DK Street Journal

Albemarle's battery-materials business earned $723m of adjusted segment profit in the June quarter, more than three times what it made a year earlier. Since August 21 the shares have fallen 22.6%, and nothing the company has published explains it.

What repriced was not this year's earnings but next year's balance. Mineral Resources restarted the Bald Hill mine and Core Lithium restarted Finniss, Chinese futures peaked at RMB162,500 a tonne (about $24,000) during 24-27 August and then slid on firming expectations that Jiangxi mines restart, and lithium futures went on to a five-month low as traders handicapped a 2027 glut. That matters to these two specifically because neither is paid the spot price. About 40% of Albemarle's estimated 2026 salts volumes sell under index-referenced contracts carrying floors, sometimes ceilings, and a three-month price lag — so a September break in the index reaches the income statement in the fourth quarter. The equity is marking a realised price nobody has reported yet.

What a lithium dollar actually is at Albemarle

Albemarle, the Charlotte specialty chemicals group that sells lithium carbonate, hydroxide and chloride into battery supply chains alongside bromine flame retardants and refining catalysts, shipped 65,000 tonnes of lithium carbonate equivalent at $19.53 per kilogram in the June quarter, volumes up 11% and price up 60%. Management put that realised figure roughly 15% below prevailing market pricing, the gap being contract lag plus the dilutive effect of selling spodumene concentrate beside salts. Revenue rose 31% to $1.74bn, gross margin reached 33.9% against 14.8%, and the full-year Specialties outlook was raised. "We continue to see resilient demand fundamentals across our core markets, including energy storage, electric vehicles, and semiconductors," chief executive Kent Masters said in the August 6 release.

The balance sheet behind that is already defensive. First-half capital expenditure was $170m, down $132m year on year — cut before the rally and not restored. The mandatory convertible preferred that raised about $2.3bn at a 7.25% coupon converts on March 1, 2027, and diluted share count has already climbed from 117.7m to 136.2m as earnings turned positive. Albemarle owns 49% of Talison, operator of Greenbushes, where a June 9 fire pushed the third chemical-grade plant's ramp into the first quarter of 2027 while concentrate guidance held. Its conversion spread is squeezed from both ends when the index falls: feedstock it buys from a venture it minority-owns, plants whose unit cost is set by utilisation.

SQM's cushion, and the landlord

SQM, the Santiago brine producer of lithium, iodine and potassium nitrate fertilizers, had a bigger quarter still: revenue up 137% to $2.47bn, gross margin 51.1% against 24.3%, lithium realised near $21.80 per kilogram. It raised its 2026 global demand forecast by 200,000 tonnes on August 19 rather than trimming it. But lithium is 78% of consolidated first-half gross profit against iodine's 16% — the iodine book, at a record $73.4 per kilogram, is a floor rather than the bulk of the business.

Chile takes a widening cut on the way up. The 2018 lease with development agency CORFO carries a progressive royalty rising to a marginal 40% above $10,000 a tonne, and payments to the state exceeded $1.6bn in the first half against $4.23bn of revenue. Under the Codelco partnership, the state's share of operating margin on new production runs near 70% to 2030 and 85% thereafter. A schedule that escalates with price also relaxes when price falls — an automatic stabiliser Albemarle does not have.

The verdict

The physical market moved far less than the equities. Cochilco's August average of $19,525 a tonne is 12.2% off the May peak but still more than double the 2025 average; a single-digit dip in Chinese carbonate in mid-September produced a 5.68% fall in SQM in one session. Nothing in either reported quarter earns that. What the selloff does earn is the arithmetic ahead: lagged contracts guarantee fourth-quarter realised prices follow the index down even if spot holds, and consensus has Albemarle at $11.48 a share this year and $9.81 next, so its 9.9x forward multiple is struck on a denominator analysts expect to be a cycle top. SQM's forward estimates are flat to slightly higher into 2028 at a similar multiple. The pair is not one trade. One is a converter with dilution arriving in March; the other is the low-cost brine producer whose landlord absorbs part of the fall.

SQM's own management supplied the bear case nobody needed a futures screen for: battery-storage cells are being produced faster than they are deployed, and a 12- to 18-month oversupply is possible if approvals catch up too slowly. The next print is the first to contain any of September, and the only line on it that settles this is the price per kilogram Albemarle actually collected.

Twilio has disclosed nothing about its business since August 6. Its valuation has moved regardless: the shares closed at $258.27 on September 21 against $225.30 a month earlier, about 15% higher, and now carry 14.48x trailing gross profit against 12.45x in late August and roughly 7.2x in February. No new financial information arrived in between.

That matters because the company's own last word pointed the other way. Twilio, which sells cloud communications interfaces that let developers embed messaging, voice and email into their software and bills per message and per minute rather than per seat, guided third-quarter revenue to $1.505-1.515bn — 11-12% organic growth, after posting 17% in the June quarter.

The dollar that isn't kept

Twilio's reported revenue grew 22.0% in the second quarter to $1,499.1m. Gross profit grew 20.4%, and gross margin slipped to 48.42% from 49.07%. The gap is the carrier fee. United States operators raised application-to-person messaging pass-through charges three times in 2026 — T-Mobile and US Cellular in January, AT&T in April, Verizon in May — and both companies bill those charges on at no markup. Revenue rises; gross profit does not. Twilio is absorbing roughly $190m of incremental carrier fees this year, a drag of about 170 basis points on gross margin, while per-message prices fall 5-8% a year on their own.

Strip the surcharge and the business is still improving. Dollar-based net expansion reached 116%, from around 108% a year earlier; operating income more than doubled to $84.5m; free cash flow rose 34% to $353m. "We are in a powerful new chapter at Twilio, marked by another quarter of organic growth acceleration as well as record profitability and free cash flow," chief executive Khozema Shipchandler said with the results. On the call he called the artificial-intelligence opportunity "very early innings", with under 6% of voice interactions machine-driven. The reported earnings line is no help in judging any of it: second-quarter net income of $1,067.2m against $84.5m of operating income is a one-off tax item, which is why gross profit and the 43.5x forward earnings multiple are the anchors that work.

Owning the network did not help

Bandwidth, a Raleigh company that routes voice and messaging over an internet-protocol network it owns and licenses market by market, was supposed to be the structural alternative to reselling someone else's routes. It grew revenue 22% to $220m — and gross profit 9.6%, with gross margin down more than four points to 35.72%. Some $68m of the quarter was messaging surcharge; the cloud communications line it strikes net of pass-through grew 12%. "AI is changing how enterprises buy communications infrastructure," chief executive David Morken told investors on July 29. The shares fell 29.3% that session and sit 31.3% below their July 9 peak, at 5.63x trailing gross profit against roughly 8.2x at that high. Bandwidth funded the network with $275m of zero-coupon convertible notes in June and still runs an operating loss.

Both stocks rose sharply on September 21 — Twilio 5.8%, Bandwidth 7.3% — with no company announcement discoverable for either; the likelier reading is a shared sector move, and the whole per-minute complex has been marked up, RingCentral included.

The verdict

The hypothesis that owned infrastructure earns better economics than resale is inverted by these numbers: the asset-light reseller held its gross margin within a point while the network owner gave up four. What separates the two is not the rail but the mix — Twilio's improvement reaches the gross profit line, Bandwidth's does not. Twilio's business has earned a re-rating; September's leg is not part of it, arriving as management guides growth down by five points.

The surcharges lap in 2027, and when they do the reported growth rate falls back to whatever customers are actually paying for. Consensus already assumes it: 11.4% revenue growth next year for Twilio, 4.5% for Bandwidth.

Formula One held five races in its June quarter against nine in the same quarter a year earlier, and the four missing weekends took contracted money with them. Liberty Media, whose Formula One Group tracking stock holds the exclusive commercial rights to the world championship, reported revenue down 30.4% to $934m, operating income down 68.6% and net income of $5m. The company attributed the fall in part to "a decline in contractual media rights revenue and a decline in title sponsorship revenue related to event mix".

That sentence is the mechanism the whole heading turns on. Sports media rights are treated as the safest cash flow in entertainment — negotiated years ahead, escalating on a schedule fixed at signature — but they are recognized per event, so no event means no fee. The Bahrain and Saudi Arabian Grands Prix were cancelled because of the war in the Middle East, cutting the 2026 calendar from 24 races to 22; Liberty's accounts now assume 23 after an October date at Sepang partially replaced Bahrain, and it expects a full calendar in 2027.

Everything else at F1 went up

The per-race meters did not deteriorate. Nearly 3.7m people attended race weekends this season against 3.4m for the same events last year, every 2026 race sold out, and five circuits set attendance records, Silverstone drawing 564,000 over four days. The US media deal repriced upward by two-thirds at renewal: Apple pays roughly $150m a year under a five-year exclusive streaming agreement beginning this year, against the $90m ESPN had been paying. "Demand for our brands remains robust and resilient," chief executive Derek Chang said with the August 6 results.

The equity has nonetheless de-rated, and honestly: the tracking stock's forward price-to-earnings multiple of 53.1x sits above its trailing 47.7x, meaning consensus earnings are falling faster than the share price. Analysts see recovery in 2027 — revenue of $5.28bn, up 10.8% — which is what a race-count problem should look like.

TKO's contracts stepped up; its shares did not

TKO Group Holdings, the New York parent of UFC and WWE, is the same asset class billed differently. Its June-quarter revenue rose 18.2% to $1,547.1m, with gross margin widening to 64.1% from 56.0% — contracted fees landing against a largely fixed cost of producing the slate. UFC media-rights revenue was $325.2m as the Paramount agreement began in January, and WWE's rose 29% to $359.7m on the ESPN deal that started in September 2025. The Paramount contract is $7.7bn over seven years, back-weighted so the escalator is written in at signature. Adjusted EBITDA rose 23% to $649.9m and full-year guidance went up to $2.275–$2.305bn.

"Sports has become the anchor of premium media, commanding unrivaled live audiences and cultural relevance," president Mark Shapiro said on the August 3 call. The shares fell 6.7% over the past twelve months anyway. The available explanation is arithmetic: consensus has revenue growing 23.3% this year and 1.0% next, as the two step-ups annualize. TKO trades at 41.3x forward earnings against 63.5x trailing — the compression is the doubling in expected earnings, and the risk is what happens after.

Paid by the ticket instead

Live Nation Entertainment owns almost none of what it sells. It promotes, operates more than 250 venues, and collects a fee per fee-bearing ticket through Ticketmaster, which holds roughly 86% of ticketing at major concert venues against about 9% for AEG's AXS. Every meter rose in the June quarter: 90m fee-bearing tickets, up 8%, gross transaction value above $10bn, event-related deferred revenue at a record $6.4bn and sponsorship revenue of $383m. "We've seen no consumer issues to date," chief executive Michael Rapino told analysts on July 30.

What the meters cannot settle is the April 15 jury verdict, in which a federal jury found Live Nation and Ticketmaster liable on every antitrust count and set damages at $1.72 for each primary ticket sold pursuant to the conduct. The remedy phase — divestiture, structural limits or behavioral conditions — remains undecided, and the shares fell 8.1% in the past month. On negative GAAP earnings the only usable anchor is 24.3x trailing enterprise value to EBITDA, which consensus has collapsing toward the mid-teens by 2027.

The tier nobody watches is deflating

Madison Square Garden Sports, which owns the Knicks and Rangers and nothing else, is the useful control. Its contracted local media-rights revenue fell to $125.7m in fiscal 2026 from $157.4m after MSG Networks restructured its debt and the rights agreement was renegotiated downward. The stock rose 85.2% over the year regardless, on franchise scarcity and a Rangers spin-off. Sports rights are not one market: the national streaming tier is inflating while the regional tier writes itself down.

So the segment's 17% twelve-month gain was never a rights trade — it was the team equities, MSGS alongside the Braves and Manchester United lines, while TKO, Formula One and Live Nation averaged a 4.7% loss against the S&P 500's 15% gain. Nothing is being un-priced that was priced in the first place. F1's fade is earned by a calendar; TKO's is an annualization cliff sitting behind rising cash flow; Live Nation's is a number a jury wrote. The one thing they share is duration — escalators running into the next decade, race contracts, franchise terminal values — and the 10-year Treasury yield touched 5.011% on September 14, its highest since 2023, discounting all of it harder.

The war that pushed those yields up is the same war that deleted two race fees. It is rare to watch a macro shock arrive twice at the same income statement, once through the discount rate and once through the calendar.

The price utilities pay for uranium under multi-year contracts reached its highest level since 2008 this year, around $90 a pound, according to Sprott, and Cameco — the Saskatoon miner that sells concentrate to nuclear utilities across three continents — raised the price it expects to collect. Its shares still fell about a tenth over the thirty sessions to 21 September, to roughly $92.

What moved was not uranium. Between 8 and 18 September, Constellation Energy fell 15.4% and Talen 10.5%, while an S&P 500 tracker slipped 0.8%; reporting on Constellation's week attributed the move to rising long-term Treasury yields and investors stepping back from the artificial-intelligence buildout trade. Nuclear fuel names were repriced alongside the electricity generators they are assumed to supply — which matters because a fuel dollar is earned nothing like a megawatt-hour is.

How the pounds are actually priced

Cameco does not get paid the spot price. It delivers under utility contracts struck years earlier, and on the July 31 second-quarter call management described the shape of them: floors in the high $70s a pound and ceilings at $160, escalated. Those terms cap participation in a spike and cushion a fall. The company has commitments for average annual deliveries of more than 28 million pounds of uranium oxide over five years, and second-quarter realized price came in at C$91.30 (US$65.53) a pound, up 12% year on year. In the same release it raised 2026 realized-price guidance to C$91.00–96.00 from C$85.00–89.00, with unit cost of sales also lifted, to C$63.00–67.50 from C$61.50–65.00, and production held at 19.5–21.5 million pounds.

The reported quarter was ugly regardless: revenue down 7.2% to $814.1m, gross margin from 29.3% to 21.1%, net income down 92% to $25.2m as the 49%-owned Westinghouse swung to a $10m loss from $126m of earnings, the prior year having carried roughly $170m from the Dukovany reactor project in the Czech Republic. Chief executive Tim Gitzel framed the volume shortfall as deliberate: the company is "not prepared to dilute the value of our assets by committing supply into contracts that do not appropriately reflect the durability of market fundamentals," he told investors on July 31. Westinghouse also sits behind a conditional Department of Energy commitment of up to $17.5bn in loans for long-lead items on as many as ten AP1000 reactors — conditional, and a 2030s in-service story.

Two other ways to sell the same atom

Centrus Energy sells separative work units and enrichment services rather than pounds, and its backlog reached $4.5bn at 30 June, extending to 2040. The income statement went the other way: June-quarter operating income fell 69% to $10.4m on gross margin of 28.3%, down from 34.9%, and consensus 2026 earnings per share of $2.55 sits 43% below the $3.90 reported for 2025. Its trailing and forward multiples — 60.4x and 59.6x — are essentially identical, because profit is falling as fast as the shares. It also priced $500m of stock and warrants at $199.64 on 9 September; the stock closed at $151.82 on 21 September.

Uranium Energy is the unhedged mirror. It booked no revenue in two of its last three quarters; its one recent sale was 200,000 pounds at $101 a pound, against a quarterly spot average of $80.76. It held 1,456,000 pounds of purchased inventory at 30 April — about $131m of metal against a $4.97bn market value.

The verdict

The group is not one trade and should not be marked as one. Over twelve months the producers with contract books rose — Cameco 6.9%, NexGen 11.6%, Denison 11.3% — while the pre-revenue reactor developers collapsed, NuScale down 78.3% and Oklo 63.8%. Cameco's fall is a very high multiple compressing, from about 65x forward earnings in May to 61.4x now, still 36x 2027 consensus earnings of $2.56; the fuel meters improved through it. Centrus's fall is earned by its own income statement. Uranium Energy's is neither, because there is no delivery book to judge.

Behind all three, the arithmetic accumulates quietly: utilities need roughly 150 million pounds a year to replace what they burn, and 2025 was the thirteenth straight year they contracted for less. That bill does not go away when yields rise.

Nutex Health, which runs 21 micro-hospitals and freestanding emergency rooms across eight states, collected less revenue in its June quarter than a year earlier and still turned its best profit in years. Revenue fell 13.6% to $210.75m. Net income was $65.8m, against a $17.7m loss in the same quarter of 2025. Same-hospital visits rose 6.3%, so almost none of that came from the waiting room.

It came from one line in the cost stack. Total arbitration costs fell $52.3m, booked as a reduction to contract services expense, and gross margin went from 51.2% to 67.0%. Nutex stays out of network almost everywhere on purpose, bills the insurer, and when the offer is low it files the claim into the federal independent dispute resolution process created by the No Surprises Act. That process is now enormous: arbitrators awarded providers nearly $15bn in 2025, more than triple the prior year, and the programme costs roughly $3bn a year to run. Nutex's earnings line is, in a real sense, set by rulemaking.

How the dollar is actually earned

Nutex submits 50% to 60% of its claims into arbitration, prevails in more than 85% of determinations and collects over 80% of what it is awarded. Management has said normalised revenue per visit has run near $4,200 since the process opened in July 2024 and should stay around there — an award rate, never a posted price. The friction shows on the balance sheet: receivables stood at $351.7m at 30 June, up from $319.4m at year-end and equal to roughly 152 days of billings.

Two things cut the cost of running that machine at once. Vendor HaloMD's fees were amended to a pay-on-collected basis, retroactive and extended to 2029, effective 30 June. And a federal final rule dated 28 May 2026 cut the per-party administrative fee from $115 to $15 and allowed up to 50 line items to be batched into a single dispute. "We will not have to record 100% of the IDR costs on every potential legal determination win, as we will now be only accruing costs using a similar collection percentage that we use for our accrual of revenue," chief financial officer Jon Bates told investors. Chief executive Dr Tom Vo was blunter about the strategy: "We can stay out-of-network and still do well... the IDR process is a tool for us to get that fair and reasonable rate." The accruals are converting: operating cash flow of $109.7m in the first half, up 40%, against $31.1m of net long-term debt.

Whose dollar it is not

The tempting story is that provider gains are payer losses in one book. They are not. No Surprises Act protections, and therefore arbitration, do not apply to Medicare Advantage or Medicaid; they cover employer, marketplace and individual plans. Alignment Healthcare, a Medicare Advantage insurer whose shares fell 35.5% in five sessions to 18 September, was hit by hospital and skilled-nursing cost trend flagged at the Baird healthcare conference plus whistleblower accounting allegations. Different money. Insurers and benefit consultants say the arbitration bill is being passed into next year's premiums instead.

Privia Health, filed under the same industry heading, earns from a third pocket: a management fee on the collections of independent practices it does not own, plus lagging annual settlements from the Medicare Shared Savings Program. Its second quarter was strong — practice collections of $970.0m and care margin of $132.1m, implemented providers up 10.1% to 5,644, and guidance raised across every metric to full-year adjusted earnings before interest, tax, depreciation and amortisation of $145–155m, about 16.7x its $2.50bn market value. The shares fell 9.0% on the print and 9.8% more since, on the timing of shared-savings cash: management expects only 70% to 80% of the year's adjusted EBITDA to convert to free cash flow, and that assumes the 2025 settlements land by year-end. Over twelve months Privia is down 12.2% while Nutex is up 124.6%. BrightSpring Health Services, the third name under the label, grew revenue 23.0% on specialty pharmacy scripts at administered rates and trades near 31x forward earnings — an engine sharing no mechanism with either.

What the market is paying for

Nutex is cheap and the cheapness is the argument rather than an oversight: 7.5x forward earnings and 4.1x trailing enterprise value to earnings before interest, tax, depreciation and amortisation, against a consensus carried by three analysts that has earnings per share flat in 2027 and down 9.1% in 2028. The step-change is earned, banked and turning into cash. What it is not is compounding: visit growth is single-digit and capped by construction at three to five hospitals a year, and the cost relief came from a fee schedule and a vendor signature that can each happen only once.

A coalition of 67 health care and advocacy groups asked Congress this month to reopen the No Surprises Act. Nutex's margin has an address in Washington, and now everyone knows it.

NXP Semiconductors sold $1.938bn of automotive chips in the June quarter, up 12.1% year on year and 17% once divestitures are stripped out, and told investors its order book now stretches as far as 18 months out with book-to-bill above 1.0 across its end markets. The Dutch company — which sells microcontrollers, radar and Ethernet processors, connectivity and security chips into cars, factories and phones — guided the September quarter to $3.75bn, plus or minus $100m, an 18% annual gain.

That matters beyond one print because NXP is the cleanest read available on whether industrial and automotive demand is genuinely recovering rather than being restocked. Roughly 59% of its quarterly sales, $2.072bn, move through distributors, who finished June holding 11 weeks of inventory. Lean channels plus orders exceeding shipments is what a demand-led upturn looks like from the inside; a restocking bounce shows up as weeks of channel stock climbing while end demand sits still.

The trough was real

NXP's 2025 revenue fell 2.7% to $12.27bn and net income dropped 19.5%, so the four quarters since — minus 2.4%, then plus 7.2%, 12.2% and 19.5% — are a recovery off a genuine bottom, not a continuation. Gross margin came back to 57.3% from 53.4%, and operating income grew 55.9% on that 19.5% of revenue growth. The sockets driving it are structural: software-defined vehicle architectures and multi-year platform commitments for in-car Ethernet switches and processors, in an advanced driver-assistance component market of roughly $35-38bn growing at low double digits, where NXP and Texas Instruments compete with Renesas on cost and power for camera and perception modules rather than with Nvidia's high-end autonomy compute.

The other two accelerated as well

Analog Devices, the Massachusetts maker of data converters, amplifiers and power-management parts, reported its first $4bn quarter in August, revenue up 40% with gross margin at 67.3%. Industrial revenue grew 53%, and communications — now about 80% data center — grew 84%. Asked on the August 19 call how durable that is, the finance chief grounded it in "bookings momentum, design activity, and pipeline visibility," and said it was not "hopes and dreams." Management also said book-to-bill was "not unusually elevated" and that customers are "running lean," which is the same demand-not-inventory signal NXP's channel weeks give.

Texas Instruments grew 22.8% to $5.46bn in the June quarter with gross margin at 61.4%, up from 57.9%. Far from a fixed-cost squeeze, Dallas cut 2026 capital spending guidance to $2-3bn against depreciation of $2.2-2.4bn, targeting more than $8 of free cash flow per share, and reported $2.74bn of quarterly free cash flow helped by $1.6bn of CHIPS Act incentives. Its capital event is the $7.5bn agreement to buy Silicon Labs, announced February 4 and closing in the first half of 2027.

One repricing, three outcomes

All three peaked within days of June 18 and fell into September — NXP 27.2%, Texas Instruments 17.4%, Analog Devices 13.5% — in a duration repricing already visible across power-analog names, with the 30-year Treasury at a 19-year high in August and the Federal Reserve raising rates on September 16. Only NXP's decline hardened into a confirmed downtrend, on September 9.

What the businesses earn and what the prices say now diverge unevenly. Texas Instruments is the most expensive of the three at 31.3x forward earnings, with the lowest free-cash-flow yield at 2.2% and a consensus that models 2029 earnings of $11.32 a share below 2028's $12.04 — a de-rating with an argument behind it. Analog Devices gave up roughly 13% of its multiple of gross profit in a month while trailing gross profit grew 51%, yet at 18.4x forward gross profit it remains the richest of the three on that measure. NXP, at 7.2x forward gross profit, 13.3x enterprise value to earnings before interest, taxes, depreciation and amortization against Analog Devices' 29.1x, and a 5.1% free-cash-flow yield, is the one whose price contains no recovery at all.

The risk is that NXP's meters are self-reported and its next print is not yet scheduled. Texas Instruments reports on October 20, guided to $5.65-6.15bn — the first outside check on whether industrial and automotive orders held through the quarter that repriced them.

Cencora told investors this month that a slice of the drug volume it ships to Walgreens started going elsewhere on 1 July. The volume sat outside the prime vendor agreement that governs the bulk of the relationship; that contract is unchanged, and the company reaffirmed adjusted earnings guidance of $17.75 to $17.95 a share for the fiscal year.

What makes a partial-volume disclosure worth reading is the denominator. Walgreens and Boots together were about 25% of Cencora's fiscal 2025 revenue, with its top ten customers at roughly 66% — the deepest single-customer concentration among the three companies that source and ship America's drugs. Walgreens has been owned by Sycamore Partners since a roughly $10bn take-private completed in August 2025 and has accelerated store closures and distribution-hub consolidation since. Because McKesson, Cencora and Cardinal Health together move more than 90% of US drug distribution by revenue, volume leaving one of them largely lands at another.

The revenue line is not the business

Cencora, which buys branded, generic and specialty medicines from manufacturers and delivers them to hospitals, retail chains and physician offices, converted $321.3bn of fiscal 2025 revenue into $3.648bn of operating income — a margin of 1.14%. The June quarter shows where the money actually comes from: revenue rose 5.1% to $84.75bn while gross profit rose 35.7% to $3.607bn. Obesity and diabetes drugs explain much of the top line and almost none of the profit — GLP-1 sales rose $2.3bn, or 25%, roughly 56% of all revenue growth, squeezing overall gross margins even as adjusted gross margin widened 0.61 of a percentage point on the OneOncology acquisition. Management raised expected adjusted operating income growth to 13-14% while holding revenue growth guidance at 4-6%.

"We delivered strong results driven by execution across Cencora and investments to advance our specialty positioning and our pharmaceutical-centric strategy," chief executive Bob Mauch said on the 5 August call. Specialty and biosimilar distribution, not retail chain replenishment, is the margin.

The same meters, three different prices

McKesson, whose US Pharmaceutical arm also runs oncology practice management, grew June-quarter gross profit 12.4% on 7.7% revenue growth, with segment operating profit up 19% to $894m. Cardinal Health, which distributes drugs and also manufactures its own gloves, syringes and surgical kits, reported pharmaceutical segment profit up 21% to $645m on 6% revenue growth. Over the twelve months to 18 September, Cardinal gained 50.2% and McKesson 26.7%; Cencora gained 6.2%, and its shares were in a downtrend as recently as July.

The prices now differ more than the businesses do. Cencora is the cheapest of the three at 4.80x forward gross profit and 17.3x forward earnings — about 15.6x the $19.81 fiscal-2027 consensus, against roughly 20x before this year's de-rate. McKesson is the most expensive at 6.52x forward gross profit; Cardinal sits between them, and its re-rating outran a 16.4% lift in earnings estimates while free cash flow was guided down to $3.5-4.0bn from $5.0bn.

What the policy does and does not touch

Section 232 pharmaceutical tariffs take effect on 29 September for manufacturers outside the exempt annex, at 0% for those signing most-favored-nation pricing agreements, with generics excluded entirely. For distributors the practical effect is more branded list-price deflation, which fee-for-service contracts largely neutralize — McKesson said on 5 August that more than 95% of its branded book now runs on those terms.

So the lost Walgreens volume is the cheapest thing Cencora sells, and the profit engine it is building — specialty, biosimilars, oncology services — is untouched by it. That is the case for the gap between its gross profit growth and its share price being an error. Against it stands a plain risk the other two do not carry: a quarter of the revenue line depends on a private-equity owner actively shrinking its store base, and the disclosure gives no size for what has already gone.

Cencora reports its fiscal fourth quarter in early November, as it has for the past two years. That release is the first document in which the July shift has to be measured rather than described.

Coupang's customers came back faster than its margins did. The South Korean retailer — it owns the inventory it sells, delivers it overnight on its own Rocket network, and charges a WOW subscription for the privilege — told investors on 4 August that retained and returning shoppers were spending about 16% more than a year earlier. The same quarter carried an operating loss of $556m.

The distance between those two facts is the bill for a data breach disclosed in November 2025, and it is the whole story of a stock that has lost 57% of its value in a year. Coupang committed ₩1.685tn — about $1.17bn — in compensation, four single-use vouchers of roughly $34 each to every one of 33.7m affected customers including cancelled members, paid out from 15 January. In June, South Korea's Personal Information Protection Commission fined the company about $409m, a record penalty, and the charge landed in second-quarter results. Against a company capitalized at $25.7bn, that is real money spent to keep a customer base that, by the chairman's own May account, had recovered nearly 80% of its lost paid membership.

What actually broke

Strip the regulatory charge and the quarter's operating loss narrows to roughly $146m, which management framed as a modest sequential improvement in loss margin. The damage that persists is in gross profit: 28.2% of revenue in the June quarter against 30.0% a year earlier, with gross profit down 2.6% in absolute terms — the only shrinking gross profit among the large emerging-market marketplaces. Fulfillment capacity was built for volumes that never arrived, volume-based supply-chain savings went with them, and marketing stayed elevated to win the members back.

Top-line growth has decelerated for four straight quarters, from 17.8% to 3.9%, to $8.856bn. But roughly six points of that is the won: revenue grew 10% on a constant-currency basis. "The vast majority of our customer spend never moved," founder and chief executive Bom Kim told investors on the 4 August call. "That group is spending at the highest levels in our history and compounding similarly to before last year's data incident." Asked to confirm the recovery timetable, he was flat: "That's correct. We're talking about margin recovery for product commerce next year." Consensus is not waiting: 2026 revenue of $37.01bn, up 6.3%, a loss of $0.51 a share, and no return to profit before 2027.

The month that had nothing to do with Coupang

That is a de-rating a company earns. The puzzle sits beside it. MercadoLibre, which runs Latin America's largest marketplace and the Mercado Pago wallet and lending book, grew revenue 49.8% last quarter to its first $10bn, with 15-to-90-day arrears on a $16.4bn credit portfolio at 7.0%, near historical lows. Sea Limited, which pairs the Shopee marketplace with Garena's games and the Monee lending arm, grew revenue 48.1% and operating income 28.4%, with 90-day non-performing loans on a book of $11.1bn steady at 1.0%. Neither loan book is deteriorating. Both stocks fell with Coupang over the past thirty days, by 7.0% and 13.5% against its 11.7%, through a week in which the Federal Reserve raised rates a quarter point on 16 September and long yields pushed above 5%. Three incompatible business models, one direction: the likelier reading for the month is the discount rate rather than the mechanics.

Over twelve months the dispersion tells the truer story — MercadoLibre down 27.8%, Sea 46.6%, Coupang 57.2% — and only one of the three has a broken earnings line to justify its place at the bottom. Coupang is priced accordingly, at 2.45x forward gross profit and 0.72x trailing sales, with no forward price-to-earnings ratio because there are no forward earnings to divide by.

The vouchers have been paid and the fine is booked. What Coupang has left to prove is that warehouses sized for a customer base it briefly lost fill back up before the middle of next year.

The meter turned back up

American frac crews went back to work in the first half of September. Primary Vision's count of active fracturing spreads, which had fallen four straight weeks to 178 in the week to 4 September, recovered to 184 and then 187 by 18 September — fifteen more crews than were working a year earlier. Over the same eight trading sessions Liberty Energy, the Denver pressure pumper that fracs and wirelines North American land wells and owns two Permian sand mines, lost 9.5% of its market value, and ProPetro, the Midland-based Permian frac and cementing specialist, lost 12.1%.

It is the second time this year the two pure plays have moved against their own activity meter, and the reason is not in their completions businesses. Both have spent the past year committing frac cash flow to gas-fired generation for data centers — capacity whose megawatts arrive between 2027 and 2029 while the deposits leave now. What the market is arguing about at Liberty is no longer what it charges per stage.

What the pumping dollar did

Liberty's June-quarter revenue rose 14% to $1.19bn and gross margin recovered to 17.5%, from 9.7% a year earlier and 6.2% in the March quarter. Chief executive Ron Gusek credited equipment mix rather than the cycle on the July 23 call: "we didn't give up as much price on that next-generation equipment over time. That had real durability to it," he told analysts.

Below the gross line the picture inverts. Operating income fell 70% to $12.7m, a 1.07% operating margin, because depreciation and maintenance on a fixed horsepower base consume almost everything the pumping hour earns. Liberty's annual operating margin has now compressed three years running — 16.0% in 2023, 8.9% in 2024, 2.0% in 2025 — on revenue that fell from $4.75bn to $4.01bn.

The power bill

Against $602m of consensus 2026 cash profit, Liberty is spending roughly $1.5bn of capital this year, most of the July increase being deposits on engines and turbines. It has locked three gigawatts of generating equipment through 2029 from Bergen Engines and Wärtsilä, targeting 17-18% unlevered returns funded through special-purpose vehicles, and ended June with $559m of cash against $736m of net debt. The $332.6m Wärtsilä supply contract signed on 22 June does not begin delivering until 2029.

What it does not have is a firm offtake. Its most contract-like commitment remains a January partnership with Vantage Data Centers anchored on a 400-megawatt reservation for 2027, and a separate 330-megawatt reservation was cancelled by the developer, with Liberty collecting termination fees. Barclays analyst Eddie Kim cut his target to $23 from $32 after the second quarter, writing that "the company has no energy service agreement or firm contract, which is what investors had been hoping for".

Halliburton prices the same stages

Halliburton — completion and production on one side, drilling fluids, bits and wireline on the other — is the test of whether per-stage economics are rolling over. Its Completion and Production unit earned $474m on $3.2bn of revenue in the June quarter, a 15% margin, up 8% sequentially, with North American revenue flat year on year but up 7% from the March quarter. For the third quarter it guided that segment's revenue flat to down 2% with margins up 125 to 175 basis points — expanding completion margins on shrinking completion revenue. Chief executive Jeff Miller told the July 21 call that pricing traction and filled white space in North America land supported that trajectory.

ProPetro is the counter-example among the pure plays. Revenue fell 6.2% to $305.8m, a fourth consecutive year-on-year decline, and it posted a $4.8m operating loss. Its power arm is further along than Liberty's — contracted capacity rose to about 350 megawatts from 240, including a 60-megawatt behind-the-meter hyperscaler site already generating positive earnings — but it is budgeting $400-450m for power against $125-145m for completions equipment. "We believe the outlook for our completions business continues to improve as market fundamentals tighten and pricing momentum builds," chief executive Sam Sledge said with the second-quarter results.

What the discount is actually for

Liberty is the cheapest of the three on enterprise value to trailing cash profit, at 6.83x against Halliburton's 8.26x and ProPetro's 8.11x, and trades at 0.75x trailing sales versus 1.36x in mid-May. That cheapness is an artifact of a measure that ignores the $1.5bn: on forward earnings Liberty is at 69.9x against 25.6x trailing, because consensus has earnings per share falling from $0.89 to $0.28, and its trailing free cash flow yield is minus 10%. Halliburton's forward multiple of 14.4x sits below its 17.6x trailing with a positive 6.1% cash yield. ProPetro is the richest of the three measured against gross profit, at 13.4x, while losing money at the operating line.

The two legs of the fade should not be conflated. Over three months the damage was specific — Liberty down 28.9% and ProPetro 29.0% while Halliburton slipped 3.7% and SLB rose 6.3% — and it is dated to a power budget without a contract. The last eight sessions were shared: SLB fell 11.5%, Baker Hughes 10.8%, Halliburton 8.6%, after the International Energy Agency on 11 September cut its 2026 world oil demand forecast a further 940,000 barrels a day to an annual decline of 2.5m. That leg belongs to crude. Neither leg is an indictment of the per-stage business, which by Halliburton's own guidance is getting better.

Liberty's first Wärtsilä engines are due in 2029. The pumping business has to pay for them every quarter between now and then.

CrowdStrike booked more new recurring revenue in its July quarter than in any quarter of its history — $332.8m, up 51% from a year earlier — and raised a full-year target it had already raised once. Six days later, Palo Alto Networks told investors that its own subscription book would grow roughly a third as fast in the coming year as it had in the one just finished, and the shares fell about 10%.

These two companies, plus Fortinet, are the three largest listed pure-play security vendors, and they bill for the same threat in three incompatible ways. CrowdStrike, run by George Kurtz out of Austin, sells Falcon modules priced per endpoint per year, so its meter is recurring revenue added. Palo Alto sells firewalls and attached subscriptions and is deliberately bundling product to convert customers onto a contracted book that leads its income statement by years. Fortinet sells FortiGate appliances through distributors and recognizes product revenue on a hardware cycle. Since early May, all three have re-rated by within eight percentage points of one another.

Three meters, three directions

CrowdStrike's ending recurring revenue reached $5.84bn, up 25%, and reported revenue grew 25.8% in a fifth consecutive quarter of acceleration. Falcon Flex — the pooled commitment that lets a customer reallocate spend across modules — ended the quarter above $2.29bn of recurring revenue, growing 101%, with 935 new Flex accounts signed. The worry about Flex was that it pulls revenue forward; instead, new-logo Flex business hit a record 34% of all net new recurring revenue. "Our top 10 deals by deal value were each Flexes," Kurtz said on the August 26 call. Free cash flow of $377m was 25.6% of revenue. Management said both gross and net retention rose but did not put numbers to either.

Palo Alto's fiscal fourth quarter, reported September 1, grew revenue 34% to $3.41bn and next-generation security recurring revenue 63% to $9.10bn, with contracted obligations of $21.2bn. Then came the guide: $11.075bn to $11.175bn for fiscal 2027, or 22-23% growth, as the $21.1bn CyberArk acquisition closed in February and the Chronosphere deal enter both sides of the comparison. Jefferies put organic net new growth at about 45%, stripping those out. "Platformization is the only viable strategy for real-time defense, fighting AI with AI," chief executive Nikesh Arora told investors. In the reported accounts it shows as damage: gross margin fell to 67.6% from 73.2% as purchase-accounting amortization landed in cost of revenue, producing a GAAP net loss of $282m.

Fortinet's June quarter was the odd one: product revenue up 52% to $773m on FortiGate unit growth and higher selling prices, with billings up 33% and operating margin at 33.7%. It is the only one of the three with a substantial GAAP earnings base — and consensus models the refresh as a pull-forward with an end date, with revenue growth stepping from 20% this year to 11% next.

One re-rating

CrowdStrike's price against trailing gross profit went from 32x in early May to 60x now; Palo Alto's and Fortinet's expanded 82% and 78% over the same stretch. Fortinet remains the cheapest of the three by that measure and CrowdStrike the most expensive, at 53.5x forward gross profit and a 0.65% free-cash-flow yield against Fortinet's 2.50%. CrowdStrike's forward earnings multiple of 189x compares with roughly 80-90x in early May — while consensus still models $6.01bn of revenue this fiscal year, unchanged. The single largest session in the group's recent advance produced no earnings at all: on September 14 CrowdStrike rose 15.4% and Palo Alto 13.8% after Anthropic's Dario Amodei published an essay on pacing AI risk.

One technical note matters for anyone reading the chart: CrowdStrike split its stock four-for-one, trading adjusted from July 2. Adjusted, it is up 118% over six months — the strongest of the three, not the laggard its unadjusted price history suggests.

The verdict divides cleanly. CrowdStrike's business earns the direction of its move: accelerating revenue, accelerating recurring revenue, a raised outlook. It does not earn the level, because the estimates underwriting that multiple have not moved while the multiple roughly doubled. Palo Alto's meters are pointing the other way into a comparable expansion, and Fortinet's cycle has a modeled end. Three businesses that agree on almost nothing operationally were marked up together, which means the November and December prints are the first real test of whether the market was paying for security economics or for the word.

Kurtz, for his part, publicly rejected the premise of the essay that added tens of billions to his own market value in a single session, arguing that securing AI's progress matters more than slowing it down.