DK Street Journal

Waste Management told investors in late July that it would collect less waste this year than it had planned to, and make more money doing it. The company, which hauls household, commercial and industrial refuse to transfer stations and landfills it owns, cut its 2026 revenue outlook to $26.275-26.475bn on a roughly $250m shortfall in collection and disposal work, held adjusted earnings before interest, taxes, depreciation and amortization guidance at $8.15-8.25bn, and raised its margin target to 31.0-31.2%.

That trade — tonnage for price — is the whole equity case for the business, and it has stopped being paid for. Waste Management closed at $207.63 on 25 September against $207.69 exactly two years earlier, while consensus 2026 earnings of $8.11 a share now sit 19% above the $6.81 it reported for 2024. Nothing in the operating record explains a flat two years; a lower multiple, 25.5x forward against 29.2x trailing, does all of the work.

Two meters, pointing apart

A collection dollar is billed per lift and per ton, not per unit of economic activity, so price and volume move independently. In the June quarter core price was 5.7% and collection-and-disposal yield 3.6%, against volume down 1.8% — down 0.4% excluding prior-year wildfire cleanup — with full-year volumes guided toward a decline of about 1% on the same basis. "Volumes were softer than we anticipated," John Morris, the president and incoming chief executive, told investors on the 29 July call, noting residential volumes down 2.9%, partly from continued shedding of lower-margin contracts.

The result was operating margin of 18.7% against 18.5% a year earlier, revenue up 4.0%, and free cash flow of $1.10bn, up 34.5%. Healthcare Solutions, the segment built on the $7.2bn Stericycle purchase, lifted its adjusted EBITDA margin to 19% from 17% and has produced $32m of annualized cross-selling EBITDA; leverage is back to 2.96x, inside the company's target range. The renewable natural gas program is the one growth line management does not price — some $600m of planned 2026 EBITDA billed at gas and credit values that fell this year, prompting executives to defer selling some credits into the fourth quarter.

What the escalators do next

Republic Services, which collects and disposes of non-hazardous solid waste across 41 states, discloses the contract machinery most plainly. Just under 20% of its restricted book is indexed to headline consumer price inflation and about 35% to other indices, with roughly a 12-month lag — so billed increases step down a year after inflation does. It shows up as open-market pricing of 7.8% against restricted pricing of 4.1%. "We're always going to take price over volume," chief executive Jon Vander Ark said on the 6 August call, with residential volumes down 4.3%. Republic raised full-year guidance that day, helped by lifting its recycled-commodity assumption to about $135 a ton. Its shares carry 29.2x forward earnings against 30.0x trailing and a 5.2% free-cash-flow yield: barely a de-rating at all, on 5.6% expected earnings growth.

Casella Waste Systems, the northeastern collector and landfill owner roughly one-sixteenth Waste Management's size, is the opposite case. Revenue rose 16.9% last quarter on five acquisitions adding some $165m of annualized revenue, while operating income fell 22.4% and net leverage sits at 2.7x. It trades on the highest enterprise value to EBITDA of the four listed collection names, 15.9x, at a 2.1% free-cash-flow yield against Waste Management's 4.3%, and BofA Securities cut it to Underperform with a price target of $81 on 23 September.

The verdict

One chart, three situations. Casella's decline is operating deterioration and the market has read it correctly. Republic's multiple has scarcely moved, so calling the whole group de-rated overstates it. Only Waste Management shows margins, cash and leverage all improving into a falling share price, with no downgrade wave behind it — UBS lifted its target to $270. The likelier reading is rate competition: a mid-single-digit free-cash-flow yield behaves like a bond substitute when long-dated Treasury yields hold at levels unseen since 2006. What investors will no longer pay a premium for is the price-over-volume trade itself — they are marking the risk that indexed escalators reset lower while tonnage keeps shrinking, and price stops covering the gap.

The executive who called the volumes soft takes over as chief executive on 4 January, with a single quarterly print between now and then to show that the softness was the shedding and not the economy.

Kinder Morgan's largest new Gulf Coast gas line cleared federal permitting three months ahead of schedule — and the shares spent the month going the other way. The Federal Permitting Improvement Steering Council announced on 8 September that the $1.7bn Mississippi Crossing project, 208 miles of large-diameter pipe built to move up to 2.1 billion cubic feet a day into Alabama with service as early as the second quarter of 2028, had finished its federal review early. Kinder Morgan's 50-day average price slipped below its 200-day on 21 August and has stayed there since; DT Midstream's crossed on 9 September.

That matters because of how these companies are actually paid. Kinder Morgan does not sell gas to data centers; it sells reserved capacity. Its own disclosure puts 96% of 2026 budgeted cash flow in take-or-pay, fee-based or hedged form — 65% take-or-pay, which bills whether a molecule moves or not — on remaining contract terms averaging 3.3 to 14.7 years. A book like that is a bond with pipes attached, and the 30-year Treasury yield topped 5.53% on 24 September, its highest since 2004, from under 5% in early July. The longer the contracted stream, the harder it re-prices.

The business improved while the price fell

Kinder Morgan's second-quarter revenue rose 10.8% to $4.48bn, operating income 17.8% to $1.35bn, and operating margin reached 30.1% against 28.3% a year earlier. Adjusted earnings before interest, taxes, depreciation and amortization hit a record $2.20bn, up 12%, and the company raised full-year guidance. The backlog fell $500m to $9.6bn in the quarter because more than $650m of projects went into service, not because anything was cancelled; the remaining $8.5bn is expected to earn back its cost at roughly 5.6 times first-full-year project cash earnings. Chief executive Kimberly Dang told investors on the 22 July call that additions from a more than $10bn opportunity set should more than offset the roughly $1bn entering service in the second half, with nearly $400m already under contingent board approval. The stock now trades at 12.5 times enterprise value to EBITDA, the cheapest of the three, and 19.7 times trailing earnings against 21.7 times in May.

Where the de-rating is earned

DT Midstream is the purer version of the same idea and the weaker case. It is about 95% demand-based with an average tenor near eight years, and in 2024 roughly 92% of pipeline revenue came from firm service. But revenue growth halved from 27.3% in the fourth quarter of 2025 to 11.0% in the second quarter of 2026, and operating-income growth decelerated from 38% to 11.6%. Its forward earnings multiple of 25.9 sits barely below a trailing 26.6 — consensus expects almost no 2026 expansion — while the revenue growth is back-loaded to 19.1% in 2029 and 30.7% in 2030, precisely the shape a 5.5% long bond punishes. The commercial momentum is real: chief executive David Slater said on the 30 July call that "our entire asset footprint is kind of lit up like a Christmas tree right now, and we've never seen that before while we've owned these assets," and the company took a final investment decision on $300m of projects including a 380 million cubic feet a day interconnect feeding an Ohio data center's power plant. Analysts split rather than converged in September: Jefferies cut its target to $145 while Morgan Stanley upgraded the shares.

Targa Resources, which gathers and processes Permian gas and fractionates natural gas liquids, is the one whose contracts are least protective — about 80% fee-based, with the balance on percent-of-proceeds terms that hand it a share of commodity sales. It has held an unbroken uptrend since 12 December 2025 and rose over the same three months the others fell. Its quarter was genuinely strong — operating income up 52.7%, adjusted EBITDA up 38% with full-year guidance moved to the top of the range — and buybacks do not explain the share price: $80m repurchased in the quarter against a $59.6bn market value. It is also the most expensive of the three at 16.6 times EV/EBITDA, on a trailing free cash flow yield of 1.2% against Kinder Morgan's 5.7%.

What nothing explains

Rates carried the direction. Regulated utilities fell harder over the same three months — Southern down 14.9%, NextEra 14.4%, American Electric Power 14.2% — while the S&P 500 rose. But rates are not the whole answer: Enterprise Products, MPLX and ONEOK, sitting in the identical bond market and the identical gas market, all rose, and Henry Hub gas itself gained 12.4% in thirty days to $3.27 per million British thermal units. Same discount rate, same commodity, opposite outcome.

So the split resolves against the comfortable reading. DT Midstream's de-rating is paid for by its own numbers — decelerating growth and cash flows that arrive at the end of the decade. Kinder Morgan's is not: revenue, margin, guidance and permitting all moved its way, and the decline is a re-pricing of duration rather than of performance. And the defense investors thought they were buying — capacity locked under multi-decade take-or-pay terms — is now the feature being discounted, while the name carrying commodity margin it does not control is the one holding. An optics of safety has inverted.

Kinder Morgan reports third-quarter results on 21 October. The test is not the earnings line but whether the newly signed firm capacity Dang promised actually lands in the backlog — because for now, being paid regardless of flow is worth less, not more.

Three vendors sell into the same artificial-intelligence build-out and bill for it three different ways, and the one whose meter is turning slowest has repriced the hardest.

Okta, which sells cloud identity — single sign-on, multi-factor authentication and lifecycle management — on annual per-user subscriptions, ended its July quarter with 5,255 customers paying more than $100,000 a year in annual contract value, 6% more than a year earlier. Cloudflare, which runs a global edge network sold as security, performance and serverless compute and charges for what crosses it, grew its count of customers above that same $100,000 threshold by 27%, to 4,698. Datadog, whose observability platform bills on telemetry volume consumed, grew its roughly 4,720-strong equivalent base about 23%.

That gap answers the question the whole group is being priced on: whether AI agents get billed as identities, as consumption, or not at all. Roughly $250bn of combined market value rides on it, and right now the market is paying it backwards.

The meters

Okta's revenue grew 10.6% to $805m, a fourth consecutive quarterly deceleration. Its forward meter is better than its reported one — current remaining performance obligation reached $2.585bn, up 14%, running ahead of billed revenue — but management guided that figure back down to 11-12% growth for the current quarter. What growth exists is concentrated: customers above $1m in annual contract value grew 22%, to more than 600, and newer governance, privileged-access and threat-protection products were 30% of bookings at roughly 40% average uplift per deal.

Cloudflare's revenue grew 35.9% to $696.1m, its fourth straight acceleration, with dollar-based net retention up six points year over year to 120%. Datadog's grew 35.6% to $1.121bn, also a fourth straight acceleration, and chief executive Olivier Pomel told investors on the August 13 call that "revenue growth for our non-AI customers also accelerated again this quarter to the high 20s percent year-over-year" — the acceleration is not confined to AI customers.

Does consumption billing scale like software?

Partly. Cloudflare's non-GAAP gross margin fell 320 basis points year over year to 73.1%, and chief financial officer Thomas Seifert attributed it to more network cost landing in cost of revenue as paid traffic grew relative to free traffic — the toll on selling compute at the edge, where non-human traffic passed half the network for the first time. It rose 30 basis points sequentially, the first such improvement in eight quarters. Datadog's slipped to 78.6%. Okta's, the seat-seller's, expanded 2.7 points to 79.6%, and its operating margin more than doubled to 13.3%.

The headcount irony runs the same direction. Cloudflare, the consumption vendor, cut roughly 1,100 jobs — about 20% of staff — on May 7 in an "agentic AI-first" restructuring costing up to $165m; the $150.7m charge is the main reason its GAAP operating loss widened to $205.7m. Okta, the per-seat vendor agents are supposed to destroy, employs 6,366 and shrank its diluted share count 1.2%. Both consumption names diluted about 2%.

What the price paid for

In late July, Okta traded at 10.29x trailing gross profit, Cloudflare at 54.90x and Datadog at 30.44x. Today Okta is 13.50x, Cloudflare 67.92x — and Datadog 30.26x, effectively unchanged. Okta was near 5x as recently as early May, when it was down 30% for the year; it bottomed at $62.93 on April 10. The claim that this corner of software never de-rated is false in both directions: CrowdStrike is down 45.5% over twelve months, ServiceNow 26.2% and Zscaler 24.5%.

Datadog's flat valuation has nameable causes. Snowflake is paying about $1bn for Observe, moving into its market, and chief financial officer David Obstler said on August 13: "Regarding our largest customer, we have seen a usage reduction, which is incorporated in our Q3 and full year 2026 guidance." The resulting third-quarter guide of $1.135bn-$1.145bn implies 28-29% growth. Consumption cuts both ways.

Okta's re-rating rests on a story rather than a number. At Oktane on September 22 it launched an agentic-enterprise reference architecture and confirmed Agent SSO is generally available inside the core product; six brokers lifted targets into a $215-$240 range this week. But finance chief Brett Tighe said on the August 26 call that agent-identity revenue is "still immaterial. Still very small. We're very early innings… for FY '27, we don't think it's going to be material." Consensus models 9.9% revenue growth for the year ending January 2028 — a fade, not an inflection. Cloudflare, meanwhile, costs about 2.4 times what Datadog does for each percentage point of gross-profit growth it delivers.

So the verdict splits. Datadog has earned its advance with accelerating revenue and no help from a re-rating; Cloudflare has earned part of one and is paying for the rest with network cost and dilution; Okta's business — decelerating revenue, 6% growth in its large-account base, margins genuinely inflecting — explains the profitability half of its move and none of the growth premium now attached to it. Billing model is not destiny, and the market has stopped treating it as evidence.

Todd McKinnon told the Oktane audience on September 23 that in agent identity "there is going to be massive growth and the market share is going to be set probably in three years." His own finance chief does not expect material revenue from it this fiscal year. The price has already paid for both.

Packaging Corporation of America, which makes containerboard and converts nearly all of it into shipping boxes, retail cartons and point-of-sale displays, told customers in late July it wanted $140 more for every ton of containerboard from 1 September. That is roughly double a normal move, and the trade press called it unprecedented. Cascades, Smurfit Westrock and International Paper followed within days at $100 and $80 a ton. When Fastmarkets RISI published its September data, it recognized $70 a ton on linerboard and $100 on corrugating medium.

That gap matters more here than in most industrial goods, because containerboard's selling price is not negotiated quietly per customer but published by a trade journal, and box contracts reset off it. And the increase was manufactured by withdrawing supply: the US corrugated industry has permanently closed roughly 3.5–4m tons of capacity since 2023, about 2m of it since the start of 2025. Demand went the other way. Box shipments fell 1.9% year-over-year in the first quarter, containerboard production fell more than 8%, and Bloomberg Intelligence forecasts a 1.5% decline in shipments for the full year. The Association of Independent Corrugated Converters, the buyers' side, publicly opposed the third increase of the year, noting that producers historically announce large numbers and secure smaller ones.

The cost leg moved with the price leg

Old corrugated containers, the recycled fiber that feeds recycled-grade mills, rose every month of 2026 from roughly $80 a ton toward $100–130; that is still below the five-year average of about $140, so this is normalization off a depressed base. Freight and electricity rose too — International Paper cited exactly those three inputs when justifying its own increase. The result is visible in the June quarter: gross margin fell year-over-year at all four names despite a rising index.

Packaging Corporation grew June-quarter revenue 14.7% to $2.49bn and produced no additional operating profit at all, with operating income flat and net income down 20.5% to $192.1m. Its volumes are the best in the group: total corrugated shipments rose 24.3% per day against a year earlier, and excluding the Greif containerboard business bought for $1.8bn in September 2025, legacy shipments still rose 4.1% per day to a quarterly record. Greif added $0.14 of earnings per share. None of it reached the operating line, and the pattern is four quarters old: revenue growth of 6.0%, 10.1%, 10.6% and 14.7%, against operating income down slightly in each.

International Paper is the same squeeze without the growth. June-quarter revenue fell 11.3% to $6.0bn, operating income fell to $45m from $186m — a margin of 0.75% — and the company posted a $12m net loss. Its answer is supply: five sites closing across six states for an expected $230m earnings improvement, the Georgetown pulp mill shut, a paper machine at DS Smith's Kemsley mill in the UK closing while another is modernized, and $500m of DS Smith synergies targeted by 2027. "We are taking cost and complexity out of the business," chief executive Andy Silvernail told investors on the second-quarter call, where management also cut its second-half demand assumption from a 1% uptick to generally stable. Its North American box volumes rose 1.7% a day — share gain in a shrinking market.

Smurfit Westrock, the Dublin-headquartered combination of Smurfit Kappa and WestRock and the largest of the four by market value, reported the widest gap between commentary and accounts. Revenue rose 1.1% to $8.03bn; operating income fell 44% to $309m, halving the margin over five quarters. Paper markets "are as strong as I have seen in my lifetime within this industry," chief executive Tony Smurfit said on the 29 July call, with the company sold out of nearly every grade. Management's own explanation of the timing is the key to the group: the mill-side increases will be recovered through its converting operations through this year and into next.

Graphic Packaging shows the cost leg with no index to offset it. It sells folding cartons, cups and food containers off bleached and recycled paperboard grades to consumer-goods and quick-service restaurant buyers, not linerboard, and it re-guided 2026 inflation to at least $150m from $60–65m. Gross margin fell to 13.35% from 19.19%, operating income 56%, and consensus 2026 earnings per share of $0.72 is 61% below the $1.48 it delivered in 2025 — against $5.5bn of net debt.

What the shares did, and what they did not settle

September was broad rather than papery: three quarters of the 3,344 symbols priced daily are down over the trailing month, the average decline 4.3%. Within that, International Paper (−15.5%) and Graphic Packaging (−18.3%) were outliers; Packaging Corporation (−4.3%) and Smurfit Westrock (−7.1%) were not. Over twelve months the four do not move together at all: Packaging Corporation is up 11.6% and Smurfit Westrock 5.7%, while International Paper is down 24.0% and Graphic Packaging 52.1%. The clearest company-specific news in the month was a valuation call — BofA Securities cut its Packaging Corporation target to $260 from $280 on 11 September while keeping a Buy — and JPMorgan raised its target to $320 twelve days later.

So the verdict splits. The de-rating at Graphic Packaging is doing rational work: earnings estimates cut 61% and the shares at 0.86x book. International Paper's decline has outrun a business whose volumes are growing and whose 2027 consensus of $3.04 a share puts it near 12x, though its trailing price-to-earnings and enterprise-value multiples are unusable after 2025's $3.5bn net loss, so the cheapness rests on 0.77x sales and 1.28x book. Packaging Corporation, at 22.6x forward earnings and 13.6x trailing enterprise value to EBITDA against Smurfit Westrock's 8.6x, is priced for a 2027 step-up to $13.61 that its last four quarters have not begun to deliver. Smurfit Westrock has advanced without paying for it. What none of them has yet shown is a quarter in which the published price beat the cost of making the board.

Fastmarkets has proposed adding a corrugated box cost index alongside its containerboard prices. When the benchmark that sets an industry's revenue needs a second benchmark to describe what buyers actually pay, the distance between an announcement and a dollar has become the business.

Ralph Lauren raised its outlook for the year to March 2027 on 6 August, after a quarter in which it charged roughly 15% more for the average item and still grew comparable store sales at its own stores and websites by low double digits. Revenue rose 14% to $2.0bn and adjusted earnings per share 22% to $4.59, and the company lifted its full-year constant-currency growth outlook to about 5% to 6%.

That combination is the whole premium-apparel argument in one quarter. A brand like this earns three different kinds of money — units shipped to department stores at a margin fixed months before the selling season, sales through its own 504 stores and digital shops where realized price sets the gross margin, and royalties on fragrance, eyewear and home goods on inventory it never touches. All three worked at once, and the shares have fallen 15.9% since, from $401.37 on 10 August to $337.47 on 23 September, without a single session down more than about 3% and without a disclosed company event.

The meters that worked

Gross margin reached 73.7%, up 140 basis points, which the company attributed to price realization and channel and geographic mix "more than offsetting incremental pressure from tariffs and other product costs." North America wholesale revenue rose 22%, though roughly 15 points of that came from resumed shipments to a returning luxury account and a timing shift out of the prior quarter; European wholesale rose 8% in constant currency. Asia revenue rose 24% to $589m on comparable sales up 23%, with digital commerce up 32%. Inventory fell 3% in constant currency — nothing waiting to be marked down.

For the full fiscal year to March 2026, revenue grew 14.6% and operating income 26.5%, against revenue growth of 2.9% and 6.7% in the two preceding years.

What a broken meter looks like

PVH Corp, which owns Calvin Klein and Tommy Hilfiger and sells mostly through other retailers' order books, is the counter-example. June-quarter revenue fell 3.2% to $2.097bn, wholesale fell 6% across every region, Calvin Klein fell 7% and licensing revenue fell 13% — the royalty stream contracting rather than carrying the profit pool. The company also took a $439m non-cash goodwill impairment, and its spring-2027 European wholesale order book is already committed down mid-single digits. "We are really leaning into cost as a real driver of EBIT margin expansion," chief executive Stefan Larsson told investors on 3 September. PVH trades at 6.3x forward earnings and 0.73x book.

Oxford Industries, which sells Tommy Bahama and Lilly Pulitzer almost entirely through its own stores, restaurants and websites, is the third case and the cleanest demand signal. It fell 15.7% in the session after its 3 September report on 2.05m shares — roughly eight times the surrounding daily volume — after cutting full-year sales guidance to $1.43bn–$1.47bn and adjusted earnings to $1.60–$2.00 a share against $2.11 last year, while raising the quarterly dividend to $0.70, an annualized $2.80 the guidance no longer covers. Chairman and chief executive Thomas Chubb named the cause: "What we did this year is I think we just went too far too fast in shifting up the pricing tiers." Entry-level Lilly Pulitzer dresses under $200 went from about half the assortment to a third. Tommy Bahama comps turned positive. At 14.8x forward earnings on a consensus that has been cut 21%, it is the expensive one of the three.

The refunds nobody's margin earned

One reason reported profitability across the group is unreadable: the Supreme Court struck down tariffs imposed under the International Emergency Economic Powers Act on 20 February 2026, and refunds began flowing in April. PVH booked $107m in the June quarter — about $1.80 of earnings per share, leaving an underlying operating margin nearer 6% than the reported 11.1%. Oxford booked $42m, worth $2.07 of its $3.25 reported earnings per share. Ralph Lauren disclosed no such benefit; its margin gain is the only one of the three that the business itself produced.

The verdict

Two of the three de-ratings are earned. PVH's money is other retailers' commitments and those commitments are shrinking; Oxford priced its core customer out of its own dress racks and told investors spring 2027 is the first corrected season. Ralph Lauren's decline is the one nothing in the filings explains — sixteen analysts rated it a buy against two holds as of 10 September, and forward estimates have risen, not fallen. The likelier reading is a sector-wide repricing of discretionary consumption: over the same thirty days the S&P 500 exchange-traded fund rose 0.6% while Kontoor fell 21%, Lululemon 17%, Tapestry 16% and Nike 12%, with Abercrombie & Fitch the notable gainer.

What the reader is left holding is a valuation question rather than an operating one. At 18.3x forward earnings, down from 21.2x in August, Ralph Lauren is priced the way it was priced in 2024 and 2025, when it grew low single digits and its Asian business had not yet compounded at this rate. Either the market has decided that 15% price increases cannot repeat into a slowing consumer, or it has stopped distinguishing between the brand that raised guidance and the two that cut it.

Revenue is a dealer's decision

Thor Industries — the Elkhart, Indiana builder of travel trailers, fifth wheels, Class A, B and C motorhomes and European campervans, every one of them sold through an independent dealership — reported its fiscal fourth quarter on 22 September and then declined to tell investors what the coming year looks like. The company said insights from September industry events made it prudent to delay, and that it expects a relatively flat retail environment in fiscal 2027.

Withholding a forecast is the news, but the mechanism underneath it is the story. Thor is not paid for recreational-vehicle demand in any form its revenue line makes visible. It recognizes a sale when a unit ships to a dealer who has borrowed against that unit on a floorplan credit line, so the top line measures how much financed inventory the dealer network is willing to carry. Six days before the print, the Federal Reserve raised its benchmark rate a quarter point to 3.75%-4.00%, its first increase since 2023, with 16 of 18 officials penciling in another and the 10-year Treasury yield back at 5%. Because the hike answered an inflation impulse Chairman Kevin Warsh traced to spiraling oil prices, one decision repriced the dealer's floorplan rate, the retail buyer's loan and the fuel to run the vehicle at once.

The destocking case, and what breaks it

Thor's own disclosures make the channel-correction argument well. North American dealer inventories ended fiscal 2026 at roughly 64,000 units against roughly 73,300 a year earlier, a 12.7% reduction, with towable stock down 20.6% sequentially — while total North American backlog rose 7.5% to $1.645bn and the towable segment's backlog rose 74.6% to $916.6m. Orders piling up as shipments fall is a channel taking delivery late, not one cancelling.

The quarter itself was ugly regardless. Revenue fell 8.4% to $2.312bn and operating income halved, down 50.8% to $50.5m; gross margin narrowed 2.3 percentage points to 12.4% on lower volumes, unfavorable mix, increased promotional activity and higher material costs. North American towable sales fell 22.7% on a 19.7% decline in unit shipments — revenue falling faster than units, so mix and discounting deepened the hole rather than cushioning it.

What breaks the clean destocking read is the industry data. Statistical Surveys counted US retail registrations down every month of 2026 through June — 14.5% in January, 24.1% in February, 21.5% in June — with May marking the ninth consecutive year-over-year decline. Against that, wholesale shipments through April fell 13.5%, and the industry body cut its 2026 forecast to a 314,000-unit median from 349,000. Retail is contracting at least as fast as the factories are shipping. Supplier LCI Industries has marked down further still, to 280,000-300,000 wholesale units.

"Fiscal 2026 was a structurally important year for THOR," chief executive Bob Martin said in the 22 September release. "With numerous macro elements working against our industry, we took action to address our margin profile, our production footprint and the supply chain pressures weighing on the business." Martin told the trade press that the inflection point the industry expected never arrived, naming interest rates, fuel costs and inflation. The shares still rose 5.8% on the day, to $73.80, because revenue beat consensus even as earnings per share missed — from a level near a 52-week low. Thor trades at 0.89x book, below its own net asset value, and 19.6x forward earnings on a $3.70 consensus that has already been cut.

Winnebago, and the recovery in the price

Winnebago Industries, the smaller builder whose Grand Design and Winnebago towables sit alongside Newmar motorhomes and the Barletta and Chris-Craft boat lines, reported the same shape a quarter earlier: revenue down 9.9% to $698.7m and operating income down 23.8%, with production aligned to retail demand as dealer ordering stayed measured. "Consumers who are drawn to the outdoor lifestyle remain engaged, but continue to navigate affordability pressures," chief executive Michael Happe said on 25 June, citing cumulative inflation, elevated interest rates and geopolitical uncertainty in the timing of discretionary purchases. He cut the industry wholesale outlook to 290,000-310,000 units.

Winnebago is the cheapest asset in the group at 0.63x book and 9.3x trailing enterprise value to earnings before interest, taxes, depreciation and amortization. But its 11.7x forward earnings against 20.2x trailing is not cheapness; it is a forecast. Closing that gap requires earnings per share to rise from $0.91 to $2.34 — more than a doubling, embedded in today's price.

Polaris earned the opposite quarter and was sold anyway

Polaris, which builds side-by-sides, ATVs, snowmobiles, motorcycles and pontoon boats out of Medina, Minnesota, ran the other way: June-quarter revenue up 9.2% to $2.023bn, gross margin up five percentage points to 23.6%, operating income up nearly sixfold to $132.0m, North American retail up 4% with a fifth straight quarter of off-road share gain. Guidance was raised for the second time this year. The stock fell 9.9% the next session.

The reason is in the same filing: North American off-road dealer inventories were up about 7% in the quarter, faster than retail grew. Polaris is refilling the channel Thor is draining — and it owns a slice of the credit doing it, through Polaris Acceptance, its 50/50 floorplan venture with Wells Fargo Commercial Distribution Finance, which carried $1,788.9m of net dealer receivables at 30 June and obliges Polaris to repurchase repossessed product up to roughly $89.5m this year. Chief executive Mike Speetzen said in July that promotional activity from rivals had been surgical rather than broad; on 18 September UBS cut its target to $61 citing promotional headwinds. At 3.63x book and 16.9x forward earnings — with 2025's $465.5m net loss leaving no usable trailing figure — Polaris is the most expensive asset here against the weakest cash conversion.

BRP, the Can-Am and Sea-Doo maker taking side-by-side share from Polaris, grew revenue 19.8% last quarter and raised full-year guidance, and its shares fell 10.9% over the following month anyway. Its free cash flow yield, near 19%, is among the highest in the complex. Chief financial officer Sébastien Martel quantified the tariff carried into fixed model-year prices on the 3 September call: "We are talking about a net exposure this year of CAD 200 million and next year of CAD 225 on an annualized basis." He also described signing up 90% of BRP's US dealers to a new in-house retail financing program in two weeks — building the credit layer Polaris already owns.

The verdict

Thor's de-rating is earned by its income statement and only half-explained by its channel: the backlog and the emptied lots are genuine destocking, but registrations falling faster than shipments means the restock, when it comes, restocks into a shrinking retail market. Polaris's is not earned by its profit and loss at all — it is earned by the 7% inventory build sitting on dealer floorplans that just got more expensive, and by a promotional cycle its own guidance assumes away. Winnebago sits between them, priced for a doubling of earnings it has not begun to deliver.

A dealer who financed a trailer in March is carrying it at a rate that has since gone up, against a buyer whose loan costs more and whose fuel costs more for the same reason. That last variable is not settled: crude freight futures dropped roughly a quarter intraday on 22 September only because an Iranian official floated reopening the Strait of Hormuz within a week if Washington lifted its blockade.

BWX Technologies ended June with $8.4bn of orders on its books, up 40% from a year earlier, and a month later raised every line of its 2026 outlook. "We had strong second quarter 2026 results that were ahead of our expectations," said Rex D. Geveden, president and chief executive, on 3 August, lifting adjusted EBITDA guidance to $662m-$672m, earnings to $4.70-$4.80 a share and free cash flow to $345m-$360m. The shares are down 31% over ninety days.

That is the problem with the label. Three companies sit under one nuclear-fuel-cycle heading and are paid in three unrelated ways — by billable progress against a defence book, by the unit of enrichment work, and by the pound of uranium under contracts written years in advance. Only one of the three has a business that actually deteriorated, and it is not the one the fuel price would predict.

Paid by the Navy

Roughly three-quarters of BWXT's revenue comes from its Government Operations segment — about $2.35bn in 2025 against roughly $853m commercial — making precision naval propulsion reactors and components as a security-cleared sole-source supplier. June-quarter revenue rose 18% to $901.6m, but reported operating income fell 12.2%, the fourth straight quarter of that pattern, with gross margin down to 22.4% from 25.1%. The mix explains it: commercial work, grown largely by acquisition, carries an 8.0% operating margin, and management guided Government Operations to high-single-digit revenue growth rather than low teens because cost-reimbursable naval contracts book less revenue when costs fall — while raising that segment's margin target to about 20.5%. The company also agreed on 31 July to sell its medical isotope business to Nordic Capital in a deal valued at up to $800m, and the US Army picked its BANR design in late August for the Janus microreactor program, worth up to $2.2bn across five vendors. Consensus 2026 earnings went up during the decline, to $4.74 a share. The stock trades at 29.9x forward earnings against roughly 47x in May, and 16.8x trailing gross profit against 25.9x in February — on gross profit that grew.

Paid by the separative work unit

Centrus Energy, with 467 employees, is the only US-based producer of high-assay low-enriched fuel and sells enrichment services to utilities. Here the physical meter genuinely contracted: the volume of separative work units sold fell 23% year over year, with average unit cost up 13%. Revenue still rose 14% to $176.1m on uranium resales, but operating income fell 69% to $10.4m and gross margin compressed to 28.3% from 34.9%. "This was another strong quarter of financial and operational progress," chief executive Amir Vexler said on 5 August, citing commercial wins and a contingent enrichment backlog grown to $3.0bn. On 9 September the company priced $500m of stock and warrants, the common warrants covering 6,992,382 shares. Consensus has 2026 earnings at $2.55 a share against $3.90 in 2025 — roughly a third lower — and the stock trades at 59.4x forward earnings versus 60.3x trailing: no growth priced at all. Urenco announced a roughly 50% expansion of the largest US commercial enrichment plant in June, which dates the scarcity.

Paid by the pound

Cameco's uranium book improved. It realised US$67.79 a pound in the June quarter against a spot indicator near US$89.50, and lifted 2026 realised-price guidance to C$91.00-96.00 from C$85.00-89.00, with production held at 19.5-21.5 million pounds. "The long-term uranium price strengthened further, supported by increased on and off-market contracting activity," chief executive Tim Gitzel said on 30 July. Reported net income still fell 92% to $25.2m, because the damage sits in the reactor affiliate: Westinghouse, 49%-owned, swung from a $126m contribution to a $10m loss on Cameco's share as the Czech Dukovany construction contribution lapped, and is guided to a full-year net result between a $75m loss and a $10m loss. Cameco now trades at 58.2x trailing gross profit against 38.3x in May — dearer after a 21% fall in market value, because profit shrank faster than price.

What the rest of the complex did

The thirty-day leg is mostly not a nuclear story. The 30-year Treasury yield sat near 5.3-5.4% in late September, its highest in about two decades, and regulated utilities fell almost as hard as the nuclear names — NextEra down 8.4%, Dominion 7.8%, Entergy 6.1% — after the sector gave back an early-year gain of more than 11% to finish among the worst of the eleven main industry groups. The pre-revenue reactor developers took it worst: Oklo down 73% and NuScale 79% over twelve months, against BWXT's 20%.

So the verdict divides cleanly. The fuel meter is not deteriorating — term uranium reached roughly US$96 a pound in September on Kazakh supply constraints, a nominal record, some $6 above spot. Centrus's decline is earned: fewer units sold, thinner margins, dilution, and a flat trailing-to-forward multiple on falling earnings. Cameco's is half-earned, concentrated in Westinghouse rather than in pounds, and its valuation has not yet done its work. BWXT's is the one nothing in the reported numbers explains, beyond a discount rate applied to payoffs dated 2028 and later and a margin line that has ground down for four quarters.

The tightest constraint is a calendar, not a price. Waivers permitting Russian enriched uranium into the United States all terminate by 1 January 2028, and Centrus's own centrifuges at Piketon are not scheduled to produce commercially until 2029.

Accenture agreed on 18 September to commit at least $1bn over five years to a new kind of staffing: teams of embedded evaluators who red-team models, run alignment assessments and test safeguards. The counterparty, the AI developer Anthropic, committed the same amount, and the arrangement is non-exclusive on both sides. Accenture's Faculty unit leads the work; the shares rose about 6% premarket.

That matters because of how Accenture — the Dublin-based firm that sells strategy, systems integration, cloud and outsourced operations to large corporates and governments, with 799,000 employees — actually earns money. Revenue is headcount multiplied by utilization multiplied by a realized bill rate. Every hour software automates is an hour nobody invoices, which is the entire bear case on the industry and the reason the stock fell nearly 18% in one session on 18 June, its worst day on record, after management trimmed full-year local-currency growth guidance to 3-4% citing weak US federal work. On Thursday 1 October, at 8:00 a.m. Eastern, Accenture reports its fiscal fourth quarter and issues its first full-year fiscal-2027 guidance.

What the last print actually said

Reported revenue growth decelerated to 5.6% in the May quarter, roughly 3% in local currency. New bookings fell 2% to $19.32bn, split $10.26bn consulting and $9.06bn managed services — so the hour-billed half booked more work while the outcome-priced half grew faster, 5% in local currency against consulting's 1%. Gross margin slipped to 32.77% from 32.87%.

The labor meters, though, look healthy. Utilization rose to 93% from 92%, attrition excluding involuntary terminations fell to 14% from 16%, and headcount grew about 1% against 5.6% revenue growth — revenue per employee of $91,489, up 5.7% over twelve months. "Demand for large scale reinvention remains strong—104 quarterly client bookings of $100 million or more year-to-date, up 13%," chair and chief executive Julie Sweet said on 18 June. What vanished that day was the scoreboard: Accenture retired its advanced-AI bookings disclosure, arguing AI now runs through everything it sells, after cumulative bookings of about $11.5bn since 2023.

The mirror and the control

Cognizant, the Teaneck outsourcer whose delivery is mostly offshore across financial services, healthcare, products and communications, tells the same story in a different accent. June-quarter revenue of $5.5bn grew 4.5%, decelerating, with gross margin down 30 basis points and net income off 1.4% — yet adjusted operating margin expanded for a sixth straight quarter and full-year adjusted earnings guidance was raised to $5.70-$5.82 a share. Headcount fell 900 sequentially to 356,700. "Our organic revenue growth momentum continued in the second quarter and was at the high end of our expectations," chief executive Ravi Kumar S said on 29 July. The shares fetch 10.28x forward earnings, below the roughly 14-15x they carried in May.

Genpact is the test that breaks the thesis it was meant to prove. The former General Electric back office runs accounts payable, record-to-report and procurement for banks and insurers, and it is the only member of this group accelerating — revenue growth of 5.6%, then 6.7%, then 7.1% — with gross margin widening to 36.50%. Its Advanced Technology Solutions segment rose 24.1% to $363.3m, 27% of revenue, sold as annuitized recurring contracts with minimum volume commitments rather than per-person billing; chief executive Balkrishan Kalra told investors the firm is on track for more than $1bn of agentic contract value in 2026. The legacy back office grew about 2%. Genpact trades at 8.22x forward earnings and is the worst thirty-day performer of the five names here, down 9.9%.

What the rebound is made of

From the 22 June low Accenture has recovered 47% and Cognizant 41%, while Infosys and Tata Consultancy Services — running the same offshore pyramid, and sold just as hard on the day Accenture cut guidance — fell 4.4% and 1.9% over the same stretch. Delivery model did not separate them; listing venue and the direction of consensus revisions did. Over the past thirty days all five fell.

So the honest split: the business earns the part of the recovery attributable to pricing discipline — utilization at 93%, falling attrition, revenue per employee up 5.7% while the payroll barely grew. It does not earn the rest. Accenture at 14.54x trailing earnings against 21.4x at its fiscal-2025 year-end is a de-rated share price partly restored, not an order book reaccelerating; consensus itself models fiscal-2027 revenue growth of 3.9%. And Genpact proves the labor meter is not what is being repriced: the one company whose revenue is least tied to the hour is the one being sold hardest.

Accenture spends around $9bn a year buying capability — $4.18bn of it this year on three operational-technology security firms — and has now agreed to spend $1bn more on evaluators it did not previously employ. Next Thursday it has to say what all of that is worth in fiscal 2027, without the AI bookings number it used to publish.

Four customers

The American generic drug business has four customers. Red Oak Sourcing, ClarusONE, Walgreens Boots Alliance Development and Econdisc Contracting Solutions — joint ventures pairing the big pharmacy chains with the big wholesalers — together buy more than 90% of the generic volume purchased in the United States. A manufacturer that wins approval to copy an off-patent molecule does not enter a market. It negotiates a per-pill price with effectively four counterparties, and that price goes down.

So the past twelve months in generics are not the revival the group average suggests. Among six US-listed makers of generics and active ingredients, three roughly doubled or better — Teva up 100.6%, Amneal 96.4%, Viatris 78.3% — while Amphastar, Dr. Reddy's Laboratories and ANI Pharmaceuticals all fell, ANI by 21.9%. The split does not sort by who is best at generics. It sorts by who sells something the four buyers do not price, and whose balance sheet was being marked for distress a year ago.

Teva sold itself out of the contract

Teva Pharmaceutical Industries, the Tel Aviv maker of generics, specialty medicines and active ingredients, reported June-quarter revenue of $4.14bn, down 0.8%. Gross profit grew 2.4%, all of it mix: margin reached 52.0% against 50.3% a year earlier. Underneath sit two opposite businesses. Austedo, its tardive dyskinesia treatment, delivered $696m, up 40% in local currency, and full-year guidance rose to $2.45–2.60bn — a midpoint that reaches in 2026 the $2.5bn the company had targeted for 2027. With Uzedy and Ajovy, the three brands Teva prices itself grew 43% and are guided to roughly $3.7bn this year, about a fifth of group revenue. Meanwhile US generics including biosimilars fell 31% to $660m, mostly because generic Revlimid drew additional entrants.

"The innovative portfolio is reshaping our financial profile with stronger revenue growth, margins, and free cash flow," chief executive Richard Francis told investors on the 29 July call.

The other half of the doubling is credit, not commerce. Net debt to EBITDA fell to 2.42x from 3.10x against a 2.0x target for 2027, and on 4 September S&P Global Ratings raised Teva to BBB−, the last agency to restore investment grade; Teva then refinanced $4.9bn at lower coupons. That cuts the cost of the capital structure without changing what a pill earns.

The scarcity rent has a clock on it

Amphastar Pharmaceuticals, a California maker of difficult injectables and inhalation products — enoxaparin, glucagon kits, naloxone, and the BAQSIMI nasal glucagon franchise bought from Eli Lilly — is the test of whether manufacturing difficulty buys durable pricing. On its own numbers, it buys a lease. Gross margin fell from 54.5% in 2023 to 49.5% in 2025 while operating income dropped 31.7% to $140.4m, and the March quarter bottomed at 41.1%. June's recovery to 50.8% came from brand-new approvals — ipratropium bromide inhalation aerosol carrying 180-day exclusivity, teriparatide and iron sucrose — which chief financial officer Bill Peters set against pricing pressure in BAQSIMI and glucagon. BAQSIMI is the erosion in miniature: net sales fell 3% to $45.5m while prescriptions rose 17%, volume adding $6.9m and rebates and discounts taking about $8.1m away. A $100m milestone payment to Lilly falls due in the September quarter.

Integration did not protect Dr. Reddy's

Dr. Reddy's Laboratories, the Hyderabad group that makes the active ingredients it formulates, had a collapse rather than a grind: June-quarter revenue of ₹81.2bn fell 5.0%, gross margin dropped to 46.5% from 56.9%, and net income fell 68.5%. Generic semaglutide launched in Canada on the first G7 authorization, then stopped after roughly 180,000 pens, with a ₹2.4bn provision — about $27m converted — and supply now expected in November. Chief executive Erez Israeli blamed "semaglutide-related challenges" and an impurity that emerged at scale-up. But fiscal 2026 gross margin had already fallen to 52.8% from 58.5%, with operating income down 32.0% — the deterioration predates the pens.

Nor is this a reshoring trade. Generics, biosimilars and their ingredients are excluded from the April 2026 Section 232 order that set a 100% default rate on patented pharmaceutical imports — an advantage, not a cost. The exclusion is provisional: the order mandates a generics policy review within a year, and a phased generics tariff has been trailed for 2028.

What the group pays for

Viatris, the Pennsylvania group behind EpiPen, Lyrica and the Semglee biosimilar, grew June-quarter revenue 4.9% to $3.76bn with gross profit up 9.3% — its first growth after sliding from $17.9bn of annual revenue in 2021 to $14.3bn in 2025 — and still trades at 6.8x forward earnings, the cheapest of the six. Amneal, whose AvKARE unit supplies the Departments of Defense and Veterans Affairs, grew revenue 9.9% and operating income 31.1%, but at 19.0x forward and a 1.3% free-cash-flow yield its shares have run well ahead of that. ANI Pharmaceuticals, the Minnesota maker of Cortrophin Gel and the ILUVIEN and YUTIQ eye implants, grew revenue 25.9% to $266m and nearly tripled operating income — and fell 21.9% over the year, into 8.0x forward earnings and the group's lowest enterprise value against EBITDA at 6.5x.

Teva, the biggest gainer, is the most expensive on 19.4x forward earnings and 18.9x EV/EBITDA. That multiple sits on a deliberate trough — consensus has 2026 earnings per share at $2.01, down 23%, before $3.10 in 2027, which would be about 12.6x. Consensus also has group gross profit essentially flat this year. So the branded pivot and the investment-grade upgrade earn a real part of the doubling; the rest is the market paying now for 2027. Amphastar's decline is earned by its own margin line. ANI is the cell nothing in the generics story explains.

The arithmetic of the four buyers did not change this quarter. The only force that lifts a generic price is a rival walking away — and the walking away is measurable: US drug shortages now last, on average, more than five years.

Analog Devices closed a $1.5bn all-cash purchase of Empower Semiconductor on 7 July, buying a company that puts voltage conversion inside the processor package. Six weeks later the Wilmington, Massachusetts maker of data converters, amplifiers and power-management chips reported its first $4bn quarter, with communications revenue up 84% year over year and data center accounting for 80% of that segment. Optical and power within it each more than doubled.

That matters because Analog Devices is still read as the industrial and automotive analog cycle in listed form — distributor channels, lead times, restocking. The reading has stopped describing the company. Power delivery inside an artificial-intelligence rack is now a contested socket, and three firms are being paid for it three different ways: per part, per socket, and — in one case — per patent.

The record quarter the market marked down

Analog Devices' growth is accelerating, not fading: revenue rose 30.4% year over year in fiscal first quarter 2026, 37.2% in the second and 39.6% in the third, while operating margin widened from 31.5% to 40.1% across the same stretch. Industrial, still 49% of revenue, grew 53%; automotive, a quarter of the business, grew 16%. "ADI delivered a strong third quarter, exceeding the midpoint of our revenue, margin, and earnings outlook as we capitalized on broad-based demand," chief executive and chair Vincent Roche said on 19 August. On the same call, management called intermediate-to-core conversion — feeding 6,000-amp, sub-one-volt processors — one of the fastest and largest growing analog opportunities, and said book-to-bill was not unusually elevated, with customers running lean and minimal restocking.

The shares have gone the other way. Analog Devices closed at $385.23 on 23 September, 13.5% below its 22 June close, though still up 56% over twelve months, and its trend turned negative in the second week of September — three weeks after the record print. Trailing price/earnings has fallen to 45.5x from 57.9x in mid-August, most of that delivered by a higher earnings base rather than by the 1.3% decline in market value; forward earnings are 29.9x against consensus for 66% earnings growth this fiscal year. The company guided fourth-quarter revenue to a record $4.3bn, plus or minus $100m, and does not report again until 1 December. Across the wider group the drawdowns are worse — onsemi is down 44.7% from its high, Microchip 26.6%, Texas Instruments 18.0% — which makes sector de-rating the likelier reading than anything discoverable about Analog Devices itself.

The other two meters

Monolithic Power Systems, which designs the power-management chips inside Nvidia reference boards, is the direct socket-content read, and it is accelerating: enterprise data revenue of $380.6m in the June quarter, up 164% year over year and 38.8% of the company, with the full-year growth floor for that segment raised from 85% to 130%. "Our results demonstrate the strength of our diversified model," founder-chief executive Michael Hsing said on 30 July. The stock has fallen since early September and sits 19.8% below its high, at 36.9x trailing gross profit against the roughly 39x that marked the top of its range five weeks ago.

Vicor, the Andover module maker whose vertical power delivery patents cover the architecture the industry is migrating onto, is the exception — and the reason is legal. On 16 September it licensed those patents to a leading AI original-equipment manufacturer on terms that let the licensee buy the modules from someone else, then on 21 September doubled its third-quarter sequential growth guidance to more than 20% citing royalties from it. Its shipped business is strong underneath — advanced products up 45% sequentially to $94.2m and one-year backlog up 26% to $379.7m — and the reported 1.6% revenue growth is an artifact: the year-ago quarter contained a $45m patent-litigation settlement, and core revenue grew 49.3% without it. Gross margin's rise to 58.0% from 55.2% is likewise part mix, since royalty dollars carry almost no cost of goods. At 79.1x forward earnings, Vicor is dearer than when a near-identical share price implied roughly 68x in May.

What the money actually buys

Royalty income is real revenue, but it is paid on someone else's shipments at a rate that rises as litigation escalates — it does not scale with Vicor's own fab, and the second fab that chief executive Patrizio Vinciarelli said in July is necessary on the path to $2.5bn of revenue is a fixed cost that royalties will not fill. Analog Devices and Monolithic Power are the two here converting AI rack demand into shipped silicon, and both are being priced as analog cyclicals rolling over while their data-center lines compound. Nothing in either company's disclosure explains the de-rating; the shares are carrying a sector verdict their own numbers contradict.

Between now and 1 December there is no Analog Devices print to settle it. The next news out of this group may well come from a docket rather than a fab.