DK Street Journal

Chewy paid $400m for Modern Animal, a chain of veterinary clinics, and funded it with an inaugural $600m term loan — the first material debt in the capital structure of a company that has never needed any. The deal takes Chewy Vet Care from 18 locations to 47, with 60 expected by the end of fiscal 2026 and embedded revenue approaching $290m at steady state, according to the company's disclosure. An online retailer of about 100,000 products from 3,000 brands is becoming a landlord of medical practices.

The reason sits in the price data. Veterinary services prices rose 7.8% year over year against pet supplies at 1.5%, while pet food inflation ran just 1.4% in February 2026 — well under the food-at-home rate. And US dog households shrank 4.2% between 2019 and 2025. Selling more bags of kibble to more dogs is not the available growth.

What Chewy's quarter actually showed

On September 9 Chewy reported second-quarter net sales of $3.33bn, up 7.3% — an acceleration from 4.8% the prior quarter — with operating income of $98.4m, up 41%. Autoship subscription sales grew 9.3% to a record 84.6% of net sales, active customers reached 21.7m and net sales per active customer hit $602, each up 3.8%. Full-year adjusted EBITDA margin guidance was raised to 6.7–6.8%.

The shares fell 9.1% that session and have since traded at $20.44, down 43.3% over twelve months, with the 50-day average below the 200-day since late July. Two things explain the reaction. Gross margin was flat year over year at 30.4%; the entire margin gain came from selling and administrative costs falling 70 basis points to 18.4% of sales. And chief executive Sumit Singh told investors plainly: "We are not assuming a meaningful consumer recovery for the balance of this fiscal year." Roughly $15m of the quarter's EBITDA beat was timing — tariff refunds, rebates, gift-card breakage.

Chewy price-matches Amazon on the merchandise itself — the two carried identical prices 93% of the time — so the incremental gross-profit dollar has to come from sockets a marketplace cannot occupy: a sponsored-ads network sold to the brands it resells, running at roughly 2% of net sales against a 1–3% entitlement with about two-thirds flowing to the bottom line, a pharmacy, and the clinics. Modern Animal's mature locations earn EBITDA margins above 20%.

Petco is doing it from the other end

Petco, which runs roughly 1,500 leased pet stores plus grooming, training and 300 in-store veterinary hospitals, reported on September 2: net sales of $1.489bn, up 0.05%, comparable sales up 0.6% for a second consecutive positive quarter, adjusted EBITDA of $122m including a $6.8m tariff refund. Selling and administrative costs rose $1m. That is expense discipline against fixed occupancy, and it is working: $170m of debt retired over nine months against $1.48bn outstanding, a 2.0x leverage target, seven consecutive quarters of hitting cash-flow goals, in CFO Sabrina Simmons' account on the call. Its veterinarians are employees, not franchised partners, and hospital expansion begins in 2027. Chief executive Joel Anderson was blunt about the fleet: "We have not been the best custodians of the physical part of our brand."

Petco is also importing Chewy's mechanism, rolling Autoship into stores; it already accounts for about half of Petco's digital sales.

The verdict

Neither company is being paid for demand, because there isn't any to speak of — the US pet market grew 3.7% to $158bn in 2025. What Chewy earns is the mix shift: a subscription base compounding faster than the company, an ad network at a media margin, and now clinics bought with borrowed money. What nothing yet explains is a 43% twelve-month decline against 26.2x forward earnings versus 39.3x trailing, a 6.7% free-cash-flow yield and consensus fiscal-2027 earnings growth of 48.5% — though 26x forward on a 7%-growth retailer is not cheap in absolute terms. Petco at 5.6x trailing enterprise value to EBITDA and 0.58x book is a $700m equity stub beneath $1.48bn of gross debt, and the operating-lease liability on 1,500 stores is not disclosed in its results, so true leverage is worse than the 3.4x the guidance implies.

The market values a dollar of Chewy's gross profit at seven times what it pays for Petco's. Both are now buying the same veterinary dollar — one with a term loan it never needed before, the other with cash it is currently using to pay down debt.

Clearway Energy has promised $3bn of corporate capital between 2026 and 2029, and the cheapest piece of that plan is the piece it will not sell. The company is a holding vehicle for roughly 5,000 net megawatts of American wind and solar plus about 2,500 net megawatts of natural gas generation, spun out of NRG in 2018 and staffed by 60 people. Its plan called for $500m to $1bn of new equity. It has raised $50m.

That matters beyond one funding decision, because equity is the marginal money in a yieldco. The remainder of the plan — more than $500m of retained cash flow and more than $1.5bn of corporate debt, $600m of it already raised this year — is largely committed. The share price sets the price of everything else: the drop-downs bought from sponsor Clearway Group, the third-party projects, and ultimately the cash available for distribution per share that governs the dividend. Management has deprioritized third-party project acquisitions until the stock recovers.

The spread that vanished

What repriced was the alternative. The 30-year Treasury closed 11 September at 5.36%, its highest since 2007, against the 5.27% reading of a week earlier; the Treasury had already doubled its long-end liquidity-support buybacks to at least $4bn per operation effective 9 September, and yields rose anyway. Clearway's quarterly dividend, raised to $0.4602 in February and to $0.4676 in May, annualizes to $1.87 — about 6.0% at $31.14, or roughly two-thirds of a percentage point over the long bond. A contracted annuity that clears the risk-free rate by that little is not an expensive thing to replace.

Meanwhile Clearway's debt is cheap: $600m of senior notes priced at 5.750% in January, against a leverage target of 4.0-4.5x and a BB rating. Its repowering program is deploying roughly $600m at cash-available-for-distribution yields of 11% to 12%. Borrowing at under 6% to earn 11% is obvious arithmetic; issuing shares at 0.69x book to do the same is not.

The business kept working

June-quarter revenue was $481m, up 22.7%, with operating income up 36.5% to $116m — growth accelerating from 18.8% in the March quarter. All three ERCOT wind projects, more than 600 megawatts, were restructured onto contracts running past 2040 and accretive to cash from the first month, at a time when North American wind power purchase agreements cleared at a record $79.40 per megawatt-hour. The one genuine deterioration was weather: 2026 CAFD guidance cut about 8% at the midpoint, to $430m-$470m, on poor first-half wind.

"While we are lowering our 2026 financial guidance due to factors outlined in our mid-July operational preview, our team is focused on maintaining our trademark operational excellence to uphold our historic track record of meeting our financial targets," chief executive Craig Cornelius told investors on 5 August. The 2027 target of $2.70 or better per share was reaffirmed; the shares sit at 11.5 times it.

The shares themselves have fallen 9.3% in thirty days and 17.4% in three months, with the 50-day average under the 200-day since 20 July. The decline was not Clearway's own: Brookfield Renewable fell 12.0% and Ormat 16.3% over the same month. The only discoverable company news was an 8 September leadership change, with Steven Ryder becoming chief financial officer while keeping the same job at the sponsor — a dual hat at a vehicle whose central governance question is what it pays that sponsor for assets.

The mirror: a yieldco with no yield

XPLR Infrastructure, the former NextEra Energy Partners, barely moved: down 0.5% over the same thirty days, because it has no distribution to be repriced against a 5.36% bond. It suspended one and redirected the cash into buying out convertible equity portfolio financings — outside investors who can otherwise claim the underlying projects. In the June quarter it completed the first minimum buyout of one such financing for about $150m and repaid $500m of converts. Roughly $2.5bn of buyouts remain, falling due at annual intervals from 2028 through 2034, with about $470m planned for 2027 and an intention to sell the assets inside one financing next year. Revenue grew 6.1% to $363m; operating income fell a third to $60m as margin dropped from 26.3% to 16.5%. It trades at 0.33x book and 10.0x forward earnings against 17.0x trailing.

What the move earns

The guidance cut explains a slice of Clearway's decline and the rest is the discount rate — which is a real loss, not a misunderstanding, since a vehicle valued as a spread over Treasuries has less spread. But the operating evidence points the other way: recontracting is repricing upward, not down, and the 2027 cash target is fundable out of retained cash and debt alone. The exposed part of the plan is the back end, the 2030 range and the 2 gigawatts of late-stage solar-plus-storage behind it, which needs an equity market Clearway currently declines to use. XPLR is what that looks like taken to its conclusion: the same contracted megawatt-hours, routed to financiers for the rest of the decade, with growth arriving through sponsor joint ventures because issuing units is no longer an option.

Clearway will roll its five-year growth target forward into 2031 on the third-quarter call, alongside a roadmap to a payout ratio below 70%. It will be setting that target in a market where its own shares are the most expensive money it can raise.

Howard Hughes Holdings spent $2.1bn of cash in June on a Bermuda specialty re/insurer, and its master-planned communities now largely exist to pay for what comes next. The company completed the acquisition of Vantage Group Holdings on June 4 at roughly 1.5x year-end 2025 book value, funded from cash plus $1bn of non-voting exchangeable perpetual preferred stock issued to Pershing Square, with a further $200m of capital put into Vantage at closing. Pershing Square manages Vantage's $3.4bn investment portfolio without charging a fee.

Executive Chairman Bill Ackman was explicit about the destination. The goal, he told the August 6 earnings call, is to make Howard Hughes "disproportionately an insurance holding company as opposed to a real estate company with an insurance operation." On the same call: "The priority for every incremental dollar of free cash flow is to put it into Vantage." That is the stake. Shareholders own a land bank carried at decades-old cost whose monetization — management expects roughly $2.5bn to $3bn of excess real-estate cash over five years — is earmarked for underwriting risk. The market capitalization, about $3.67bn, is below the $5.6bn of "projected margin effective residual value" chief executive David O'Reilly attributed to the remaining wholly owned land bank on that call, before Teravalis and Floreo.

The land is still selling

Nothing in the operating record explains a stock at its lowest close in a year. Howard Hughes sold 119.7 residential acres in the second quarter at an average $1.309m per acre for $156.7m, per its earnings release, above the $1.2m average of the first half. Pre-tax earnings from the communities segment rose 32% to $134.7m and net new home sales across the portfolio rose 12%. In Hawaii, 527 units closed at The Park Ward Village, delivering $226.6m of net proceeds and $130.9m of gross profit; the pipeline behind it is 1,293 units, 78% pre-sold, better than $4bn of future revenue — though the company expects only about 40% of that to be recognized across 2026 and 2027.

Three things do argue for a lower price, and all are compositional. Growth is concentrated in Texas: Bridgeland's net new home sales rose 17% and The Woodlands Hills' 34%, while Summerlin managed 2%. That maps onto its market — Las Vegas net new-home sales fell 15% in the first half to 4,284 with permits down 25% as mortgage rates near 6.8% cut absorption per community. Second, this year's segment guidance of $343m to $391m is, by management's own account, essentially flat against 2025 excluding one large Summerlin superpad sale, so the 32% is a comparison effect. Third, Teravalis — 37,000 acres in Buckeye, Arizona, planned for 100,000 homes — needs water: Arizona will no longer issue new certificates of assured water supply on Phoenix-area groundwater.

The recurring stub is small and steady. Operating-asset net operating income was $70.5m in the quarter, up 2%, guided to $279m–$290m for the year. Against that sit near-term maturities of $230m for the rest of 2026 and $596m in 2027, with $1.2bn of liquidity including $698m of cash, per the 10-Q.

What the insurance arm actually looks like

Vantage is not a shell. It wrote $473m of gross premium in the quarter, up 29%, and carries $1.8bn of book value. It also ran a 101.6% combined ratio against 94% a year earlier, hit by $18m of Iran-conflict catastrophe losses and $19m of adverse prior-year development, even as the accident-year ratio excluding catastrophes improved to 91.4% from 96.2%. Marc Grandisson, formerly chief executive of Arch Capital, is executive chair; David Gansberg, Arch's former co-president, is chief executive designate, targeting mid-teens returns on equity. The holding structure already cost something: S&P cut Vantage's core subsidiaries to BBB from A− and then withdrew the ratings, citing the credit quality of the new leveraged parent rather than underwriting, while AM Best affirmed A− with a positive outlook.

The external management economics are modest for now: a $3.75m quarterly base fee, $15m a year, plus 0.375% a quarter of market-cap appreciation above a reference of about $3.93bn. The stock is below that line, so the incentive fee pays nothing. Pershing Square paid $100 a share for nine million shares in May 2025; that tranche is down roughly 38%.

The control that re-rated

The same mechanism elsewhere is being paid handsomely. St. Joe, which owns 170,000 acres in Northwest Florida and sells lots to builders, has spent a decade converting acreage into a membership club, hotels, marinas and leased commercial space. Its second-quarter revenue rose 23% to $158.8m with net income up 37% to $40.5m — "the highest in the second quarter in the company's history," chairman and chief executive Jorge Gonzalez told investors on July 31 — with residential, hospitality and commercial margins all up. It bought back $32.7m of stock in the quarter and trades at 4.9x book and 17.3x trailing EV/EBITDA. Howard Hughes trades at 0.73x book and 8.6x, and its 12.4x trailing earnings is flattered by condo closings and land gains that do not repeat on schedule. Shares of the two have moved almost symmetrically over twelve months in opposite directions.

So the de-rating is not the land repricing. Price per acre rose, Houston volumes rose, and the condo backlog is contracted. What is being repriced is the destination of the cash: a self-liquidating property portfolio whose proceeds leave real estate for a Bermudian balance sheet with one quarter of loss-affected results, a withdrawn S&P rating, and $1bn of preferred sitting ahead of the common. Investors who bought a land bank at a discount to appraised residual value are being asked to keep owning it while the discount funds something else. The Las Vegas slowdown and the Arizona water problem are real and quantifiable; neither is worth a quarter of the equity in a year.

On September 30, Ackman, chief investment officer Ryan Israel, O'Reilly and Grandisson will present strategy at the annual shareholder meeting. The land has been answering for itself all year; the question on that stage is what the insurer is worth to the people who still own the acreage.

Archrock, which rents natural-gas compression packages to gas gatherers and processors under fee-based contracts, told investors on August 5 that it no longer expects the top of its 2026 profit range. Full-year adjusted EBITDA guidance came down to $865m–$885m from $865m–$915m. The reason was a lubricant bill, plus the cost of refurbishing idle units for redeployment and higher stock compensation — not weaker demand for horsepower.

That is an unfamiliar way for the oil price to matter here. A compression operator's earnings are fleet horsepower multiplied by utilization multiplied by the monthly rate charged per horsepower, less field labor and consumables. Crude reaches that arithmetic in exactly two places: lube oil is a refined crude product, and expensive crude changes when customers schedule maintenance. With WTI trading near $96 after the US-Israeli conflict with Iran and the effective closure of the Strait of Hormuz, both channels opened at once — and the shares have been priced since late June as though something larger had gone wrong with demand.

The cost shock, quantified

Kodiak Gas Services, the large-horsepower operator concentrated in the Permian, sized the lube-oil headwind at roughly $18m annualized and said it was offsetting it through maintenance optimization and vendor supply agreements. USA Compression Partners, the master limited partnership that owns the largest US fleet by horsepower, flagged about $1m a month of incremental lubricant cost in the second half as contracts renew; it has no direct pass-through clauses and is trying to recover the money through renegotiation, with consumer-price escalators as a partial offset.

None of it won. Kodiak lifted its monthly rate to $23.80 per horsepower, up 4.5%, ran record fleet utilization of 98.2%, and still widened compression adjusted gross margin to 70%, 170 basis points better than a year earlier. Archrock's own contract-operations margin improved to 71%. Natural Gas Services, the Midland-based manufacturer-operator, reached record utilization of 88.3% at $28.06 per horsepower per month, more than 5% higher year on year, and raised its 2026 EBITDA guidance to $103m–$108m from $92.5m–$97.5m. USA Compression averaged $22.84 per horsepower, up 7%.

The supply side explains most of that pricing power. Caterpillar engine lead times are stable at roughly 195 weeks by Archrock's account — new orders land in 2029 — and have been extending further as data-center demand absorbs capacity. USA Compression has already contracted about half its 2027 deliveries. On the demand side the volumes compression is bolted to keep growing: the Energy Information Administration expects gross LNG exports of 16.7 billion cubic feet a day in 2026 against 15.1 in 2025, with the Permian adding 1.4. Permian rig counts have fallen — 242 in late January, down about 20% year on year — but well productivity has kept output rising anyway, so the associated-gas stream has not thinned.

The one that shrank

Archrock is the exception, and its problem is its own. June-quarter revenue fell 3.1% to $371.2m, operating income dropped 53.7% to $93.0m, and the operating fleet declined to 4.5m horsepower from 4.7m, with utilization at 94.4%. Aftermarket Services — parts and maintenance sold to customers who own their own units — fell 35% to $42m as those customers deferred work while oil was expensive; Archrock's management characterized the deferral as temporary, "pay us now or pay us later." Consensus has 2026 earnings per share at $1.72, below the $1.84 Archrock actually earned in 2025. Against that, it signed an eight-year contract covering 665,000 horsepower with an existing midstream customer, cut leverage to 2.6x from 3.3x and raised the dividend 10%.

The valuations have not sorted this out. Archrock trades at about 10.1x trailing enterprise value to EBITDA and Kodiak at 10.6x — near parity between a fleet that contracted and one whose revenue grew 21% to $391.1m. Archrock's forward price-to-earnings of 19.1x sits above its 17.6x trailing, the arithmetic of expected declining earnings, and against the 21.8x trailing it carried in May. Natural Gas Services is the cheapest of the four at 7.8x, with the lowest leverage.

What the shares are arguing

All four peaked between May 19 and June 26 and sit 9% to 22% below those highs; over the past month three of them rose. Archrock's moving-average trend signal turned decisively negative on September 8 even though the stock had bottomed at $30.70 on August 24 and recovered since, and no company-specific news was discoverable in that window — the likelier reading is a lagging signal catching up to the summer decline, with J.P. Morgan having put a Buy on the name on September 2.

So the verdict splits. The crude shock is a genuine margin tax, and it is the reason one company's guidance came down — but renewal pricing has outrun it at every operator that has reported, which is the opposite of the demand fade a 20% de-rating implies. Archrock's decline is earned by a smaller fleet and falling forecast earnings. The other three have not yet produced a number that justifies theirs.

Kodiak's next answer comes from a different business entirely: its power arm already runs a 405 MW fleet earning a 64.5% gross margin, it has an equipment agreement with Baker Hughes framing a pathway to 1.8 GW by 2030, and its first behind-the-meter contract with a hyperscaler in West Texas is still only at limited notice to proceed. The definitive version is promised before year-end.

Limbach told investors in August that it expected to sell more and earn less. The building-systems contractor, which installs and maintains heating, plumbing, electrical and control systems for hospitals, universities, factories and data centers, raised its full-year revenue outlook to $760–790m while cutting full-year adjusted earnings before interest, taxes, depreciation and amortization to $78–84m from $90–94m — roughly 11% lower at the midpoint. The shares fell 32% that session, from $77.11 to $52.46.

What makes the cut worth more than a day's trading is where the margin went. Limbach has spent years shrinking the work it wins through general contractors in order to grow owner-direct service, maintenance and small-project work sold straight to building owners — an installed base meant to pay like an annuity, at roughly double the gross margin of construction. In the June quarter that segment was 74% of revenue at $128.4m, and its gross margin fell to 24% from 29%. Management attributed the drop to the lower-margin baseline of Pioneer Power, an acquisition that supplied all of the segment's growth, and to labor and materials inflation on data-center work. The annuity absorbed job-cost inflation the way a job does.

How the profit is actually booked

Mechanical contractors recognize revenue as work progresses, which means a quarter's reported profit is a function of what it currently costs to finish the contracts in hand. One revised estimate on a large fixed-price job lands directly in gross margin. Comfort Systems lists "use of incorrect estimates for bidding a fixed-price contract" among the factors that could make its results differ materially.

At Comfort Systems the mechanism ran the other way. Second-quarter revenue rose 50.3% to $3.27bn, and gross margin reached 25.9% against 23.5% a year earlier — a fifth consecutive quarterly expansion, which the company said was bolstered by favorable project estimates and the successful resolution of change orders. Operating income grew 86%. Backlog finished the quarter at $14.06bn, 73% above a year earlier. Excluding acquisitions, backlog rose $1.39bn in three months against $3.27bn of revenue — bookings running near one and a half times the work burned.

The composition is narrower than the total suggests. Technology work, mostly data centers, was 58% of first-half revenue against 40% a year before, and the sequential backlog growth came chiefly from two Texas operations — an electrical business that added $1.00bn and a modular business that added $510.2m. New construction is now 74.8% of year-to-date revenue, up from 63.2% for 2025, so more of the company's revenue sits in the part of the model where an estimate decides the margin. Chief executive Brian Lane told analysts on the July 24 call that the constraint is self-imposed: "we only take work that we know we can perform... We don't out-kick our coverage."

Limbach's bookings held; its margins did not

Limbach's revenue grew 21.9% to $173.5m in the quarter while gross profit fell 6.4%, taking group gross margin to 21.5% from 28%. The general-contractor segment grew revenue 35% to $45m with gross margin down to 14.5% from 24.7%, which the company called primarily timing. Demand was not the problem: $182m of work was booked in the quarter, $616m over three quarters, and the general-contractor backlog was rebuilt to about $200m after quarter-end. Limbach also bought Simpcore for $30m, a data-center program-management platform overseeing customer budgets of more than $8bn, targeting $12m of revenue in 2027. "The data center cares about its time and schedule," chief executive Michael McCann said on the August 5 call. "So they will pay up for somebody who is gonna move really quickly."

The end market was not decelerating while any of this happened: US data-center construction spending reached $75bn in July, 57% above the prior year. EMCOR, the scale rival for the same fit-out work, posted record remaining performance obligations of $17.14bn, up 44%, and raised full-year earnings guidance — while its mechanical margin fell 110 basis points because it deliberately moved roughly a tenth of its work to guaranteed-maximum-price and construction-manager roles that carry lower markup on complex artificial-intelligence jobs. Legence raised its own full-year guidance in August and reports no shortage of skilled trades across an 11,000-person workforce.

What each decline earns

Comfort Systems is 18.2% below its June 22 peak of $2,066.51. A dollar of its trailing gross profit now costs 20.6x, against 27.2x in early May — but also against 9.4x at EMCOR, 10.9x at Legence and 3.7x at Limbach for physically comparable work. That is an outlier premium being compressed, and nothing in the June quarter marks down the order book; the selling arrived alongside the 30-year Treasury yield touching a 19-year high of 5.33% on August 18, a discount-rate event for long-duration backlogs.

Limbach's fall is the one its numbers earned. Consensus now has it earning $3.99 a share this year, down 4.2%, and the stock at 12.7x that figure is the cheapest in the group on every measure — but it is cheap against declining near-term earnings. The useful conclusion is about where risk now sits. Comfort Systems is shifting its revenue toward the construction work whose margin depends on estimates it has so far revised in its own favor, while Limbach has learned that selling direct to the owner does not exempt it from the cost inflation that spoils a job.

The next print settles the open question: whether 24% was Pioneer Power's arithmetic working through the average, or the price of doing business inside a data center.

Redwire, which builds solar arrays, star trackers, spacecraft structures and in-space manufacturing hardware — and, since buying Edge Autonomy, drones and reconnaissance payloads — reported record revenue of $117.1m for its June quarter, up almost 90% year on year. Gross margin came in at a record 27.8%. A year earlier the same line read -30.9%, a quarter in which cost of revenue exceeded revenue outright. Contracted backlog reached a record $542.1m, and over the trailing twelve months the company booked $1.52 of new orders for every dollar of revenue, against $0.87 a year before. The shares fell 21.6% in the thirty days to September 11.

A record quarter can sit beside a de-rating because of the way these companies are paid. Almost all of this work is fixed-price development: the builder quotes a total, and when it revises its estimate of what the job will finally cost, the whole revision lands on margin at once, catching up every prior quarter in the current one. Reported gross margin is therefore an estimating statement before it is a pricing one. Redwire's quarterly gross margin has printed 14.7%, -30.9%, 16.3%, 9.6%, 26.6% and 27.8% — a swing of nearly 60 percentage points inside eight quarters on a broadly similar revenue base. That mechanism governs all three of the spacecraft builders below.

What Redwire now is

The other thing the quarter disclosed is that Redwire is no longer mainly a space company. Defense Tech revenue was $61.9m against $55.2m for the Space segment, and management said the drone and defense-electronics half would keep growing faster. Chief executive Peter Cannito's team attributed the record margin to that mix, to a shift from development work toward production, and to estimate revisions that were net-neutral in the quarter; the guide from here is low-to-mid 20s. Total debt was cut 75% to $48.9m and liquidity ended at $607.8m. Adjusted earnings before interest, tax, depreciation and amortization were still negative $3.2m. "We are reaffirming our full year 2026 revenue forecast in the range of $450 million to $500 million," chief financial officer Chris Edmunds told investors on the August 6 call.

What shareholders own less of is the claim on all that. Diluted shares stood at 220.5m against 89.6m a year earlier, and on June 9 the company replaced a $350m at-the-market equity program with one for up to $500m. Redwire's market value is higher than in May while its share price is barely changed: issuance closed the gap.

The same mechanism, biting

Intuitive Machines, the Houston lunar-lander builder that has since become a satellite prime, showed both directions at once. June-quarter revenue more than quadrupled to $206.2m and gross profit swung to $35.9m from a loss of $11.8m. But gross margin fell from 39.0% in March to 17.4%, and the company booked a $14.7m estimate-at-completion adjustment on its IM-4 mission for payload accommodation. Backlog is a record $1.8bn, bookings through the August 13 call were $1.2bn, and the mix has moved away from NASA lunar delivery: 50% of those bookings were commercial, 30% national security. "This quarter demonstrates that Intuitive Machines is no longer dependent on a single market, customer or mission cadence for growth," chief financial officer Peter McGrath said that day. The recurring piece — a Near Space Network communications contract with a $4.82bn ceiling — has booked nothing yet; the first relay satellite flies on IM-3 in early 2027. The mood is visible in the fact that the stock fell around 9% in early September on a new spacecraft order.

The one that broke

York Space Systems, which mass-produces standardized satellite buses and listed in January at $34 a share, is the exception that makes the group legible. On August 13 it cut 2026 revenue guidance to $375m–$405m from a $570m midpoint. "We've observed a meaningful shift in how the U.S. government is acquiring spacecraft systems, moving away from a rapid succession of larger RFPs to an IDIQ approach that is slow to start, but often faster to accelerate task orders later," chief executive Dirk Wallinger said — indefinite-delivery vehicles, which award a place in a queue rather than a job. Roughly 30% of York's 2026 new business went with it. Backlog still grew 9% to $592m and the win rate held at 88%, so this is award timing rather than lost demand — and the budget churn around the Space Development Agency's next data-transport tranche is the same story from the buyer's side. Redwire sits on the identical hinge: it is one of 15 vendors on a $981m Space Systems Command vehicle that pays nothing until task orders issue.

The verdict

None of the three earns money, and their gross margins run from 17% to 28%, so what they cost is best read against the gross profit they generate. Redwire trades at 29.6 times the past year's gross profit and 26.6 times the forward figure, against 58 times in early May. Intuitive Machines is at 17.1 times trailing and 9.1 times forward, against 437 times in May. York, at 2.85 times trailing sales with almost no discount to forward sales, no longer has growth for that gap to price.

Only York's order book deteriorated. The other two declines are mostly sector-wide: Rocket Lab fell 45% over the same three months, Planet Labs 52%, AST SpaceMobile 39%, and pre-profit space names have been sold together as a theme, against a backdrop of climbing Treasury yields pressuring small caps. Layered on that is real dilution — Intuitive Machines has sold $291m of stock at an average of $26.81 against $14.35 today. The businesses explain why nobody will pay ahead for these order books; they do not explain the compression itself.

Under fixed price, a builder finds out what a contract earned only when it has nearly finished building it. Redwire's record margin and Intuitive Machines' halved one came out of the same accounting rule in the same quarter, and the next set of estimate revisions arrives with the September numbers.

Starbucks spent the past year deliberately adding back scheduled labor hours, and its stores got more profitable while it did. North America segment operating margin reached 13.6% in the quarter ended June 28, against 13.3% a year earlier, on comparable store sales up 8.1% led by a 4.5% increase in transactions rather than price. Store operating expenses rose to $3,789.3m from $3,552.4m in dollars, but held at 51.2% of segment revenue against 51.3%: the comp absorbed the reinvestment.

That arithmetic matters because Starbucks is not a franchisor. Company-operated stores produced $6,754.8m of the segment's $7.4bn of revenue, so every hour added to a schedule lands on Starbucks' own books rather than a licensee's — the reason McDonald's, whose franchisees carry the crew cost, ran a 46.5% operating margin last quarter. The turnaround Brian Niccol is running is therefore a margin question, funded alongside a $1bn restructuring announced in September 2025 that closed 400 to 500 North American stores. "Our third quarter results are proof they do," Niccol said of the plan's premises in the July 29 release. Guidance went up with it: fourth-quarter US comps of 6.5% or better, full-year non-GAAP earnings of $2.55 to $2.65 a share. The labor line is not settled — Starbucks Workers United has proposed a $17 starting wage against a company floor of $15.25 to $16.00 in 43 states.

The other machine

Dutch Bros, which runs drive-thru-only beverage shops under the Dutch Bros and Rebel brands from Grants Pass, Oregon, earns its growth by opening buildings. Revenue rose 32.5% to $550.9m in the June quarter, company-operated same-shop sales rose 8.3% on 3.4% more transactions, and it opened 48 shops, 44 of them company-operated, against a target of more than 150 for the year at roughly $1.4m of capital each. New units are not fading: chief executive Christine Barone told investors on the second-quarter call that the Melrose Park shop near Chicago was pacing to about $7m of volume, a company opening record. Shop contribution margin still fell to 30.6% from 31.1%, because cost of goods rose eight-tenths of a point to 26.1% of company-operated revenue on coffee contracted earlier and a new food program now in about 750 shops.

The number that broke the stock was the shape of the comp. "System same shop sales growth is now estimated to be in the range of 5% to 6%, with us trending towards the midpoint of that range," chief financial officer Joshua Guenser said on that call, guiding third-quarter system comps to 4-5% against 5.8% delivered. Shares fell 18.4% the next session — on a call that raised the revenue outlook to $2.1bn-$2.13bn.

What the group did

The last month was not about either company. Between August 12 and September 11, McDonald's fell 8.5%, Cava 19.8%, Texas Roadhouse 14.8%, Brinker 13.7% and Shake Shack 12.0%, while the S&P 500 tracking fund lost 1.1% and Chipotle rose 10.5%. Starbucks' 8.5% and Dutch Bros' 14.3% declines sit inside that. No Starbucks disclosure in the window explains it; the most prominent September item was an interview in which Niccol said the turnaround is working, so the likelier reading is rotation out of restaurant valuations.

And the valuations were where the two diverge. Starbucks trades at 37.8x forward earnings and 56.7x trailing, against 19.6x forward for McDonald's. Consensus has it earning $2.61 a share this fiscal year — below the $3.31 it actually earned in fiscal 2024, before annual operating margin fell to 9.6% from 15.0% — which puts the shares near 30x their own pre-turnaround earnings power, with consensus revenue growth of only 2.8%. Dutch Bros is at 45.9x forward, down from roughly 68x against its mid-June price, and 34.8x next year's consensus; Cava, the nearest fast-growing comparable, sits at 102x forward on expected earnings growth of 3.5%.

So the two falls mean different things. Starbucks' is the compression of a recovery multiple rather than a verdict on the recovery: four straight quarters of comp growth and two of margin expansion say the plan is converting, but the market had been paying in advance for earnings the company last posted two years ago and is paying less for them now. Dutch Bros' is a repricing of growth's shape — unit count doing the work while comps step down to 4-5% and forward-bought coffee keeps shop margin under the 30% Guenser has called the target — even as estimates rose through the decline. Cheaper than June is not cheap: 34.8x next year still needs the openings to keep landing at the volumes management describes.

The bean, for once, is cooperating. Arabica has fallen back near $3 a pound from last year's record and the extra US tariff on Brazilian green coffee came off last November — but chains buy forward, and better Brazilian supply takes months to reach the cup. Both companies are pouring last year's coffee price into a market that has already stopped charging it.

The largest number in Appian's story is one a court took back. On 8 January the Supreme Court of Virginia unanimously affirmed that the $2bn trade-secret verdict Appian won against Pegasystems be thrown out and a new trial ordered, faulting a jury instruction that shifted the burden of proof on damages onto the defendant. The case returns to Fairfax County Circuit Court before a new judge, for a retrial of both liability and damages. No trial date has been identified.

That matters beyond the docket because Appian, the McLean, Virginia maker of low-code software that builds workflows, forms and data models without hand-coding, is a $2.57bn company still losing money on a reported basis. A contingent recovery larger than its own market value was doing quiet work in the equity story. With it gone, the shares rest on the operating business — and that business has been sending two signals at once while the stock rose 47% in three months.

Three revenue lines, pulling apart

June-quarter revenue was $203.3m, up 19.1%, a deceleration from 21.5% in March. Cloud subscriptions, the part that should compound, reached $131.7m and grew 23%, with net expansion from existing cloud customers steady at 115%. The drag sits on either side of that line: non-cloud subscriptions growing at low-to-mid single digits, and a professional-services arm that grew 20% to $45.6m, faster than software, holding services at roughly 22% of revenue. The mix did not shift toward the higher-margin line; it held still.

Gross margin fell to 71.2% from 74.3% a year earlier. The compression came from the software side, where subscription gross margin slipped to 84% from 86% on the cost of serving artificial-intelligence features, while services margin actually widened to 31%. Meanwhile remaining performance obligations — contracted revenue not yet recognized — stood at $625.3m at 30 June, up 13%. The book grew more slowly than the revenue it feeds.

The cash line is where the quarter was won. Adjusted EBITDA of $16.2m beat guidance of $5m-8m, operating cash flow swung to a positive $12.1m from an outflow of $1.9m, and the company bought back 1.8m shares for $43.9m. Chief executive Matt Calkins, explaining why Appian pulled next year's hiring forward, told analysts on 6 August: "I think it might be an indication of strong pipeline that's unusually strong and also our confidence of being able to win in the current environment."

The federal book, and its new date

One customer set deserves separate accounting. US federal agencies supplied 25.3% of revenue in 2025, up from 23.9% in 2024 and 21.3% in 2023 — a concentration that has risen every year. The 30 September funding cliff that would have tested it was defused on 1 September, when the House passed a stopgap 370-48, funding the government to 11 December. The risk moved rather than cleared: a continuing resolution holds spending at prior-year levels, which keeps most new awards on hold, so a quarter of Appian's revenue base sits with buyers who cannot start much that is new until December at the earliest.

What the rerating is paying for

Appian now trades at 4.14x forward gross profit, against 3.07x in early May — roughly a 35% expansion in four months — and 32.1x consensus 2026 earnings of $1.08 a share. Consensus has revenue growth halving to 10.8% in 2027. Procore, the construction-management platform filed under the same software heading, did the opposite: it advanced 26% over three months and still trades at 6.65x forward gross profit, cheaper than the 7.96x it carried on 3 May, because a first quarter of GAAP operating profit and 24% growth in contracted obligations outran the share price.

The two moves are therefore bought with different currency. Procore's is paid for in reported cash flow. Appian's is paid for with a narrative — 85% of new logos taking AI-enabled tiers, a profitability beat — set against decelerating revenue, a thinner gross margin and a backlog growing slower than sales. What is not being paid for is the lawsuit: Pegasystems itself rose 15.6% over the past thirty sessions alongside Appian, which is not how a market prices a courtroom in which one side's gain is the other's loss.

Management has guided full-year cloud subscription revenue to $525m-529m, or 20% growth, against the 23% just delivered. The second half has to fund the multiple now, because the courthouse no longer will.

Repligen agreed in July to buy BioLife Solutions, the Bothell, Washington maker of the CryoStor and HypoThermosol media in which cell therapies are frozen and shipped, and the antitrust waiting period expired at 11:59pm on 3 September, leaving a BioLife shareholder vote as the last condition before a fourth-quarter close.

The price is roughly $1.5bn of enterprise value — $11.25 a share in cash plus 0.1442 Repligen shares, 64% stock and 36% cash. What that money buys is a revenue mix. BioLife's sales are about 98% consumables, at a 64% gross margin and adjusted earnings before interest, tax, depreciation and amortization of $7.4m last quarter, 26% of revenue. Repligen is buying the part of its own model that works, because the other part still does not.

Paid by the gram

Repligen sells Protein A affinity ligands, CaptivA chromatography resins and OPUS pre-packed columns, XCell alternating tangential flow and KrosFlo filtration, and SoloVPE process analytics — the disposable flow path inside a biologics plant. Nearly all of it is consumed in proportion to grams of drug substance actually made, which makes the base business an installed-base-times-utilization annuity with a lumpy machine-sales line bolted on.

The annuity is working. Second-quarter revenue of $204.1m grew 13% organically, consumables rose in the high teens, proteins climbed about 50% and analytics more than 30%, and gross margin reached 53.1%, three points better than a year earlier. Full-year organic guidance went up to 10.5–13.5%, with adjusted earnings of $2.03–2.09 a share. Equipment revenue stayed muted, though chief executive Olivier Loeillot told investors on the July 28 call that second-quarter equipment book-to-bill was "significantly above one", with a second competitive award won and a third expected, building 2027 backlog.

Two things cut against it. Reported year-on-year growth has slowed every quarter for a year, from 21.9% to 11.9%. And on that 11.9%, GAAP operating income fell 14.2%, to $11.9m.

Repligen is also third in its signature socket: independent market work puts Danaher's Cytiva at roughly 35.7% of Protein A resin, Merck KGaA at 16.2% and Repligen at 13.6%. The offsetting fact is that Repligen supplies ligand to Cytiva and MilliporeSigma under supply agreements and buys base resin from Purolite, which this year extended five ligand programs to 2032 — so some per-gram economics accrue even when a rival's column wins.

The mirror that isn't one

Avantor, the Radnor, Pennsylvania distributor and producer of lab chemicals, consumables and single-use assemblies, is filed under the same heading and runs a different machine: a 31.7% gross margin, gross profit down 3.1% last quarter, and adjusted net leverage of 3.3x. Its second quarter inverted the obvious read. The VWR distribution arm returned to 1.7% organic growth from a 4.8% first-quarter decline, while the Bioscience & Medtech Products segment — the piece that resembles Repligen — fell 5.6% organically. "Our Revival program is strengthening how we serve our customers, leading to improved business performance," chief executive Emmanuel Ligner said in the results release. Consensus still has Avantor's 2026 earnings falling about 10%, to $0.81.

What the prices already hold

The whole life-science-tools shelf bottomed in March and re-rated together; 10x Genomics has more than tripled since. Customer funding is the plausible mechanism — trailing-twelve-month biotech capital raised through the second quarter reached $95bn, up 69%. Repligen gapped down 5.8% on the BioLife announcement and up 7.8% around its results; Avantor jumped 15.8% in a single session on 39.8m shares. Since mid-August Repligen has gone nowhere and Avantor has added nearly 8%. BioLife at $34.95 sits within a third of a percent of the $35.07 its merger terms imply, so its chart is now Repligen's chart.

The verdict splits. Repligen's operating record earns the direction of its move but not its level: 80x forward earnings, 22x forward gross profit, and still 48x consensus 2028 earnings of $3.45 — a price that assumes the return of 2022, when this business ran a 28% operating margin against roughly 6% today. Avantor's 4.8x price-to-gross-profit is the cheaper number by far, but it is being paid for a second derivative that has not yet reached the estimate line, and the only Avantor segment sharing Repligen's mechanism is the shrinking one.

The vote is the last gate. Clear it, and Repligen owns a high-margin annuity it did not have to wait for anybody's capital budget to release — while still waiting for the equipment orders that only a capital budget can release.

Two companies sell the same thing at the same checkout — an instalment plan on a several-hundred-dollar purchase — and pay for the money in opposite ways. Affirm borrows it in the capital markets. Klarna, a licensed Swedish bank headquartered in London, takes it from depositors. Over the past month that one structural difference, rather than anything happening to borrowers, decided which of the two the bond market could reach.

Affirm has spent the year keeping more of what it originates. Loans held for investment reached $9.56bn gross at 30 June against $7.03bn a year earlier, while funding debt doubled to $3.3bn, according to its fiscal-2026 annual report. Its average annualised cost of funds was 5.8% in the March quarter, down about 1.3 percentage points year over year as it repriced warehouse lines and upsized a securitisation to $750m from $500m, the company told shareholders. Hold a book that size and the long end of the curve becomes an operating input. When the ten-year Treasury yield touched a one-year high of 4.79%, Affirm fell about 5% alongside SoFi on 1 September, while Robinhood, which carries no loans, sat still. Across the thirty days to 11 September the balance-sheet lenders all fell — Upstart 12.2%, PayPal 8.7%, Affirm 6.3% — while Visa and Mastercard, which lend nothing, rose.

The borrower data did not move

Nothing in Affirm's June quarter explains a de-rating. Gross merchandise volume rose 36% to $14.1bn and revenue less transaction costs — the company's own gross-profit line, capturing gain-on-sale and servicing — grew 39% to $589m, or 4.2% of volume, so the take rate widened as the volume grew. Thirty-day delinquencies on monthly instalment loans improved to 2.5% from the 2.7%–2.8% of the prior three quarters. Amazon, at 22% of volume, signed a five-year renewal, and top-five partner concentration eased to 44% from 47%.

The exposed seam is the merchant-subsidised 0% offer, where a retailer pays the discount and the lender eats the funding. Underwriting those loans is "really, really hard science," chief executive Max Levchin said on the fiscal fourth-quarter call, where small errors produce "a lot of unprofitable transactions." That 0% volume grew 30% in the March quarter against 52% for pay-in-four. Chief financial officer Rob O'Hare guided fiscal-2027 take rates to roughly this year's level — explicitly on the assumption that funding costs and funding mix hold.

The bank that the curve cannot touch

Klarna funds about 90% of its book with roughly $13bn of consumer deposits, alongside a debit and banking franchise Affirm does not have. Its June-quarter transaction-margin dollars grew 42% to $446m on revenue up 27% and volume up 18%; provisions fell to 0.52% of volume, a third consecutive quarterly decline; the company posted a $9m profit, its first as a listed firm. "Three years ago, our transaction margin did not cover our adjusted operating cost," chief financial officer Niclas Neglen said on the 18 August call. "Today, roughly $0.56 of every additional transaction margin dollar reaches the operating line."

The shares fell 22% the next session and have ground lower since, to $13.83 — within $1.56 of their all-time closing low and some 65% below the September 2025 offer price. The cause was the top line, not the credit line: full-year volume guidance was cut to $149–151bn from more than $155bn on weak German discretionary retail and about $600m of adverse currency, as the company disclosed. UBS moved to Neutral with a $16 target and Morgan Stanley cut to $17. Consensus now has 2026 revenue growth at 18.4% against 31.6% delivered in 2025, and both the finance and marketing chiefs leave in early 2027.

Price-to-earnings is unusable for either — Affirm's fiscal-2026 net income was inflated by a one-off deferred-tax release, and Klarna's 2026 consensus earnings round to a cent. Against gross profit, Affirm trades at 8.39x trailing and 6.40x forward, down from roughly 9.6x trailing in mid-August because the profit base stepped up faster than the price. Klarna trades at 2.04x forward gross profit and 1.26x forward sales.

What each side earns

The pair answers its own test unevenly. Deposit funding plainly wins on cost, and Klarna's improving loss rates prove the product is not the problem — yet it is the capital-markets-funded lender that carries a threefold premium, because Affirm's volume is compounding in America at 36% while Klarna's largest market is shrinking in dollars. Affirm's 41% advance from March is earned by the take rate and the delinquency line; the last three weeks' give-back is not, and tracked the ten-year with no company news behind it. Its 5% rebound on 11 September came on 4.5m shares against the 29.0m traded after earnings, which is an unwind rather than a re-rating.

Klarna has now said substantially all new US and German Fair Financing originations from the second half of 2026 will be fair-valued and sold — the forward-flow model Affirm built, adopted by the one lender that did not need it. The bank is learning to sell its loans just as the borrower is learning what it costs to keep them.