DK Street Journal

On 16 July a Medicare contractor published the final local coverage determination for molecular testing of solid-organ transplant rejection, and it did something a commercial customer never does: it wrote down how many tests a doctor may order. Up to six AlloSure Kidney surveillance tests in a patient's first post-transplant year, four in each of the next two, twelve in year one for lung. CareDx, which sells those donor-derived cell-free DNA tests along with the AlloMap heart gene-expression assay, rose 35.6% that session. The policy took effect on 30 August.

That is the machinery underneath this corner of diagnostics. A laboratory with an FDA-cleared or Medicare-covered assay does not negotiate a price and does not set a quantity; the MolDX program does both, and the rest of the payer market follows. The money at stake is being repriced right now: on 21 September the Centers for Medicare & Medicaid Services posted preliminary CY2027 laboratory fee schedule rates that cut hundreds of codes by as much as 15% and are meant to save roughly $1bn a year. They are finalized in November and take effect 1 January.

The company that has already been repriced twice

CareDx knows what an adverse ruling costs. A MolDX billing article in March 2023 limited reimbursement to one test per encounter; the company paused Medicare submissions on AlloSure Kidney, withholding about 3,200 tests worth roughly $8.9m of quarterly revenue, and full-year revenue fell 12.9% in 2023 to $280.3m. "We believe publication of this draft policy solidifies coverage for surveillance testing," chief executive John Hanna said of the proposal that became this July's final rule.

Second-quarter testing revenue reached $100m against $62m a year earlier on volume up 17%, or about 58,000 tests — roughly $1,720 realized per test, and about two-thirds of the growth came from price rather than units. That realization still sits well below the $2,840 AlloSure rate and $3,240 for AlloMap in the preliminary 2027 schedule, which spared the transplant portfolio while cutting elsewhere; the gap is payer mix. Guidance is now $490m–$500m and the company bought Naveris, whose NavDx head-and-neck cancer assay carries its own Medicare coverage, for $160m up front.

Guidelines are the second gatekeeper

Adaptive Biotechnologies bills clonoSEQ, an FDA-cleared blood-cancer test that finds one malignant cell in a million, at a Medicare rate of $2,007 set through the MolDX gapfill process. It realizes $1,382 — about 69% of that rate — on 36,111 tests, up 43%, with minimal residual disease gross margin at 72% against 64%. "Q2 was an exceptional quarter for Adaptive and a clear validation of our strategy and execution," chief executive Chad Robins told investors on the July call. Guidance went to $268m–$278m.

Then the guideline bodies moved. On 17 September the National Comprehensive Cancer Network's myeloma guidelines added a dedicated MRD testing page naming clonoSEQ; on 25 September the International Myeloma Society defined cure as five years in remission with at least four negative assessments at that sensitivity — a definition that multiplies billable assays per patient. Adaptive's 13.2% thirty-day gain is four sessions, clustered on those dates.

The one still in the queue

Veracyte sells the Afirma thyroid classifier and Decipher prostate test into indications where reimbursement was settled years ago, so its growth has to come from penetration — management puts Decipher at roughly 20% of the low-risk prostate market. It is the best business of the three by margin: testing revenue of $145.7m, up 19%, gross margin 72.2%, operating margin 15.3%, net income of $25.5m against an operating loss a year earlier. Shares fell 22.5% on 31 July anyway. Prosigna's centralized version and the TrueMRD bladder test were left out of guidance pending a MolDX technical assessment and the pricing talks that follow, and Decipher's low-risk volume outlook was trimmed by about 1,000 tests after an NCCN change. "Q2 was a milestone quarter for Veracyte as we launched two new products," chief executive Marc Stapley said on the 30 July call.

The same institution that re-rated Adaptive de-rated Veracyte inside two months. That is the finding: in this segment the business event is a committee document, and the share prices have already discounted the documents that exist. CareDx trades at 53.7x forward earnings — its 30x trailing figure is unusable, containing a $113m gain on the Lab Products sale — after a six-month run of 285%, and consensus has revenue growth decelerating to 6.3% in 2027, treating this year's step-up as one-time. Adaptive's 15.6x forward sales sits above its 14.8x trailing, because a Genentech milestone rolls off, against a consensus 2026 net loss of $85m. Veracyte is at 23.6x forward earnings against 31.4x trailing, 6.1x forward sales and a 4.6% free cash flow yield, and is the laggard of the diagnostics group over twelve months.

What none of this prices is the absence of coverage. GRAIL's Galleri screening test sells for $949 in cash at a gross margin of minus 28%; Medicare cannot pay for multi-cancer screening until 2028. Coverage is the whole margin.

November settles the 2027 rates. CareDx and Adaptive already know what their documents say; Veracyte's has not been written.

Accenture reports fourth-quarter results on Thursday 1 October and, for the first time, tells investors what it expects of fiscal 2027. It gets there with a shrinking order book: new bookings of $19.32bn in the May quarter, down 2% in dollars and 3% in local currency, on revenue of $18.72bn that missed consensus by $80m.

That guidance is the only scheduled disclosure left this year that speaks to the question the whole sector is priced on: whether artificial intelligence expands enterprise technology budgets or deflates the price of an hour of the work. When Accenture — the Dublin-headquartered consulting and managed-operations firm with 799,000 staff — cut full-year local-currency growth to 3–4% from 3–5% on 18 June, Indian software stocks slumped alongside it. Accenture fell 17.97% that session, EPAM 12.61%, Globant 11.18%, Cognizant 10.49% and the Infosys American depositary receipt 9.66%.

"Because the indirect impact really started in the last few weeks and mostly in discretionary spend, we do think that there will be more impact in Q4, which is why we're saying that more of the range is in play," Chair and Chief Executive Julie Sweet told investors on the 18 June call. Three months earlier she had called AI "a tailwind because it's helping us win more today and take market share".

The recovery split was not about delivery model

From that 18 June close to 28 September, the US-listed integrators came back and the Indian listings did not: Accenture recovered 36.3%, EPAM 38.5%, Cognizant 29.9%, against Infosys at -0.5% and Tata Consultancy at -6.0%. Labour arbitrage does not explain the gap. Cognizant, the Teaneck, New Jersey firm running consulting and outsourcing across financial services, healthcare, products and communications, earns $60,673 of trailing revenue per employee across 356,700 staff — within about $1,200 of Infosys's $61,869. Accenture earns $91,489.

Infosys, the Bengaluru outsourcer that sells application and operations work plus its Topaz AI suite and Finacle core-banking product, conceded on price. June-quarter revenue of $5,082m grew 2.4% in constant currency, and full-year guidance was narrowed to 1.5–3.0% from 1.5–3.5% against the 3.4% analysts had modelled. The pyramid held: utilization excluding trainees rose 1.9 percentage points to 84.9%, headcount was flat at 328,062, and operating margin reached 21.1%, above Accenture's 17.0% and Cognizant's 16.0%. What gave was the rate. Management described client requests for price reductions as broad-based across telecom, financial services, retail and utilities, arriving at renewal or mid-contract. "Over 80,000 employees are working today on coding tools such as Cloud Codex for our clients," chief executive Salil Parekh said on the 23 July call — the productivity the client is now pricing. Fixed-price contracts were 54% of revenue in the March quarter, the book where that productivity becomes revenue no longer billed.

Cognizant runs the same offshore model and was treated as the opposite case. June-quarter revenue of $5.48bn grew 4.1% in constant currency, operating income rose 7%, and trailing-twelve-month bookings reached $29.1bn. Full-year constant-currency revenue guidance was trimmed to 4.0–5.5% while adjusted earnings guidance was raised to $5.70–$5.82.

What the week did

Accenture's 50-day average crossed above its 200-day only because June's crash rolled out of the lookback; the shares fell 7.0% in the week to 28 September and sit 9.8% below their 15 September high, having surrendered the entire pop from the $1bn Anthropic model-evaluation commitment, including a 3.3% drop on 24 September. Guggenheim's Jonathan Lee downgraded the stock to Neutral, citing deal timelines pushed into fiscal 2027.

At 13.82x trailing and 11.92x forward earnings, Accenture is cheaper than at any of its last five fiscal year-ends, the nearest of which was 21.4x. Cognizant is cheaper still at 9.85x forward, below the 14–15x that looked undemanding in May. Infosys is the only one of the three whose forward multiple, 13.22x, sits above its trailing 12.56x — because consensus fiscal-2027 earnings of $0.796 a share are below the $0.83 it delivered last year.

So the split is not a growth-differential story: constant-currency growth is within two points across all three. It is a story about which direction forward estimates are moving. Cognizant's earnings guidance went up as its price fell; Infosys's went down and its price went with it; Accenture's shares recovered a third of their value on no new disclosure at all, while every disclosure it does have — bookings, guidance, analyst estimates — got worse. Deflation of roughly 2–3% a year in traditional services pricing sits underneath all of it, against Gartner's forecast of $6.31trn of worldwide technology spending in 2026. The budget grows; the price of the hour falls.

Thursday is the first morning since June that Accenture has to put a number on next year. Infosys reports on 23 October, Cognizant on 4 November.

Array Technologies, the Albuquerque maker of single-axis trackers that tilt panels across utility-scale solar farms, took $500m of new bookings in the June quarter and raised its full-year outlook on 5 August. Total executed contracts and awarded orders reached $2.5bn at 30 June, up 37% year on year and the largest the company has reported. Trailing book-to-bill ran at 1.5x.

The shares fell 46% over the three months to 28 September, the steepest decline among the six US-listed solar names — worse than Enphase, worse than Sunrun, worse than First Solar. The reason is not in the order book. It is in what sits above the common stock in the capital structure, and in a date in July.

The gate that closed on 4 July

Every one of those six names closed at a local peak on 30 June and fell continuously through the first half of July. That was the week the begin-construction safe-harbour deadline for the 30% clean-electricity investment credit passed: projects underway before 4 July lock the credit with a four-year window to 2030, while anything starting later must be in service by the end of 2027 or risk losing it to permitting and interconnection delay. A statutory gate, rather than a demand signal, synchronised the break across rooftop and utility-scale alike.

Since then the two sides have delivered opposite numbers. Enphase, which sells microinverters through distributors and installers, reported June-quarter revenue down 19.6% and US sell-through down 34% — its third consecutive decline of roughly a fifth. Sunrun, the largest US residential solar financier, added 31% fewer subscribers and booked 44% less net subscriber value per customer, at $9,444, and cut its 2026 cash-generation guidance to $200-375m from $250-450m. Wood Mackenzie and the Solar Energy Industries Association now forecast a 23% contraction in US residential installations this year. Enphase trades at 7.34x forward gross profit against 6.55x trailing — the forward figure sits above the trailing one only because consensus models the gross-profit line shrinking, and both are far below 14.55x in mid-May. Sunrun trades at 0.53x book.

Array's meter moved the other way. June-quarter revenue of $342.1m was 5.6% below last year but 53% above March, gross margin widened to 29.1%, and the raised guidance put 2026 revenue at $1.4-1.5bn with adjusted earnings of $0.68-0.75 a share and adjusted gross margin a point higher than before, at 27-28%. Over "95% of the order book is domestic," chief executive Kevin Hostetler told analysts on the second-quarter call, with half the book now tied to specifications set by developers and utilities rather than contractors. Newer products account for about half the backlog, against 15% two years ago.

What the common actually owns

The company generated $91.9m of operating cash in the first half and ended June with $307.3m of cash, enough to fund the $153m base price for its AWM acquisition. It also carries $670m of convertible notes and a Series A redeemable perpetual preferred with a $498.2m liquidation preference and $106.4m of dividends already accrued. Common equity is minus $202.1m. The market capitalisation is $589m — less than the preferred claim standing in front of it.

That is why the usable anchors here are odd ones. Trailing price-to-earnings and price-to-book are both negative and meaningless; the shares change hands at 5.10x consensus 2026 earnings of $0.75, on a trailing free-cash-flow yield of 22.9%. But that yield is measured against the common alone, and the cash it counts services the notes and accrues to the preferred before any of it belongs to a shareholder.

The verdict

Very little of Array's decline is earned by the tracker business. Orders grew, revenue guidance rose, margin guidance rose with it, and 80% of the book is expected to convert within six quarters. What the price is discounting is the two things the backlog number cannot settle: whether projects safe-harboured before 4 July all reach construction, and how much of the enterprise the common stock is entitled to when they do. On the first question Array has more visibility than the residential names; on the second it has less than almost anything else in the sector.

The order book converts over six quarters. The preferred's dividends accrue in every one of them.

Ramaco Resources has two listed share classes, and only one of them is about coal. The Class B stock is a tracking share covering the company's coal royalties, logistical tolling fees, intellectual property and possible revenue from exploratory rare earth elements and critical minerals — in practice, the Brook Mine project in Wyoming. Through September it was marked down roughly twice as hard as the rare-earth miners it now resembles, while the seaborne coking coal benchmark that sets prices for the Class A mines sat near a 17-month high.

That split is the whole story of the equity. Ramaco's coal business — underground and surface mines at Elk Creek, Berwind, Knox Creek and the RAM Mine in West Virginia, Virginia and Pennsylvania, selling to blast-furnace steelmakers and coke producers — has posted an operating loss for six consecutive quarters. So what an investor owns here is substantially a claim on a project that will not produce until 2031 and needs $3.2bn of capital to get there. Both share classes together are worth about $780m.

The coal book is a cost story, not an earnings one

June-quarter revenue fell 5.3% to $144.8m, with an operating margin of -12.6% and a net loss of $15.4m. The realized coal price averaged $116 a ton, down 6% from a year earlier on weakness in US high-volatile markets, though up $2 from the March quarter. Costs are the part that is working. "Second quarter mine cost at $99 a ton represented the fourth consecutive quarter of sub 100 cash costs," chairman and chief executive Randall Atkins told investors on 5 August. Roughly $17 a ton of margin does not cover the rest of the business: consensus has Ramaco losing $0.95 a share this year, losing $0.05 next year, and reaching a first profit of $0.58 in 2028 — on an estimate range running from -$1.23 to +$3.60, which is the Brook Mine question, not the coal one.

What the market pays for Wyoming

On 29 July Ramaco released an independent conceptual study by Hatch Associates superseding an earlier Fluor report. Running its own model on Hatch's capital and operating inputs, Ramaco produced a potential net present value of $8bn and average annual adjusted earnings before interest, taxes, depreciation and amortization of $1.3bn, against capital cost of $3.2bn plus $0.8bn of contingency and first production in 2031. Hatch has since been retained to lead a Preliminary Feasibility Study, with interim revised economics expected by year-end.

Against that, the Class A stock carries a market capitalization of $483m and the Class B shares — 63.68m outstanding — $297.4m. Trailing free cash flow yield is -28.1%: this equity funds a project rather than harvesting one.

The move was a minerals move

From 31 August to 28 September the Class B shares fell 41.6%. Over the same stretch MP Materials fell 15.8%, USA Rare Earth 19.2% and Energy Fuels 24.8%, as reports that a US–China summit could ease trade tensions deflated the scarcity premium that had powered the complex. A mechanical event compounded it: Ramaco said on 4 September that negative trading in its Class A stock the prior day traced to Yorktown Energy Partners distributing about one million Class A shares to its limited partners on 2 September.

The coal producers are the check. Australian premium hard coking coal was $266.50 a tonne on 28 September, down 5.5% on the month but 39.5% higher than a year earlier, after hitting a 17-month high — driven by Chinese supply, where safety checks across Shanxi after a fatal mine blast slowed restarts and pushed mills into the seaborne market. Warrior Met, which mines premium low-volatility coal in Alabama, sold a record 3.7m short tons in the June quarter at $138 a ton against cash cost near $93, generated $103.4m of free cash flow, and has finished Blue Creek at $1,022.9m with 2026 capital spending guided down to $130–150m. It trades at 14.56x forward earnings against 22.06x trailing. Alpha Metallurgical, with nineteen mines in Central Appalachia, realized $118.71 a ton against $103.07 of cost and raised its full-year cost guidance to $103.00–107.00 from $95.00–101.00; at 17.73x trailing enterprise value to EBITDA it is valued at 1.7 times Warrior's 10.52x on a fraction of the cash margin. Since 18 September Warrior has added 4.3% and Alpha 0.8%. Only Ramaco's Class B kept falling, another 12.2%.

So the September give-back in the two producers looks like the unwinding of an August squeeze the benchmark never endorsed — coal prices rose over the twelve months in which these equities were supposedly pricing collapse. At Ramaco the residual is different in kind. With coal steady and the tracking stock alone still sliding, what is being repriced is the critical-minerals option bolted onto a loss-making coal miner, and the discount to management's own $8bn — better than nine-tenths of it — is the market's price for a 2031 start date and a capital bill the coal book cannot fund.

The interim economics Hatch is due to deliver by year-end will be the first set of Brook Mine numbers Ramaco did not model itself.

A record price, fewer contracts

The price utilities actually sign for uranium reached a nominal record last month while the companies selling it spent the year being marked down. Under multi-year contracts, the long-term price closed August at $96.50 a pound, above the $95 peak set in 2007; spot sat at $89.30 on 28 September, roughly where it began 2026. Cameco, the largest listed producer, raised its 2026 realized-price guidance to C$91–96 a pound from C$85–89 and held production at 19.5–21.5 million pounds. Its shares are 35% below their 28 January high.

What has been marked down is time, not tonnage. The 30-year Treasury yield broke 5.5% on 25 September, the highest since 2004, with roughly $570bn of artificial-intelligence-related corporate bond issuance crowding the long end; a higher discount rate falls hardest on cash flows dated in the 2030s. Over twelve months the nuclear names ranked themselves by distance to cash flow: NuScale Power fell 79%, Oklo 67%, Centrus Energy 54%, Uranium Energy 33% — and Cameco, the only one with a delivery book, rose 1.7%. Two dated tests now land inside 48 hours.

Cameco: the hole is Westinghouse

Cameco's June quarter read like a collapse — net income down 92% to $25.2m — and none of it came from mining. Its 49% share of Westinghouse, the reactor-services business and AP1000 vendor it co-owns with Brookfield, swung to a $10m loss from $126m a year earlier, when the Dukovany two-reactor project in the Czech Republic contributed about US$170m. "Our year-to-date financial and operational performance reflects the value of aligning our marketing, operational and financial decisions with strengthening industry fundamentals," chief executive Tim Gitzel said on 31 July. The quarter's realized price of C$93.13 a pound — about US$68 converted — sits well under spot because the book was struck years ago with floors in the high $70s and ceilings near $160, and it lags in both directions.

The de-rating is real in dollars and illusory in multiples. Market capitalization has fallen from $50.3bn in May to $37.9bn, yet trailing price-to-gross-profit rose from 38.6x to 56.1x, because gross profit shrank faster than the price; forward earnings put the shares on 59.1x, with consensus 2026 profit essentially flat before a 72% step up in 2027. The unpriced item is Westinghouse itself, which confidentially filed a draft registration statement on 31 July.

Centrus: the customer left

Centrus, the only US-owned commercial enricher and supplier of separative work units to utilities in the US, Japan and Belgium, is where the business genuinely deteriorated. Separative-work volumes fell 23% year over year while enrichment pricing rose 3%; gross margin narrowed to 28.3% from 34.9% and operating income dropped 69%. Then the Department of Energy amended the high-assay low-enriched uranium contract on 30 June into a three-month, $15.0m option for cascade maintenance and storage with no production, expiring 30 September, requested no fiscal-2027 operating funds, and said it does not currently intend to exercise further options.

Five weeks after chief executive Amir Vexler called it "another strong quarter of financial and operational progress" on 5 August, Centrus priced $500m of shares and warrants at about $199.64. The stock closed at $140.37, leaving every warrant strike out of the money. A $4.5bn backlog running to 2040 sits behind a trailing price-to-earnings of 55.9x against 55.1x forward: no growth priced, consistent with consensus 2026 earnings 43% below last year's.

Uranium Energy: no revenue, by choice

Uranium Energy, the in-situ recovery producer that restarted Burke Hollow in South Texas in April, booked zero revenue in the quarter ended 30 April and a $52.3m loss. Fully unhedged, it withheld 1,456,000 pounds carried at $127m rather than clear into spot; its last sale moved 200,000 pounds at $101. With no earnings anchor, the usable readings are price-to-book of 3.18x and price-to-sales down from 373x in May to 226x. Full fiscal-2026 results come on 29 September.

What was actually taken back

Three different payment meters, three different verdicts. Cameco's earnings damage is a reactor-construction accounting swing and its cheapness is arithmetic, not value. Centrus's fall was written in Washington and confirmed by a dilutive raise — so the idea that an enrichment premium held while mined pounds fell is simply wrong, since Centrus is the second-worst performer of the three. Uranium Energy's decline has no fuel-market cause at all; the likelier reading is a discount rate that punishes waiting, and a market unwilling to fund a producer that will not sell. Constellation Energy, which mines nothing and enriches nothing, fell 12.9% over the same September sessions in which Cameco fell 13.6% and the S&P 500 fell 0.7%.

One number belongs to the fuel market rather than the bond market: reported term contracting through 31 August exceeded 38 million pounds, about 15% below last year, at the highest long-term price ever printed. Utilities will pay the record. They still will not be the first to sign.

Whirlpool spent last year telling investors that a steel tariff had finally arranged the American appliance market in its favor. It has instead become the one large appliance maker in the country paying that tariff on every unit it builds at home.

Section 232 of the Trade Expansion Act taxes steel and goods derived from it. In June 2025 the Commerce Department added eight household-appliance categories to the derivative list — combined refrigerator-freezers, washers, dryers, freezers, dishwashers, ranges and ovens, food waste disposals — at 50% on the steel content. A proclamation effective 6 April 2026 went further, applying the duty to the full customs value of covered articles rather than to metal content, at 50% for primary metal articles and 25% for derivatives substantially made of steel, aluminum or copper. The asymmetry is structural. A company that manufactures domestically buys the taxed input; the protective half of the bargain only reaches competitors whose finished goods actually cross the border and land in a covered annex.

Whirlpool — refrigeration, laundry, cooking and dishwashers sold to retailers and builders under the Whirlpool, Maytag and Speed Queen names among others — sits on the wrong side of that. Chief executive Marc Bitzer called the new regime "finally the environment which allows a level playing field" and said footprint optimization and the Section 232 update "meaningfully strengthened our competitive advantage as a domestic producer," arguing that Asian rivals' access to tariff-free Chinese steel handed them roughly $70 of cost advantage per product. The same company then guided to a $225m tariff cost for the year despite that domestic base.

The importer's bill went the other way

The duties that were taxing the importer are gone. On 20 February 2026 the Supreme Court ruled 6-3 that the International Emergency Economic Powers Act does not authorize the President to impose tariffs, voiding the reciprocal rates. A flat 10% surcharge under Section 122 of the Trade Act of 1974 replaced them within hours and lapsed on 24 July, giving way to 12.5% Section 301 rates; Vietnam went from 46% to 12.5%, Thailand from 36% to 12.5%.

SharkNinja — Shark vacuums and steam mops, Ninja air fryers, blenders and coffee systems, sold through retail, e-commerce and direct — had already moved nearly all production out of China to Vietnam, Thailand, Indonesia, Malaysia and Cambodia. Its 2026 outlook now assumes minimum rates of 10% to 12.5%, and it filed $247.1m of refund claims that customs accepted in July, to be booked as a reduction of cost of sales in the third quarter.

The two books

Whirlpool's June-quarter revenue fell 6.8% to $3.52bn while gross profit fell 26.8%, four times the rate of the sales decline — fixed-cost deleverage. Gross margin compressed 343 basis points to 12.6%. In North America, segment operating margin was 2.7% against 5.9% a year earlier on $2.4bn of sales, with US industry demand down 3.4%. Pricing has not yet reached the per-unit economics: a 10% promotional price increase in April, its largest in a decade, was followed by 4% on list in July. The demand backdrop is a frozen resale market — existing-home sales ran at a 3.98m annual rate in August, down 1.2% year on year — which strips out the trade-up replacement that carries the mix and leaves lower-margin break-fix demand.

The balance sheet is the louder problem. The dividend was suspended in May to fund more than $900m of debt paydown. A rescue equity raise of roughly $800m priced common stock at $69.00, against $31.70 today, lifting weighted-average diluted shares toward 71.3m from 57.4m. In June the company swapped roughly 1%-coupon paper for $1.0bn of 7.500% and $1.0bn of 7.875% second-lien secured notes, and 2026 interest expense guidance rose to $350m from $300m. Guidance for 2026 ongoing earnings has been cut four times, from about $7.00 to $2.50-$3.00; consensus sits at $1.83 against $5.66 actually earned in 2025. That is why the forward earnings multiple of 17.2x is above the 10.6x trailing figure — the denominator collapsed faster than the price. Enterprise value is 9.2x trailing earnings before interest, taxes, depreciation and amortization, a figure consensus expects to fall 26.8% this year.

SharkNinja's June quarter grew 22.2% to $1.77bn. "Q2 was a standout performance for SharkNinja, with net sales growth accelerating to 22.2%, our fastest pace since 2024," chief executive Mark Barrocas said in the 5 August release. Adjusted gross margin still slipped about 70 basis points to 48.7% on US tariff cost, currency and retailer activations; operating income rose only 6.4% and net income fell 7%. The fastest-growing part of the book sits beyond US trade policy entirely: international sales rose 36.6%, against 15.5% at home.

The verdict

Incidence flipped by court ruling rather than by pass-through, and the market has priced that faster than either management has narrated it. Whirlpool's de-rating is earned, but only partly by steel: a shrinking industry, a $50m step-up in interest and a dilutive rescue equity explain more of it than the tariff line does. SharkNinja's advance is genuine volume and geography — yet the apparent discount in its forward multiple is not. The refund is worth roughly $1.75 a share before tax on 141.5m diluted shares, about $1.31 after; strip it from the $6.45-$6.55 adjusted guidance and the forward multiple sits near 34x, barely below 36.9x trailing. Roughly half those duties were expensed in 2025, so third-quarter gross margin will carry a credit for last year's costs.

Both prints are close. Whirlpool's third quarter is most commonly dated 26 October, SharkNinja's 5 November. The settling lines are narrow: whether North America margin climbs back toward last year's 5.9% now that both price increases are in the base, and what SharkNinja's gross margin looks like with the refund taken out.

Two Septembers ago these two shares closed within a dime of each other. Section 232 was meant to make the American plant the cheap place to build an appliance; so far it has mostly made steel more expensive there.

On 24 September Starbucks said it would close about 250 underperforming cafes in North America, from a base of more than 18,000, and take roughly $300m of restructuring charges — around $200m for early lease exits and separation pay, $100m in non-cash asset disposal and impairment. The same announcement cut the fiscal-2026 net new store target to about 440 from 600–650, with all of the net openings international. The company said the closing stores were ones where it would be "unable to create the physical environment our customers and partners expect" or saw no path to financial performance.

The home market is therefore being made smaller in the year its sales line is running strongest since the slump — and that is the trade underneath chairman and chief executive Brian Niccol's turnaround. Comparable sales are being bought with labour hours, remodels and now lease terminations, while the shares change hands at about 36 times the fiscal-2026 consensus of $2.61 a share, a figure still 21% below the $3.31 Starbucks earned in fiscal 2024.

The comp is real, and it has a price

In the June quarter North America comparable sales rose 8.1%, led by a 4.5% gain in transactions against 3.5% in average ticket — more visits, at higher prices, simultaneously, which is the harder combination. North America segment operating margin widened to 13.6% from 13.3%, credited to sales leverage, lapping the prior-year Leadership Experience event, and lower inflation paired with tariff refunds. "Our Back to Starbucks plan was built on the belief that an extraordinary cup of coffee, human connection and customer experience win the day, every day. Our third quarter results are proof they do," Niccol said in the 29 July results release. Full-year adjusted earnings guidance went up that day, to $2.55–$2.65 from $2.25–$2.45.

The cost side is disclosed too: Starbucks has committed over $500m of incremental labour spending across twelve months, a figure RBC said exceeded its modelling when it cut the stock to Sector Perform in March. And the reported top line no longer describes the same company: consolidated revenue fell 1.4% to $9.32bn even with comps up, because China was deconsolidated after Boyu Capital bought 60% of the retail operations in a $4bn deal that closed in April, moving 7,991 coffeehouses to a licensed model with Starbucks keeping 40% and the brand.

Not a coffee problem

The shares are down 11.9% over thirty days to $94.98, and they fell 0.5% on the closure announcement itself — the de-rating preceded the news. Starbucks is also mid-pack: over the same window Cava fell 22.9%, Texas Roadhouse and Shake Shack 20.5% each, Domino's 16.9%, Chipotle 16.8% and McDonald's 10.8%, against 0.6% for the S&P 500 tracker. Placer.ai counted August restaurant visits down 2.4% year on year while total retail visits rose 0.3%, with gasoline above $4 a gallon and food-away-from-home prices up 3.4% — though a Labor Day shift out of the month flatters the decline. The bean is not the culprit: arabica closed at $2.71 a pound on 23 September, down 28.2% in thirty days from a 52-week high of $4.23, the 50% tariff on Brazilian coffee has been zero since November 2025, and packaged-coffee names barely moved, Keurig Dr Pepper down 3.3%.

The one that fell twice as far

Dutch Bros, the drive-thru beverage chain with 1,225 shops, fell 24.7% over the same thirty days and 47.7% over three months. Its June quarter grew revenue 32.5% to $550.9m, posted a thirteenth straight positive same-shop quarter at 5.8%, transactions up 3.4%, and a record $2.2m systemwide average unit volume, and raised full-year guidance to $2.10–$2.13bn of revenue and $385–$390m of adjusted earnings before interest, taxes, depreciation and amortization. The soft edge was third-quarter comp guidance of 4% to 5%, which cost the shares 18.4% in a single session on 6 August, and a beverage, food and packaging line up 80 basis points to 26.1% of shop revenue on higher coffee costs — 2025's record beans reaching the profit and loss now, with more pressure guided for the second half.

The verdict

A sector-wide traffic repricing hit both, but only one of them was carrying a recovery premium into it. Starbucks at 54.8 times trailing earnings and 21.7 times enterprise value to trailing earnings before interest, taxes, depreciation and amortization, against McDonald's at 19.2 times trailing earnings and 14.8 times, is being asked to pay for margins it has not rebuilt — consensus has revenue up 2.8% this fiscal year and 1.6% next, so the whole case is margin. That part of the fall the multiple earns. Dutch Bros is the harder one: its forward multiple has compressed from about 74.9 times the 2026 consensus in late June to 39.4 times now while that consensus rose, leaving 39.4 times against expected earnings growth of 39% — roughly one times growth, where McDonald's sits near three. Its weak point is cash, not demand: a trailing free-cash-flow yield of 1.46% against Starbucks' 3.35%, with 185-plus shops a year to fund at about $1.4m each.

Closing 250 cafes to make the remaining ones work is a coherent plan, and it is the second time in two years Starbucks has paid to shrink toward it — last September's round cost about $1bn. The fiscal fourth quarter will be the first to carry both the new charge and a North America store count moving the other way from its comp.

Both of the listed companies that sell mainly to American police departments raised their guidance in August, and both have been marked down hard since. Axon Enterprise, which sells TASER conducted-energy weapons, body cameras and the Axon Evidence cloud that stores the footage, has lost 40.1% of its value over twelve months. Cadre Holdings, which makes Safariland body armor, duty gear and Med-Eng bomb suits for police, corrections and federal agencies, is down 28.3%.

The tidy explanation is their shared paymaster. Neither company has a commercial customer of consequence; every dollar either earns is appropriated by a city council, a statehouse or Congress. If that dollar is tightening, both equities deserve less regardless of what they shipped last quarter. So the question is whether the squeeze is visible where it would appear first — in orders.

The buyer is genuinely tight

The National Association of State Budget Officers' spring survey put median general-fund spending growth across the states at 0.6% for fiscal 2027 against 2.5% median revenue growth, with thirteen states reporting $26.3bn of projected gaps before balancing measures. Federally, the fiscal 2027 request would cut the Office of Justice Programs' State and Local Law Enforcement Assistance account to $1.619bn, $781m below fiscal 2026, trim Byrne Justice Assistance Grant formula money by 9% and cut the COPS school violence prevention program by a third — the exact channels through which a small department buys vests and camera subscriptions. Cutting the other way, the Justice Department is readying up to $3.5bn in law-enforcement grants. The money is being reshuffled more than withdrawn.

Neither order book shows it

Axon is paid on ten-year bundled Officer Safety Plans, so a budget shock reaches its revenue line late and its bookings early. Bookings accelerated. "Future contracted bookings grew more than 40% year over year to $15.1 billion, reflecting the broad-based momentum across products and end markets," chief executive Rick Smith told investors on August 5. Annual recurring revenue reached $1.6bn, up 39%, with net revenue retention of 126%. Second-quarter revenue of $904.4m grew 35.3%, a tenth consecutive quarter above 30%, and Axon raised full-year guidance to growth of 32-34% from 30-32%.

Cadre sells the least discretionary thing in the ledger — a vest replaced on a cycle — and its backlog reached a record $368m, a second straight record, helped by a $61m indefinite-delivery contract with the Federal Bureau of Investigation for ballistic panels and a $50m Med-Eng award. Guidance went up too, to $749m-$769m of sales from $736m-$758m, and the $0.10 quarterly dividend was maintained. The nearest thing to a warning from either company was Cadre's call commentary about a growing focus on balancing state and local budgets.

Where the two part company

Cadre's de-rating is written by its own income statement. Only 5% of its 32% second-quarter growth was organic, the rest bought with TYR Tactical, closed in January; full-year organic growth is guided to 3-5%. Net income of $11.4m fell 6.6% on that 31.8% revenue gain, after a first quarter in which gross margin dropped to 37.7% from 42.4%. Consensus has 2026 earnings of $1.00 a share against $1.02 last year — 21% more revenue, no more profit, as acquisition debt and intangible amortization absorb it. Bank of America moved to Underperform on August 13 with a $26 target; the shares closed at $26.06 on Monday. At 26.0x forward, against roughly 36x a year ago and 16.2x trailing enterprise value to EBITDA, the reset matches the arithmetic; a 7.0% trailing free-cash-flow yield is the argument against it.

Axon's month was two sessions and two events. The shares fell 11.4% on September 1 and 9.8% on September 15, the day it announced a zero-coupon convertible that priced at $1.15bn including the over-allotment, due 2031, convertible near $652 against a stock now around $425. Two days after a September 23 Senate Judiciary subcommittee hearing on automatic license-plate-reader surveillance — Axon was invited, did not appear, and was told subpoenas were possible — Goldman Sachs cut its target to $675. Axon now trades at 55.1x forward earnings against 171.1x trailing and a reported five-year average forward multiple near 78x; Motorola Solutions, the closest public-safety comparable, is at 25.5x forward on 13.3% growth, which makes Axon the cheaper of the two per point of growth. An 80.6x trailing EV/EBITDA and a 0.39% free-cash-flow yield are what the bear owns.

So the shared-customer story is real in statehouses and invisible in both order books, and the two stocks fell for unrelated reasons: Cadre because a consumable business bought its growth and is paying for it, Axon because a very expensive multiple met dilution and a subpoena threat in the same month. An appropriation deferred one year reappears the next. A hearing that decides what a camera network may record does not, and Axon has not yet sat down at it.

Expedia Group agreed on 22 September to sell hotel rooms inside Muse, the personal AI agent Meta launched in the United States on 8 September that can search, compare and complete a booking by itself. Expedia remains the merchant of record and the payment settles inside the assistant. Investors read the arrangement as a loss for the seller rather than a win: Expedia closed 5.5% lower that session and Booking Holdings 4.2%, with travel intermediaries selling off again the following day as Booking gave up a further 3.4%.

An online travel agency is not paid for travel demand. It buys demand — largely from search engines it cannot price — and then levies a take rate on the gross bookings it collects. The question a four-week drawdown of roughly a fifth at both Booking and Expedia asks, against a broad market that was flat, is whether an agent owning the traveler's first question permanently raises the cost of that demand. It has been asked without new data: neither company has reported since early August, and the hotel market itself is firming, with CoStar and Tourism Economics raising their 2026 US revenue-per-available-room forecast to 4.4% from 2.8%.

The counterparty is the one accelerating

Expedia, which runs Brand Expedia, Hotels.com and Vrbo alongside a white-label supply business selling travel inventory to airlines and banks and the Trivago metasearch site, has now accelerated for four straight quarters: June-quarter revenue rose 14% to $4.315bn, and operating margin reached 23.9% against 14.0% a year earlier. The growth is coming from other people's traffic. Business-to-business revenue rose 23% and its gross bookings 21%, against 8% for the consumer brands, while Trivago advertising revenue grew 48% — and full-year revenue guidance went up, to 9–10%, on 5 August. Chief executive Ariane Gorin told the Skift Global Forum on 23 September that Expedia has scrapped its single all-purpose agent for narrower "point agents" and will judge agent partnerships case by case; she also disclosed that those tools have not lifted bookings yet, though they yield more than 60% more data on traveler intent. Over twelve months Expedia's shares are up 18.6%.

Booking bought more traffic to grow bookings less

Booking Holdings, which owns Booking.com, Agoda, Priceline, KAYAK and OpenTable, gives the bears something real. Revenue growth halved to 8.1% in the June quarter, from 16.2% in the March quarter. Gross bookings of $51.0bn produced $7.352bn of revenue — a take rate of 14.4%, down from roughly 14.5% — while marketing spend rose 11% to $2.37bn, faster than bookings at 9% and room nights at 5%. Management attributed the overshoot to "changes in traffic mix, incremental investments in paid marketing at attractive ROIs, and a shift of merchandising spend to performance marketing". Roughly a third of its demand arrives through channels it pays for; direct booking runs in the mid-60% range, according to Morgan Stanley's Matthew Cost, who initiated coverage at Overweight with a $230 target on 16 September on the argument that AI is a new acquisition channel. Against that: operating margin improved to 34.0% from 33.1%, full-year guidance is untouched at about 9% constant-currency growth, and the company returned a record $4.1bn in the quarter. It also lost its route into flights by acquisition permanently when the EU General Court upheld the veto of the eTraveli deal on 9 September. Booking now trades at 15.7 times forward earnings against a ten-year average CNBC puts at 22.3 times; Expedia is cheaper still at 12.7 times, on 7.5 times trailing earnings before interest, taxes, depreciation and amortization.

Trip.com broke, and search had nothing to do with it

Trip.com, the Ctrip, Qunar and Skyscanner operator, is the member with no Google dependence and the worst numbers. Revenue growth decelerated to 5.5% in the June quarter and the period produced an operating loss of RMB1.462bn — because China's market regulator imposed a RMB5.18bn ($765m) penalty for price-parity terms forced on hotels. Underneath it, international agency revenue grew 50%. "AI-assisted orders through TripGenie on Trip.com increased by approximately 400% year-over-year and nearly 60% of TripGenie interactions are now booking related," executive chairman James Liang said on the 15 September call. The shares sit at 1.04 times book and under five times trailing operating cash earnings, with RMB100.5bn ($14.8bn) of cash against a $25.9bn market value — cheap because the fine is real and the domestic partner transition is unfinished, not because an agent took the customer.

What the selloff earns

Only one of the three causes is shared, and it is the one supported by no company data: the agent risk. Booking earns part of its markdown — a take rate that slipped while the traffic bill outgrew the bookings is the mechanism bears describe, arriving early. Expedia's is the harder to justify, since it is the counterparty to the deal that triggered the selling and raised guidance during the slide. Trip.com's markdown is earned by a regulator in Beijing. Treating these as one exposure to agentic search is the error the September tape made.

Booking reports on 27 October and Expedia two days later. If an AI agent is raising the price of finding a traveler, it will show up as a marketing line before it shows up in a headline — and nobody has seen one since 4 August.

The instruments that grasp, cut and suture inside a patient on a da Vinci robot run on a counter. When the counter is exhausted the hospital buys another set, and that recurring purchase — not the robot — is where Intuitive Surgical, which sells the da Vinci surgical system, the Ion lung-biopsy platform and everything consumed on both, earns the larger part of its revenue. In the first half of 2027 the company will slow the counter down.

Intuitive will introduce a subset of EndoWrist instruments for its fourth- and fifth-generation robots with increased useful lives and lower customer cost per use, aimed at benign procedures where price decides whether an operation is done robotically at all. It has done this before: in the comparable 2020 extended-use cycle, instrument-and-accessory revenue per procedure took a hit of roughly 7%. The size of this one is not public. Asked by Wolfe Research's Mike Polark whether pricier da Vinci 5 and force-feedback instruments could hold revenue per procedure flat, chief financial officer Jamie Samath declined to quantify it on the 16 July call: "When there is an opportunity to drive incremental growth by reducing customer costs through innovation, we are willing to do that." The figure is promised for the third-quarter report on 20 October.

The numbers got better

Everything else points the other way. June-quarter revenue rose 19% to $2.89bn — 18.5% precisely — with worldwide procedures up about 16%. Gross margin widened to 67.8% from 66.3% a year earlier and operating margin to 33.6% from 30.5%. Full-year da Vinci procedure guidance was reaffirmed at 13.5% to 15.5%, and non-GAAP gross-margin guidance was raised to 68–69% after a $36m refund of tariffs already paid. Instrument-and-accessory revenue grew 18% — still slightly ahead of volume.

The soft spot is American. US da Vinci procedure growth slowed to 12%, which management ties to the 1 January expiry of enhanced Affordable Care Act premium tax credits; marketplace premiums for roughly 22 million enrollees more than doubled on average, deferring elective benign cases while urgent after-hours procedures grew 26%. The attribution is contested: BTIG's annual hospital capital-equipment survey found most executives had seen no impact on procedure volumes.

The shares did not walk down, they gapped: a 14.1% fall on 17 July, the session after the print, a low of $332.02 on 23 July, then a 25% recovery helped by an Oppenheimer upgrade to Outperform with a $500 target. Intuitive still sits 26.2% below where it began the year. It trades at 46.9x trailing earnings against roughly 58x twelve months ago, while trailing earnings per share grew 16.3%; on gross profit, 19.9x against about 25x, with gross profit up 15.3%. Forward, 38.4x is not cheap against consensus earnings growth decelerating to 11.9% in 2027.

Some of the rating change lives outside the income statement. On 22 July the Food and Drug Administration granted Johnson & Johnson authorization for Ottava, the first soft-tissue robot with arms integrated into the operating table, six days after Intuitive's print and months after Medtronic's Hugo cleared. In China, domestic laparoscopic robots outsold imports through public tenders for the first time in 2025.

Where the meter actually broke

Procept BioRobotics, which sells the HYDROS robot and the single-use handpiece that performs Aquablation therapy on enlarged prostates, shows what a broken consumable meter looks like. On 25 February it disclosed that handpiece sales had exceeded procedures performed by 8% to 16% in every quarter since early 2023, leaving more than 10,000 units of excess field inventory, and scrapped the quarter-end discount program that encouraged the stockpiling. US handpiece units fell from 13,225 to 9,400 in a single quarter, and full-year revenue guidance came down to $390m–$410m from $410m–$430m. Price to trailing gross profit has compressed from about 10.7x to 4.6x. The underlying franchise is not dead — US procedures topped 13,100 in the June quarter, up 21%, on an installed base of 816 systems — but with roughly $41m of quarterly cash burn against about $194m of net cash, the promised fourth-quarter adjusted-EBITDA inflection is not optional. Chief executive Larry Wood could not fully account for the weakness in older accounts on the 4 August call: "Aquabeam, it's an older system now. It doesn't offer all the features and benefits of Hydros."

The cheap-console threat is real on price and small on scale. SS Innovations' SSi Mantra sells in India for roughly a third of a da Vinci's price, and its installed base reached 238 systems across twelve countries by early September with revenue up 39%. But gross margin fell 8.2 points to 50.9%, its FDA submission is not expected to clear before the first quarter of 2027, and at 23.5x trailing gross profit it is priced above the incumbent it is meant to undercut.

The verdict

One label covers three different stories, and only one is about demand for robotic surgery. Procept's de-rating is earned — it sold consumables into storerooms and is now working through them. SS Innovations' is dilution and the margin cost of a one-third price point. Intuitive's is the hard case, because the operating numbers improved while the multiple contracted by a fifth. What the current results cannot settle is what a trained surgeon's console is worth now that four rivals can be bought in the United States, and now that Intuitive has decided the charge per operation should come down.

Extending instrument life is the move of a company confident enough in volume to give up price. On 20 October it has to say how much price.