DK Street Journal

PG&E plans to spend $11.4bn on its system next year, down from the $13.4bn it had previously laid out, a deferral the company disclosed after California lawmakers ended their regular session without voting on a deal to change how utilities bear wildfire costs. For a company paid a return set by regulators on the capital it puts into service, cutting that capital is a choice to earn less later. PG&E also stopped giving capital plans beyond 2027.

Utility equities this year have been bought on load growth: data centers queueing for power, capital plans rising to serve them, earnings compounding off a bigger asset base. PG&E, the $35.2bn holding company that generates, transmits and distributes electricity and gas across northern and central California, is the counter-case on its own shelf. Its binding constraint is the cost and availability of credit, and it is willingly surrendering future rate base to protect it.

The two variables in Sacramento that set the profit

Chief financial officer Carolyn Burke told S&P Global Ratings' webinar series for chief financial officers on 5 October that the company's credit quality remains tied to the size and durability of California's $21bn wildfire fund, created under Assembly Bill 1054, and that while PG&E's preferred outcome is further state legislation, it cannot wait indefinitely. PG&E has named the two changes it wants: a durable source of liquidity outside the present fund, and a true maximum cap on disallowed costs that does not depend on how much of the fund is left.

That second point is the whole argument. S&P has said PG&E and Edison International could fall to single-B speculative grade if the state does nothing and the fund keeps shrinking, and cites a California Earthquake Authority estimate that the Eaton fire alone could leave roughly $11bn of the $21bn. The strategic review PG&E launched in response may reach capital allocation, business structure or separating the operations least exposed to wildfire.

Every earnings meter improved in the quarter the shares fell hardest

June-quarter operating income rose 15.2% to $1.263bn on revenue that was flat, net income rose 38.6% to $761m, and gross margin reached 84.5%. Consensus has 2026 earnings at $1.65 a share, up 10.1%, and 2027 at $1.80. The shares fell 23.6% over the 90 sessions to 9 October; they trade at 7.94x forward earnings against 9.51x trailing, and at 0.85x book value, where Dominion sits at 1.87x and Entergy at 2.56x.

Some of that is the sector's cost of capital. Twenty-six of 29 utility and power names tested fell over the same 90 days, with the utilities exchange-traded fund XLU down about 6% in the month, its steepest monthly drop in nearly two years, as the 10-year Treasury reached 5.365% on 7 October, its highest since April 2002. PG&E's decline was roughly twice Entergy's 9.8%, and the gap is Sacramento: the shares fell about 20% when the liability deal collapsed in the legislature. Those lows held until the last week of September, after which PG&E rose from $12.34 on 25 September to $13.13 by 9 October.

So the operating results account for none of the drawdown. Earnings grew, margins widened, guidance points up, and the equity still trades under the book value of the assets producing those earnings, because the market is pricing the liability tail on the balance sheet rather than the income it throws off. A sub-book price is a weak anchor in that situation: book value only bounds the loss if the exposure is bounded, and PG&E's own stated ask is a cap that does not yet exist. The capital cut is the honest read on what management believes about its financing cost, since a regulated operator forgoes earning assets only when the alternative is worse.

Lawmakers were expected to take up a bill in special session that would let insurers pursue utilities for wildfire losses. The variable that sets this company's profit sits in a chamber where it does not get a vote.

Two kinds of money arrive at an aircraft owner, and only one of them weakened this summer. AerCap, the Dublin company that buys, leases and sells commercial aircraft and engines to airlines worldwide, collects a fixed monthly rent struck at signing as a percentage of the asset's value. Willis Lease Finance, the Florida lessor of spare engines, collects that too, and also bills maintenance reserves that move with engine hours and cycles. The second meter is the one that broke.

The aftermarket has been traded as a single slowing business since mid-September, when Melius Research cut five aviation names to Hold, warning that several years of exceptionally strong commercial aftermarket growth could moderate to high-single-digit percentage growth next year. The June quarters locate the damage precisely, in the hour-based dollar, while contractual rents widened. That leaves the three owners of the metal on different sides of one slowdown.

The rent does not know whether the aeroplane flew

AerCap booked $1,677m of basic lease rents in the quarter plus $177m of maintenance rents and other receipts. Against flight equipment carried at $57.886bn, the rent line alone implies a book yield near 11.6% annualized, by our calculation. Revenue rose 10.1%. On the call, management put lease yield up about 0.3 of a percentage point year over year and net spread up about 0.5, after several flat quarters.

The second earnings engine is disclosed in full. Net gains on asset sales were $223m on 38 owned assets, a 20% unlevered margin, and 2026 guidance of about $16.80 of adjusted earnings per share splits into roughly $14 excluding gains plus $2.80 banked in the first half: about 17% of the year. Book value per share reached $119.21, up 16%, and the board added a $1bn repurchase authorization in September after $783m of buybacks in the third quarter. At 1.23x trailing book and 7.7x forward earnings, above the trailing 7.0x, the market is marking down the sale-gain slab.

A signed lease pays the same in a weak August as in a strong one.

The meter that fell

Willis Lease's short-term maintenance reserve revenue fell to $39.0m from $50.2m, a 22% drop, while its fixed rent line rose 6.7% to $77.1m and gains on equipment sales rose 16.2%. Management told investors on 4 August that the decline "appears to be the result of customers flying fewer hours on less fuel-efficient platforms such as the A320ceo and 737NGs, powered by CFM56 and V2500 engines." IATA put August traffic down 0.8% in revenue passenger kilometers, with jet fuel at $157 a barrel and load factor at 85.1%, nearly a point lower. Spare parts sales fell 30.2%; utilization slipped to 85% from 87.2%.

The shares took that harder than the accounts warrant: down 28% over thirty days, de-rated to 1.12x book from roughly 2.0x three months ago, under AerCap's 1.23x. Two cautions attach. A three-for-one forward split took effect in July, so unadjusted price histories overstate the fall, and the $0.133 dividend declared was the ninth consecutive payment at the split-adjusted rate, maintained. And 19.6% of the float was short as of 15 September, in a company with a market value of $864m.

FTAI is spending the margin on share

FTAI Aviation has moved from owning engines to selling the modules inside them, and its revenue grew 40.9% in the quarter while gross profit rose 26.0% and operating income 16.2%. Aerospace Products delivered $249.7m of adjusted EBITDA at a 29% margin, flat on the quarter in what management called "a continued reflection of the decision to prioritize market share and large customer penetration." Chief executive Joseph Adams said "the CFM56 market is supply-constrained, not demand-constrained," and that module output is increasingly directed toward third-party customers; 2026 Aviation Leasing EBITDA guidance came down $100m to $475m as a consequence. Green-time CFM56-5B cores were appraised at $2.6m to $3.4m with very limited availability, so the feedstock is dearer, too. At 29.0x forward earnings and 18.4x trailing gross profit, four times AerCap's 4.59x, FTAI carries the group's richest price on the group's thinnest margin.

So the label that groups these three as one weakening business describes one of them. AerCap rose over thirty days with its spread wider and its book compounding, because its income is contractual rent plus sales of assets into a market where its passenger-aircraft extension rate hit 85%. FTAI fell because the parts margin is being traded for volume while consensus still assumes both. Willis Lease fell hardest because the hour-based line is the purest claim on narrowbody flying there is, and that is what airlines cut when fuel costs $157 a barrel. AerCap's own squeeze is still in front of the accounts: on 23 September it priced $1.5bn of notes at 5.500% and 5.875% against an average cost of debt of 4.2%.

Rents signed years ago cannot be reopened to pay for money that costs more today, which is the one way a soft traffic year eventually reaches the fixed half of a lessor's income. The next reading on whether the hour meter has turned back up comes with third-quarter results on 28 October.

CoreWeave rents NVIDIA accelerators by the hour to companies training and serving AI models, and the price of that capacity has been going up. Short-dated contracts of roughly three to six months signed since the end of June priced at about $40m per megawatt of annualized revenue, the company said on 17 September. The money that bought the chips sitting inside those megawatts repriced faster.

Over the thirty sessions to 9 October the data-center group divided along one line: who pays for capacity out of operating cash flow and who pays for it with a lender's. Microsoft gained 8.8%, Meta 10.3%, Alphabet 6.7% and Amazon 4.5%. Every member that borrows fell, Cipher Mining by 20.9%, CoreWeave 14.8%, Oracle 12.2%, Hut 8 10.9% and Nebius 9.2%. The slide was read as the AI demand story cracking, helped by an 8 October report that OpenAI's annualized revenue was closer to $50bn than the roughly $70bn circulated to investors. The disclosures support something narrower and more durable: with the 10-year Treasury yield touching 5.36% on 7 October, its highest since 2002, the cost of building a rented-accelerator business moved, and it only moves the companies that finance the hardware.

CoreWeave's top line follows the power switch

CoreWeave defines its revenue backlog as remaining performance obligations plus other committed amounts "subject to the satisfaction of delivery and availability of service requirements." That makes the top line a function of how fast a shell gets power, not of how many customers want compute. The book reached $104.2bn at 30 June, up 246% in a year. The sequential path is where a slowing reservation would appear first: backlog ran to $99.4bn at 31 March on more than $40bn of new commitments, nearly 50% in one quarter, and then added 4.8%.

Delivery, meanwhile, caught up. Active power hit 1.5 gigawatts at 30 June after nearly 500MW was added in the quarter, against 3.7GW contracted, so 40.5% of contracted power was energized against roughly 29% three months earlier.

"Demand for CoreWeave cloud remains exceptionally strong across the entirety of the customer base," chief executive Mike Intrator told investors on 11 August. Revenue cannot corroborate that, and the counterparty list is short: the largest customer was 36% of June-quarter revenue and the second 26%, against a largest customer at 45% in the March quarter. Microsoft was the largest customer for fiscal 2025, and Meta signed an order form in March 2026 with an initial commitment of about $21bn — two buyers that can build the same halls themselves.

Two schedules management does not set

Gross margin has fallen 8.3 percentage points in four quarters, from 74.2% to 65.9%. The reason sits below it. Net interest expense was $640m in the June quarter, 24.9% of the quarter's $2,575m of revenue and more than double the $267m a year earlier, and the September quarter was guided to $860m-$940m. Total debt was $35.0bn at 30 June. New facilities price well wide of investment grade: a $3.1bn delayed-draw term loan closed in May at 450 basis points over the floating benchmark, and a proposed $2.6bn facility rated BB+ by Fitch was talked at 425 to 450 over, an 8.53% to 8.79% yield to maturity, with maturities clustering between 2026 and 2028. Free cash flow runs at minus 30.5% of market value.

The second schedule is depreciation. CoreWeave writes its accelerator-bearing equipment off straight-line over about six years, raised from five at the start of 2023; Nebius uses four years on the same class of hardware. Oracle's September quarter is the one public test of which assumption is right: on its 10 September call it reported 97.9% utilization and said renewed capacity was priced 20% above the prior contracts, on hardware four years old.

Nebius sells the same hour and funds it with stock

Nebius, the Amsterdam-based cloud that rents GPU clusters and developer tooling, reported $582.3m of June-quarter revenue, up 454%, with 98% of it from the AI cloud and an annualized run rate of $3.0bn at the end of June against $1.9bn in March. It has never posted an operating profit; the June quarter lost $175.9m at that line. Its one profitable quarter this year, $621.2m of net income in March, contained a $780.6m revaluation of its ClickHouse holding.

The build is funded from three places outside compute sales. Capital spending was $5.66bn in the June quarter against $20bn-$25bn guided for 2026, covered by an at-the-market program that issued shares for about $2.8bn at a weighted average $224, customer prepayments running at 50% to 60% of the associated capex, and a $775m asset-backed facility in July, its first secured debt. Its landmark 2026 deals yield $20m to $25m per megawatt, below CoreWeave's recent $40m, and chief executive Arkady Volozh said on 12 August: "We could sell today our entire 2027 capacity on these terms if we wanted to." The shares carry 15.88x forward sales against CoreWeave's 3.48x, with no forward earnings multiple available because consensus has the company loss-making at the operating line through 2028.

Hut 8 lets the tenant buy the accelerators

Hut 8, the energy-infrastructure platform that hosts data-center capacity and still mines bitcoin, has contracted 949MW with about $26.6bn of expected base-term value and more than $1.75bn of expected average annual net operating income. Divide one by the other and the landlord earns roughly $1.84m per megawatt a year, about a twenty-second of CoreWeave's $40m, and the difference is almost entirely the accelerator capex CoreWeave finances. Its River Bend campus carries a 15-year, $7.0bn lease on 245MW to Fluidstack, backstopped by Google and funded by a $3.25bn project bond — $1.90m per megawatt a year by the same arithmetic.

Lenders price that difference explicitly. Hut 8 closed a $1.07bn four-year secured revolver on 24 September with JPMorgan Chase as lead arranger at an initial margin of 175 basis points, under 40% of the spread on CoreWeave's May term loan. The cost of the model is time: the first River Bend data halls are targeted for the June quarter of 2027, which is why Hut 8's trailing and forward price-to-sales are identical at 32.61x, the dearest of the three, with consensus 2026 revenue only 21% above 2025. Its June-quarter net loss of $177.1m was, in the company's words, "driven primarily by $138 million loss in digital assets."

What actually got repriced

The results explain the thirty-day decline on the cost side only. Every disclosed demand meter improved: pricing per megawatt rose, the energized share of contracted power went up, and top-customer concentration came down. What deteriorated is the price and the quantity of debt, and the split follows the collateral. CoreWeave and Nebius buy depreciating silicon with borrowed or newly issued money, so a 5% handle on the long bond reaches them directly. Hut 8 signs triple-net leases with an investment-grade backstop and lets its tenant own the chips, which is why a bank syndicate funds it at bank spreads. Microsoft, Meta, Alphabet and Amazon spend from operating cash flow and cannot be de-rated by the same lender. Cipher Mining, the worst performer in the group over three months at minus 41.9%, is a separate case: its revenue fell 43% year on year as mining economics deteriorated.

CoreWeave's own multiple has done the arithmetic already, halving to 5.90x trailing sales from 11.76x in early May while revenue doubled.

The gap between a signed reservation and an energized megawatt is the only place a cancellation can surface before it reaches earnings, and it closes or widens one quarter at a time. Nebius reports next on 10 November, with CoreWeave historically a day ahead of it.

Borr Drilling, which owns a fleet of modern jack-up rigs and rents them with crew to oil companies in shallow water, added three more contract commitments on 6 October and watched its chairman spend roughly $17.5m buying its own shares through late September and early October. Both arrived while the price of new work sat below what the working fleet already earns.

That gap is the only honest read on this market, because an offshore driller is paid a dayrate multiplied by contracted operating days, recognised against contracts signed one to three years earlier. Revenue can hold long after rates break, and it can fall while the market tightens. The four weeks since Borr's shares were last at $4.12 have been read as the jackup cycle turning. What the filings show is a company deliberately buying occupancy with price.

The de-rating is in the rate, not the days

Borr's 21 contract commitments so far in 2026 — about 4,350 days and $541m of dayrate-equivalent value — were struck at an average of roughly $123,000 a day, against the $134,000 average on the 73% of 2026 rig-days already sold. New work is being signed some 8% under the rolling fleet rate.

The days themselves are piling up. Total backlog reached $1.13bn, with eight commitments and more than 2,100 firm days added since the prior report, then three more in October: a West African operator's binding letter of award for the Norve, roughly 245 days from mid-December; a one-well option exercised by Shell Nigeria on the Natt; and about 60 days of UK accommodation work for the Joro. Rigs worked 98.4% of scheduled time. Idleness is not the problem.

Price and cost are. Operating margin went from 36.0% in the June quarter of 2025 to 0.13% a year later, on revenue of $232.3m, down 13.2% and a third straight sequential decline. Adjusted EBITDA fell $44.7m sequentially to $43.8m, with $22.5m of that quarter's expense spent preparing the Odin for a US contract before it earned anything. Consensus now has 2026 EBITDA at $173.6m, a 60% fall and a further cut from the $183.4m carried a month ago, before a snap-back to $422.5m in 2027.

The company has said which of the two it is chasing: 2027 is when it plans to "turn focus again very sharply into economics and rates".

The bid came from inside

Borr rose 11.7% across the sessions of 8 and 9 October to $4.78. Noble gained 1.5% over the same two days, Valaris 1.9% and Seadrill 1.4%. Immediately before it, Drew Holdings, the vehicle of chairman Tor Olav Trøim, bought 1.5m shares on 5 October at an average $4.1314 and 1.25m more on 7 October at $4.3122, part of about 4.15m shares acquired since mid-September.

The demand case behind the buying is not imaginary. ADES Holding said on 11 August that all of its temporarily suspended Saudi offshore rigs had received resumption notices, with the suspensions event-driven rather than demand-driven; contracted jackup utilisation was holding near 90% in a market described as tight, with marketed supply roughly flat at about 405 rigs this year. Capacity came back, but earlier Aramco recalls were repriced at prevailing rates, which is consistent with Borr's $123,000.

Noble is a deepwater company on a shallow-water shelf

Noble's June-quarter revenue of $719.7m fell 15.2%, and the cause was floaters: contract drilling revenue dropped to $679m from $743m on the suspension of two rigs in Brazil, about $43m, and the end of the Globetrotter I's Black Sea work. Its 24 marketed floaters were 61% contracted, down from 68%. Its five ultra-harsh North Sea jackups went to 80% from 66%.

Its earnings sit in a $6.8bn backlog roughly equal to its $6.76bn market value, with four Guyana drillships extended to February 2029 under ExxonMobil's enabling agreement and a $0.50 quarterly dividend. Diluted shares rose from 152.9m to 161.0m since late 2024, so shrinking share count explains nothing about a stock down 7.4% in thirty days. At 9.00x trailing EV/EBITDA, Noble is the cheapest of the four offshore names on cash earnings and the only one with a positive free-cash-flow yield, at 4.2%.

What the quarter can and cannot explain

Almost none of Borr's October gain comes from reported results, which deteriorated again, or from forecasts, which were cut again. It comes from the chairman's purchases and from a Middle East market putting rigs back to work. The business evidence splits cleanly: days tightening, dollars per day loosening. Borr trades at 10.24x trailing EV/EBITDA and 1.53x book, the second-highest book multiple of the four and the only negative earnings among them; the market is paying marginally more per forward sales dollar than per trailing one, because consensus revenue falls 1%.

The clock on trading rate for utilisation is the balance sheet. "The new notes amortized at 5% per annum, equating to $101.75 million on a full year basis," chief financial officer Magnus Vaaler told investors on 12 August, with amortisation starting July 2027 on debt refinanced at 8.75% and 9.00%.

Chief executive Bruno Morand's case, made the same day, was that Middle East demand "was close to materializing prior to the onset of the conflict, and we believe this delayed activity should reenter the market once conditions stabilize." Conditions did stabilise, and the rigs came back. The question the 17 November print answers is what they came back at.

Norfolk Southern collected more revenue in the June quarter than in any quarter of its history, and almost none of the increase came from hauling more valuable freight. The Atlanta railroad, which moves chemicals, metals, autos, coal and intermodal containers over roughly 19,300 route miles in 22 states, reported revenue of $3,465m, up 11%. Operating income rose 1.8%. Net income fell 4.4%.

The record top line has been read as a franchise strengthening into its handover to Union Pacific. A Class I railroad, though, is not paid for freight demand in any form its revenue line makes legible: it collects a rate struck commodity by commodity, plus a diesel surcharge recognized inside revenue, so the honest meters are volume mix, revenue per unit excluding fuel, the operating ratio and service. All four of Norfolk Southern's went the wrong way.

The volume that grew was the cheap kind

The hierarchy is in the quarter's 10-Q. Merchandise earned $2,133m on 610.7 thousand units, about $3,493 a unit. Intermodal earned $908m on 1.06m units, roughly $857. Coal came in between, near $2,264 a car. A merchandise car is worth about four containers.

The containers are what arrived. Domestic intermodal units grew 11% while international fell 3%, and utility coal tonnage fell 8% as export tonnage surged 25%. Revenue excluding fuel surcharge reached a record $3,050m, up 5%, but revenue per unit excluding fuel rose just 1% across all three segments, against reported per-unit gains of 6% in merchandise and 16% in intermodal.

That gap is the surcharge, and it looks backwards. Class I formulas compute the charge off a diesel price from a period up to 60 days earlier, so fuel expense and fuel recovery can diverge sharply in a volatile market. Diesel set a nominal record of $6.529 a gallon in the week of 21 September against a year-to-date average of $5.040, which means the recovery landing in autumn revenue is money already spent.

Slower trains inside a record top line

Norfolk Southern's operating ratio was 67.6%, or 65.5% adjusted for $51m of merger-related expense, $15m of Eastern Ohio incident costs and $6m of restructuring. The reported figure worsened by 2.1 points year on year, of which roughly 1.1 points came from fuel price and a further 1.9 from inflation.

Service deteriorated with it: train speed fell to 19.9 mph from 21.6, terminal dwell rose to 24.0 hours from 22.7, and merchandise plan compliance dropped to 68% from 78%. Train length, train weight and fuel efficiency set records, which is what a long train run late looks like in a report.

"Norfolk Southern delivered a strong second quarter, exceeding our expectations as demand improved across key markets," chief executive Mark George said on 23 July. On the call management said network velocity was regaining momentum in the third quarter.

Union Pacific ran the same mix and held 59.7%

The larger railroad, 32,452 route miles linking Pacific and Gulf ports to Midwestern gateways, saw the identical shift: domestic intermodal up 19%, international down 14%, coal carloads down 17%. Its fuel-surcharge revenue rose to $1.0bn from $569m, so about $431m of the quarter's $735m revenue increase was diesel pass-through. Freight revenue excluding fuel grew 4%. The operating ratio was 59.7%, eight points better than Norfolk Southern's.

The next price increase gets filed, not negotiated

What Union Pacific is actually buying is single-line revenue, and whether it keeps it is decided in a docket. The Surface Transportation Board's 18 August schedule set 18 November as the deadline for protests and requests for conditions, with responses due 16 February 2027 and hearings to be scheduled later.

BNSF, CPKC and CSX have all said they will seek broad access to the combined network, CSX specifically trackage rights over Norfolk Southern's Lurgan Branch to Harrisburg, the Reading Line to Bethlehem and the Lehigh Line into New Jersey. Union Pacific has said it would walk away from the $85bn purchase if the Board orders widespread trackage rights or line sales. CSX, the Eastern railroad left outside, ran a 61.7% operating ratio last quarter on 10% revenue growth, the best of the three, and trades at 27.3x trailing against 23.6x forward earnings.

Old Dominion sets its own tariff

The truck side of the same shelf answers to nobody's docket. Old Dominion, a less-than-truckload carrier running a national service-center network out of Thomasville, North Carolina, saw tons per day slip 0.9% in August on a 2.4% fall in shipments, and raised prices anyway: revenue per hundredweight up 11.3% as reported but 4.8% excluding fuel, followed by a 4.9% general rate increase effective 5 October. "Underlying demand trends remaining relatively consistent as the quarter has progressed," chief executive Marty Freeman said on 3 September. Second-quarter operating income rose 30.0%, and the shares still fell 19.4% over three months.

Which de-rating is deserved

The results account for Norfolk Southern's weakness and not for Old Dominion's. Norfolk Southern is 5.3% off its 11 August high, yet its trailing price-to-earnings ratio of 27.1x is above the 25.9x of 31 May, because the earnings fell faster than the price; at $317.79 the shares sit roughly 13.5% below the $367.16 of Union Pacific stock and cash the deal would pay, a discount derived from the two closing prices. Old Dominion's compression, to 35.0x trailing from 43.3x in early May, is earnings catching up to a price that had run ahead of them.

The cause of the split is who sets the rate. Old Dominion files its own tariff and its ceiling is its own doors, which is why 30% earnings growth and a one-fifth share decline can coexist. Norfolk Southern's rate is negotiated car by car, its surcharge is a lagging echo of a diesel bill already paid, and the several years of single-line revenue behind its premium are an open question on a government calendar.

Third-quarter numbers from both railroads land on 22 October, the first covering the record-diesel quarter. Six weeks before anyone files what they want from the merged map.

Green Brick Partners, which entitles and develops most of the lots it builds on in Dallas-Fort Worth, metropolitan Atlanta and Florida's Treasure Coast, told investors its homebuilding gross margin reached 29.8% in the June quarter, the highest among public homebuilding peers by its own reckoning and more than eleven percentage points above the 18.4% peer average. Net new orders rose 19% to 1,079 homes, and the shares fell 10.3% in the month to 9 October anyway.

The shelf of regional builders has been traded as one position all year, priced off the cost of buying a buyer's monthly payment down from a 30-year fixed rate Freddie Mac put at 7.40% on 8 October. The gross margin lines refuse to cooperate with that reading. What separates these companies is not how hard they discount but when they bought the ground.

The spread was earned years before the rate moved

Green Brick owned 76% of roughly 52,000 lot positions at the end of June and uses no land bankers at all. Its annual gross margin has run 29.75% in 2021, 30.84% in 2023, 33.52% in 2024 and 31.16% in 2025 — a band that barely notices the rate cycle, because the profit was set at wholesale land prices years earlier. The company employs 620 people against Meritage's 1,860.

"One of the primary differentiators from many of our peers is that we do not engage in high interest cost land banking relationships that can distort a builder's economic leverage and risk and that can give a land banker indirect control over a builder's lot purchase timing," co-founder and chief executive James Brickman said on the second-quarter call. Chief financial officer Jeff Cox attributed the roughly 90 basis points of sequential improvement to the entry-level Trophy Signature Homes brand taking a larger share of sales.

What a rented lot costs at closing

Dream Finders Homes, which builds for first-time and move-up buyers from Jacksonville to Denver and originates mortgages for sale to PrimeLending, is the structure Brickman described. Its homebuilding gross margin fell to 14.2% from 16.5%, which the company attributed primarily to higher land and financing costs, while average sales price per home closed dropped 8.9% to $438,171.

Demand was not the problem: net new orders rose 15% to a second-quarter record 2,232. The company's own filing says operating margins "depend heavily on the need for continued mortgage buydowns via forward commitments", and its controlled lot pipeline shrank to 54,091 from 63,121 at the end of 2025 as options went unexercised.

Meritage Homes, the spec builder of the group, sits between the two: home closing gross margin of 18.3% against 21.1%, orders down 9% as a 19% slower absorption pace outran a 14% larger community count, cancellations at 13% from 10%, and full-year volume guided about 5% below 2025 with a warning that revenue could fall further "if market conditions require higher incentives."

One group, two verdicts

The selling was indiscriminate but the businesses were not. Dream Finders closed at $9.38 on 9 October, down 29.1% in a month and the lowest in the price record available; consensus has its 2026 earnings halving to $1.07 a share and falling again in 2027, its forward price-to-earnings of 8.8x sits above a trailing 5.9x, and it is closing an all-cash purchase of Beazer Homes at about $2.2bn. That de-rating is the results, not a mispricing. D.R. Horton fell 3.4% over the same month and Lennar 5.8%.

Green Brick is the one the numbers do not explain. At $62.86 it trades on 9.4x trailing earnings and 1.37x book, against 0.54x book at Dream Finders, and sits 21% above its own two-and-a-half-year low while its margin lead widened. Nothing company-specific landed in the window.

Both companies report third-quarter results after the close on 28 October, the land developer and the spec builder on the same day. One of them will have to say what it paid to sell a house; the other will say what it was paid for the land.

Element Solutions agreed in July to sell itself to Solstice Advanced Materials, and its own shareholders talked the board out of the deal. The two sides mutually terminated the merger agreement on 27 August, neither owing the other a payment. The shares closed at $36.32 on 9 October, 26% below the $49.09 they reached on 22 June, before any of it began.

A specialty chemicals maker whose gross margin has fallen in each of the last five quarters, and which fell out of its uptrend in September, looks like a business losing its pricing. Its own quarterly filing describes something else. Of the 56% revenue growth Element reported for the June quarter, 15 points were organic; $107m was pass-through metal and $129m came from acquisitions. Metal bought on a customer's behalf and resold at close to no mark-up lands in revenue and in cost of sales at the same size, so it mechanically shrinks the margin percentage without touching a dollar of profit.

The premium its holders read as a markdown

Solstice Advanced Materials agreed on 6 July to pay $10.00 in cash plus 0.500 of its own shares for each Element share, implied consideration of roughly $50.10 and a premium of about 15% to the prior close, valuing the business at about $14.5bn including net debt. Element holders would have owned some 44% of the combined company.

They did not want it. The shares went from $43.64 on 2 July to $42.32 on 6 July, on 18.5m shares against roughly a sixth of that the session before, and to $38.57 on 7 July: down 11.6% in three sessions on a bid struck at a premium. Seven weeks later the agreement was gone. "At Element Solutions, we work for our shareholders and other stakeholders and have heard their feedback clearly," chief executive Benjamin Gliklich said on 27 August. "The termination of the proposed transaction is a direct response to that feedback." The next session took another 4.7% off the stock on 20.7m shares, against a normal two to four million.

The revenue line that quotes the metals market

Element sells solder and assembly chemistries, circuit-board process chemicals, die attach and wafer bumping into electronics manufacturing, and palladium and its fellow precious metals are inputs it buys and bills on. Palladium futures rose 81.5% in 2025 and traded above $1,900 an ounce in late January. That is why consensus has revenue up 49% this year while earnings per share rise about 22%, and why management warns of greater quarterly earnings variance from metal price swings alongside full-year adjusted EBITDA guidance of $690m to $710m.

Measured where the metal cannot reach, the business accelerated. Electronics organic sales grew 20% with every vertical up double digits, led by Semiconductor Solutions at 31% on thermal interface materials for artificial-intelligence graphics and central processors and on advanced packaging at Asian assembly houses; adjusted EBITDA margin excluding pass-through metals rose 120 basis points to 27.8%. The shares now change hands at 6.16x forward gross profit, against 9.43x in May.

Weakness in the shelf next door is a different thing

Element sits on a list of fab-materials suppliers whose bearish complexion is real but not shared. Entegris, which supplies filters, chemistries and wafer carriers consumed per wafer start, rose 18% in thirty days; its two segments each grew and both are guided above 10% for the September quarter, and at 16.7x trailing gross profit it has more than erased a de-rating that had it at 14.5x in August. Linde, the largest industrial-gas group, and MKS, which sells vacuum and power subsystems to chipmakers, have both rolled into downtrends — MKS down 28% in three months after a 118% twelve-month gain. Those are capital-equipment and base-volume stories.

Element's is neither. The operating numbers reported since the shares peaked are better than the ones before, and nothing in the filings marks the June quarter as the turn. What changed in July was ownership, not demand: a buyer offered equity, the holders said no, and the price has stayed where the refusal left it. The open question is the one the metals create — a quarter in which palladium falls will shrink reported revenue and lift the margin percentage, and neither will mean anything about the chemistry.

No money changed hands when the agreement ended in August. What ended with it was the outside price a buyer had put on these assets, struck before the quarter in which Element's semiconductor chemistry grew fastest, and withdrawn by the shareholders who owned it.

The fees an ultra-low-cost carrier collects for a checked bag, a seat assignment and a bundle are close to a fixed sum per passenger. Frontier's came to $68.42 a head in the June quarter, unchanged from the year before, while the fuel it burned cost $4.17 a gallon against $2.36, according to the carrier's second-quarter results. Everything the fuel bill demanded had to come out of the fare.

The low-fare group's summer selloff has been read as travel demand cracking. The filings point at refining: the International Air Transport Association expects industry net profit to halve to $23bn in 2026 from $45bn as jet fuel rises nearly 70% to $152 a barrel, with the jet crack spread over crude averaging a historic $57. Because the per-head fee take will not flex, what is left for each carrier is the question of who else is flying its routes.

Frontier inherited a price, not a cost

Spirit Airlines, the original American ultra-low-cost carrier, ceased all operations on 2 May 2026, and Frontier overlapped it on roughly 35% of seats. Fare revenue per passenger then rose 54% year over year, lifting total revenue per passenger to $131 and quarterly revenue 38% to a record $1.3bn on unit revenue up 28%. Chief executive Jimmy Dempsey called it a "structural change" on the 29 July call, estimating that Spirit's removal from overlapping markets provided more than five points of unit revenue.

The cost side is intact and improving. Stage-adjusted cost per available seat mile excluding fuel was 7.42 cents, up 2%, aircraft utilisation reached 9.7 block hours a day against a management target of 11 to 11.5, and aircraft grounded by Pratt & Whitney geared-turbofan inspections were down 25% year to date on maintenance output up over 40%. Liquidity was $1.16bn against net debt of $506m, and Frontier guided third-quarter capacity up 17-18% with unit revenue up just over 20%.

It still lost money: a 7.6% operating margin deficit in the June quarter after a 28.5% deficit in March, and a $90m net loss. The equity trades at 0.28x forward sales against Southwest's 0.62x, with no earnings multiple at all — consensus carries a loss of $0.50 a share this year and $0.36 of profit in 2027, and FY2025 gross profit was $14m on $3.72bn of revenue. Competitive capacity on Frontier's routes is scheduled down more than four points in the third quarter. Frontier is pricing into a gap a dead rival left.

Southwest's unit revenue is partly a credit-card number

Southwest spent two years dismantling the model that made it cheap: bag fees on tickets bought from 28 May 2025, basic economy, and assigned and extra-legroom seating from 27 January 2026, which ended 54 years of open seating. Adjusted unit revenue rose 20.1% in the June quarter. Its 10-Q explains that passenger revenue "includes the co-brand impact" of a 2025 amendment with Chase that moved "a larger portion of co-brand Chase credit card benefits" into that line, and that other revenue rose $104m in the half "driven primarily by improved retail spend on the Company's co-branded credit cards." Loyalty revenue, the cash received for points sold, grew 2.4% to $557m in the March quarter, a quarter in which total revenue grew 12.8%.

"The revenue contribution from bag fees has exceeded our expectations so far, and we've experienced no negative impact to the operation," chief executive Bob Jordan said in January. Fuel swamped it: $3.92 a gallon, an estimated $889m year-over-year increase, larger than the quarter's $465m adjusted net income. Full-year adjusted earnings were reset to $3.25-$4.25 a share from at least $4.00 and planned capacity growth roughly halved to 1-1.5%. The shares carry 14.5x forward earnings against 25.1x trailing, a compression that exists only because consensus assumes $2.88 a share this year against $0.79 delivered last year.

The fifth of Ryanair's fuel it did not hedge set the result

Ryanair hedged 80% of this fiscal year's jet fuel to March 2027 at about $67 a barrel; the unhedged remainder pushed total costs up 11% to €3.81bn in the June quarter. Traffic grew 6% to 61.3 million passengers and ancillary revenue held at roughly €24 a head, the same as the prior full year, but the average fare fell to €48 from €51 and profit dropped 34% to €538m. With jet fuel then trading near $140 a barrel, it trimmed winter capacity and cut full-year traffic guidance to 214m passengers from 216m. "This is a full-blown crisis," Michael O'Leary told the Financial Times in September. "Everybody is going to be hurting next year."

Ryanair is the only one of the three with earnings to capitalise, and its forward multiple of 16.2x sits above 12.7x trailing: the price is falling and the estimates are falling faster, from $4.10 a share last year to $3.36 expected. It is also the only one with no debt, having repaid a €1.2bn bond in May. Some of the dollar-quoted decline is translation — the euro went from $1.1808 on 16 April to $1.1200 on 8 October.

What the three prints explain

All five listed low-fare carriers fell between 15% and 35% in the three months to 9 October, and the last thirty days were flat, which is why none appears in a short-term mover list. Four rolled out of their uptrends between late August and late September. Frontier, down 23.9% over three months, has held a mild uptrend since 11 August.

The results account for Ryanair cleanly and Southwest hardly at all. Ryanair's de-rating matches an earnings decline it guided toward itself, plus roughly five points of currency. Southwest delivered four straight quarters of accelerating revenue and a wider operating margin and fell anyway; the reconciling item is its own guidance cut. Frontier's shares followed its margin line rather than its 37.7% revenue line, and its fare power is an inheritance from a liquidated competitor.

Boeing has confirmed Ryanair's first fifteen MAX-10s for spring 2027, aircraft that burn a fifth less fuel. They arrive after the winter it has already cut.

SpaceX agreed on 9 October to buy a nationwide portfolio of 800 MHz licences from the investment firm Grain Management, and the market promptly paid three landlords for a customer that has signed nothing. Crown Castle, which leases vertical space on more than 40,000 US cellular towers to wireless carriers, closed the session up 16.8% on 14.17m shares against roughly 3m on typical early-October days. American Tower rose 10.4% and SBA Communications 9.1%.

The same session took T-Mobile US down 12.3%, AT&T down 9.7% and Verizon down 9.2%. The physics explains the split. The 800 MHz band is low-band coverage that works through walls and ceilings, which satellites cannot do, so it is hard to use from orbit and easy to use from a tower. One headline therefore read as a prospective fourth tenant for the steel and a fourth national competitor for the carriers. The belief being bought is that a new network means new leases on existing towers.

The buildout SpaceX has actually described

The deal was valued at about $8bn by the Wall Street Journal, covers up to 14 MHz of paired licences, and needs approval from the Federal Communications Commission; the parties did not disclose terms. Grain bought the portfolio from T-Mobile in August 2026, so the spectrum arming a possible fourth carrier was sold by the carrier that fell hardest on the news.

SpaceX's own stated architecture is the problem with the rally's logic. The company has described housing small cellular radios alongside its existing Starlink broadband dishes on a limited number of host rooftops rather than building new macro towers. President Gwynne Shotwell has told investors the EchoStar spectrum "does have terrestrial components, so we definitely intend to build out the terrestrial component." Intent to transmit from the ground is not intent to rent a lattice. As Wireless Estimator put it the same day, "investors are betting on a new tower tenant that hasn't signed a lease yet."

The rent already signed is the rent in bankruptcy court

A tower dollar is earned under long-dated master lease agreements whose escalators run near 3% a year whatever traffic does, and it is lost through contractual churn. Crown Castle's second-quarter organic contribution to site rental billings was $38m, growth of 3.9%, but only "excluding an unfavorable $49 million and $5 million impact from DISH Terminations and Sprint Cancellations, respectively". The escalator delivered its 3-4%; one defaulted tenant took back more than it gave. American Tower shows the same gap: 1.9% reported organic tenant billings growth, about 4% before DISH.

That loss is a forward event, not a post-mortem. DISH declared its leases excused after the FCC forced parent EchoStar to sell spectrum to AT&T and SpaceX, and the three towercos are pressing roughly $7.6bn of disputed claims, Crown Castle alone about $3.5bn, in a case whose plan confirmation now slips to no earlier than 2 December. The Brattle Group put DISH at 5.1% to 7.0% of annual rental revenues across all three. The regulator that armed the hypothetical tenant is the defence for cancelling the real one's rent.

Three landlords, three different books

Crown Castle's reported revenue fell 4.9% year on year in the second quarter to $1.008bn, mostly the fiber and small-cell businesses sold in May. Its AFFO of $1.13 a share rose 10%, but the company said the increase came "primarily as a result of a decrease in interest expense and an increase in interest income" from the proceeds of that sale. It raised full-year AFFO guidance by $5m to $4.59 a share at the midpoint, inside a $4.53-$4.65 range, and ended the quarter at 6.3x net leverage. The cost side is a ground lease under every site: about 57% of its ground is leased, and the buyout programme plus efficiencies is expected to cut $15m of annual cost.

American Tower grew revenue 4.7% to $2.749bn and has a second meter inside the ticker. CoreSite, its US data-centre arm, had its full-year revenue growth guidance lifted to roughly 15% from 13%, against roughly 4% for the global tower book; European escalators are uncapped and indexed to inflation rather than fixed. SBA Communications is the shrinking domestic control: site-leasing revenue there fell 3.7% to $452.5m on Sprint and EchoStar consolidation, international rose 30.5%, and AFFO per share fell 3.8% to $3.03. Barclays downgraded American Tower and Crown Castle to Equal Weight on collection risk and kept SBA at Overweight for having less of it.

What the single session actually repriced

The business record does not account for Crown Castle's move. Its shares closed 9 October at $79.64 against $79.66 ninety days earlier and $96.62 a year before, and have traded below their moving-average trend every day since 17 August, including the day of the jump. After the pop it carries the highest trailing EV/EBITDA of the three at 21.0x, against American Tower's 18.3x and SBA's 17.0x, on the weakest organic growth and the most leverage. American Tower's +8.1% over three months has a decelerating-but-growing book and an accelerating data-centre segment behind it; its 26.1x forward earnings sits below the 28.7x on the Desk's May record. SBA is the cheapest of the three and the only one whose domestic revenue is outright contracting. A 5.24% ten-year Treasury yield is the discount rate under all of them, but it cannot lift one name 16.8% while its customers fall by a tenth.

At the Goldman Sachs Communacopia conference on 9 September, chief executive Chris Hillabrant said "the larger opportunity depends on how quickly carriers move on new spectrum." On 9 October the market decided a company that owns no towers and runs no network had answered him.

Affirm Holdings, the point-of-sale instalment lender that lets shoppers in the US, UK and Canada spread a purchase over terms from one month to forty-eight, took in more money last fiscal year from interest on loans sitting on its own balance sheet than from the merchants who pay it to offer interest-free credit. Interest income was $2.05bn of $4.26bn of fiscal-2026 revenue; merchant network revenue, up 30%, was $1.15bn.

That matters because buy-now-pay-later shares are traded as a read on the shopper's health, and revenue is treated as the read. It cannot be. Four unlike dollars land on one transaction, and only one of them is priced off the shopper's credit at all.

Four unlike dollars, one purchase

The merchant discount rate is charged on interest-free volume, where the retailer and not the buyer pays for the credit: Pay-in-X was 16% of fiscal-2026 volume and 0% APR instalment loans another 14%. Consumer interest income comes off the longer interest-bearing loans Affirm keeps, which is why that line is roughly half of revenue. A third dollar, gain on sales of loans, was $596.6m — a credit decision recognised up front as a capital-markets spread when the loan is sold into a forward-flow agreement. The fourth is card interchange, $294.0m, from the Affirm Card, which is attached to 19% of active customers. "We need to make sure that the Affirm Card isn't just the best card in the world," founder and chief executive Max Levchin told investors on 27 August. "It's also the best way to use Affirm."

Stack those and revenue can grow in a quarter when the economics decay.

Where the credit charge actually lands

The honest meter is revenue less transaction costs measured against volume, because funding cost, the provision for credit losses and the subsidised share of volume all land there. It was $589m in the June quarter, 4.2% of gross merchandise volume, and $812m or 5.8% before the credit provision — the same disclosure read with and without the credit charge. Across the full year the measure rose to $2.08bn on $50.2bn of volume, about 4.14%, against roughly 4.03% a year earlier: a widening unit margin on volume up 37%.

Affirm has to borrow before it can lend, and the borrowing got cheaper. Aggregate funding debt of $9.8bn splits into $5.3bn of securitisation notes, $3.3bn of warehouse facilities and about $1.1bn of convertibles, and the average annualised funding cost in the June quarter fell 103 basis points to 5.8%. Thirty-day delinquency on monthly instalment loans improved to 2.5% from the 2.7%-2.8% range of the prior three quarters, with an allowance of $563m against $9.56bn of gross loans held for investment. Management guided fiscal-2027 revenue less transaction costs to 4.16% of volume, above its own prior 3.25%-4.0% range, crediting favourable debt capital markets. "We are in control of the credit reality," Levchin said on the same call. "You would see us slow down growth before you would see us have a real credit disturbance."

A re-rating with no filing behind it

None of that is what moved the shares. Affirm closed at $77.76 the session after the report, slid 12% to $68.15 by 9 September, then rose 18% to $80.70 by 9 October. Price to gross profit — the fitting measure, since a roughly $1.5bn deferred-tax release leaves a trailing price-to-earnings ratio of 14.1x against 40.5x forward — went from 8.4x trailing in mid-September to 9.5x on 9 October, 7.2x forward, entirely on price. The likelier reading is the cost of money: implied odds of an October Fed hike collapsed from near 69% to 16.5% in the first week of the month after New York Fed president John Williams cooled the talk, and a lender with $9.8bn of funding debt prices the policy path straight into its cost of goods sold. A DBS Bank upgrade on 1 October helped; TD Cowen cut its target to $110 from $124 on 6 October while keeping a buy rating.

The bank and the remitter

Klarna Group, the London-headquartered payments provider that also runs a digital retail bank, carries the same interest-free receivable from the opposite side of the balance sheet: about 90% of its book is funded by consumer deposits, roughly $11.7bn to $13bn of them at 30 June, with a capital ratio rather than a counterparty's appetite as the constraint. Its credit line is improving — provisions fell to 0.52% of volume, a third consecutive decline — and its shares are down 66% over twelve months, at 1.3x trailing sales, the lowest since the September 2025 listing and 2.1x trailing gross profit. The cause sits above credit: full-year volume guidance cut to $149bn-$151bn from more than $155bn on German retail weakness and some $600m of adverse currency, consensus growth of 17.9% for 2026 against 31.6% delivered in 2025, and a chief financial officer and chief marketing officer both departing in early 2027.

Remitly Global, the Seattle digital remittance business serving immigrants sending money to nearly 150 countries, is the contrast that defines the group: no loan book, a fraud loss line instead of a credit one, and a take rate of roughly 2.11% of send volume in the June quarter against about 2.23% a year earlier on our arithmetic — a compressing spread on volume up 27%, trading at 16.6x trailing and 17.3x forward earnings.

So two lenders' disclosed loss rates improved while the sector narrative worsened: LendingTree's 2025 survey found 41% of users had made a late payment, up from 34%. The companies' own books and the industry's survey data point opposite ways, and nothing filed closes the gap. What the August numbers establish is that Affirm's spread widened on cheaper funding with losses flat; what they do not establish is the October advance, which arrived on a rate expectation.

Affirm reports after the close on 5 November, covering the first quarter it has funded since the Fed began raising its policy rate again. Whatever that costs shows up in the one line its revenue cannot show.