). The index closed at 11,518.20, down 1.42%, giving back part of the prior four sessions while staying well above the -6.72% low of October 22.

Memory deserves its own breakdown. Between October 27 and 28, SK Hynix's quarter (+328% revenue, +632% operating profit, capacity sold through 2029) and Oracle's capex schedule pulled the whole chain for two days. On the 29th Alphabet raised the 2027 number while turning current cash flow negative, and the first thing those buyers cut was whatever had run hardest in those two days. SK Hynix's ADR fell only 1.58%, far less than SanDisk or Micron, which says multiple was sold, not orders.

Equipment held up for a simpler reason: KLA, Applied Materials and ASML book orders against fab capex plans that are locked in annually. What the market questioned on the day was whether cloud companies will keep adding money, not whether orders already signed would be cancelled. That distinction bought the equipment group a day.

XI. After the bell: Amazon's record margin, Intel's guidance breaks

Amazon's issue is not margin. Net sales of $201.9 billion rose 12%, the 13.9% operating margin is the highest in company history, net income was $20.7 billion and AWS grew 39.3%, its fastest in five years. Fourth-quarter guidance of $240–247 billion puts the midpoint at $243.5 billion, in line rather than clearly above, while capex keeps climbing. The stock had already fallen 4.25% in the regular session and lost another 6.9% after the report.

Intel's problem is more direct. Revenue of $16.5 billion rose 21% and matched expectations, with the data centre and AI group at $7.3 billion, up 78%, and foundry at $6.7 billion, up 58%. But the fourth quarter is guided to $15–16 billion, at least $2 billion below the roughly $17 billion consensus, and gross margin guidance drops from 43% to 38%. Management blamed soft PC processor demand and component supply problems in data centre, and said the PC slump lasts at least through the first half of 2027. The stock fell 23% after hours.

Put these two beside Alphabet's and you have three facets of one question: cloud is growing, chip demand is growing, but whoever pays for that growth now wants to see cash come back. Amazon's margin was the best answer available and the market still gave it nothing. Intel could not even offer certainty on demand, so its multiple took the hit directly.

XII. China ADRs: hotels and consumer up, compute down

The Golden Dragon closed at 5,411.30, up 0.20%, one of the few green prints in a US tape that ran from -0.25% to -1.42%. The internals are cleanly split: hotels and consumer higher (Atour +7.82%, H World +3.97%, PDD +2.34%, Vipshop +1.92%, Yum China +1.60%), compute and semiconductors lower (ACM -7.04%, Hesai -6.22%, Kingsoft Cloud -5.56%, GDS -2.93%).

A 0.20% gain on the index is not a China ADR rally; what happened was internal rotation. The half that was sold prices off US AI capex: GDS builds data centres, Kingsoft Cloud sells cloud, ACM ships tools to fabs, and all of them hang on the spending plans of Nvidia and the hyperscalers. The half that was bought tracks domestic consumption and has nothing to do with Alphabet's report or Intel's guidance. The index gained only 0.20% because the two halves cancelled out.

XIII. Macro: 3.4% growth, 3.6% inflation, and the curve moved down

The data was a strong-growth, sticky-inflation combination: third-quarter GDP at an initial 3.4% annualised, above the 3.2% expected and well ahead of 2.2% in the second quarter; September PCE at 3.6% year over year, faster than August's 3.4%; core PCE at 3.1% against 3.0% previously, up 0.3% on the month, all in line. Initial claims came in at 194,000 against 198,000 expected.

On the usual logic that set of numbers pushes yields up. Instead the entire curve moved 5 to 8 basis points lower. Two forces stacked: equities were falling, so risk-off money went into duration, and oil was lower, dragging inflation expectations with it. Reports that Vice President Vance wrote to Iran's leadership to restart talks fuelled expectations of a cooler Middle East, and both crude and bonds reflected it.

The front end not moving is the important part. Three-month yields held at 4.18%, which says a December hike is already fully priced and everything new on the day landed in the belly and long end — the long-run path of growth and inflation, not what the next meeting does. White House economic adviser Hassett said Trump respects the Fed's independence and believes rates could be lower. Remarks like that, made when hike expectations are already locked, rarely move pricing.

Two other items unrelated to the tape: ECB President Lagarde said she will step down on February 7, nearly nine months before her term ends, and Nvidia took a stake in Canadian AI company Cohere. The first is a source of uncertainty for European policy, the second shows a chipmaker reaching into the model layer.

XIV. Cross-asset: oil held, silver did not

Oil barely fell and the spread actually widened. December WTI settled at 102.00, down 0.76%, and December Brent at 114.88, down 0.38%, moving the gap from 12.54 to 12.88. The decline is about talks: Vance's letter let the market price a cooler Middle East. The small size of it is about shipping: Hormuz transits have not normalised and three more tankers hit mines on unauthorised routes in the previous 24 hours. Expectations moved first; the physical market did not follow.

In metals the real move was silver. Gold fell 0.45% and silver 3.34%, more than seven times as much. Silver carries a larger industrial demand share, and "small caps up, silver down" is not a contradiction on the same day: buying small caps means buying cash flow that does not depend on AI capex, while selling silver means selling the marginal industrial demand outlook. Bitcoin at -1.26% moved with the Nasdaq, and the high-duration complex contracted together.

XV. Views

• What broke in Alphabet is the valuation method, not the business
  Cloud up 103%, $561 billion of backlog, a billion Gemini users — all of it was in the report, and the stock was still marked down 12.64%. The market re-denominated growth into how much cash has to be burned to buy it, and advertising growth slipping from 14.7% to 12% is what makes that equation fail. For platform companies from here, free cash flow comes before revenue growth. The order has changed.
• The rotation went to light assets, not to defense
  Health care 2.50%, real estate 2.25%, staples 1.64%, utilities 1.56%, financials 1.52%. What those five share is light capex and visible current cash flow, not small declines. Communication services and information technology, the two heaviest spenders, were sold on the same day. The money did not leave; it moved from where growth has to be prepaid to where growth has already become cash.
• Microsoft's capex miss is the first time in this AI trade that the payer chose to slow down
  $39.4 billion, $6.6 billion below plan. Good for Microsoft's cash flow, and an order deferral for the lessors who depend on it: CoreWeave, Nebius, IREN and Applied Digital all fell at least 4.8%. The thread is not finished — Amazon's fourth-quarter guide also failed to beat, and Intel guided below. Equipment still rose because fab capex is locked annually. That gap in timing is worth watching.
• The index does not need a direction yet
  Dow -0.25%, S&P -0.30%, eight sectors up, the Russell 2000 on a four-day run, the VIX up only 1.71%. What fell was a small group of the most expensive companies. In that structure a 1.38% Nasdaq decline is not the same thing as a falling market. If communication services stops falling next week while health care and real estate keep rising, the rotation is complete and the index repairs itself. The risk is the other case: two large companies guiding below expectations after the bell turns one company's problem into a capex-cycle problem.