DK Street Journal

Fair Isaac, the Bozeman, Montana company whose FICO credit scores lenders buy through Equifax, Experian and TransUnion, committed on 1 October to eliminating about 15% of its positions, a plan carrying roughly $27.0m of pre-tax charges in the fiscal fourth quarter, with affected employees notified from the week of 5 October. Ten days later the Mortgage Bankers Association cut the forecast that governs the company's largest profit line, putting 2027 originations by loan count at 5.0 million against the 5.3 million expected this year, with the 30-year mortgage rate averaging 7.3% in the current quarter and holding above 7% through next year.

September's collapse in the shares was read as a fight over which score lenders are allowed to use. The count is the other half of the machine. The royalty is a price Fair Isaac publishes multiplied by the number of times a lender pulls a score, and the company controls only the first term. The cost cut defends the margin on the dollars that arrive; it does nothing to the number of pulls, and that number is now forecast lower.

The June quarter was a price increase

How much of the growth was price is not a matter of inference. Scores revenue reached $458.9m in the June quarter from $324.3m, up 41%, with the business-to-business half up 49% "primarily attributable to a higher mortgage origination scores unit price"; mortgage-origination revenue rose 97% while mortgage score volumes grew low single digits. Mortgage was 62% of total Scores revenue. The list price behind that: Fair Isaac took its wholesale mortgage royalty from $4.95 to $10.00 a score for 2026, after moving it from $0.60 to $4.95 over the prior five years. A roughly 102% price step produced a 97% revenue gain, so units and mix contributed close to nothing.

The other business bears no resemblance to it. Software revenue grew 2% to $215m, with Platform annual recurring revenue up 62% to $413m and net retention of 148% on Platform against 82% on the older non-Platform book that is shrinking beneath it. One consolidated revenue line cannot say which half moved.

The $10 is harder to collect than it was

On 1 October the Federal Housing Finance Agency moved Fannie Mae and Freddie Mac to a single loan-level price adjustment grid for Classic FICO and VantageScore 4.0, removing a 20-point adjustment that had made the rival model's pricing less favorable. The bureaus that distribute FICO jointly own that rival: TransUnion has cut VantageScore 4.0 to $0.99 a score and supplies it free with a FICO purchase, and Equifax has priced it at $4.50 through end-2027, also free alongside FICO. Fair Isaac's answer routes around them: a direct licence for tri-merge resellers at $4.95 a score plus a $33 funded-loan fee, or $10 flat, the first option described by the company as a 50% reduction in average per-score fees into resellers.

"We're not crazy about lenders choice as a policy. We think it's bad policy because it encourages gaming," chief executive William Lansing said on the 29 July earnings call, adding that volumes were not falling, "which suggests that they're pulling both scores." Pulling both is what United Wholesale Mortgage has said it will do; Rocket Mortgage said in early October it will make VantageScore 4.0 its preferred model on all eligible loans from this quarter.

The cheap number nobody has cut

Nothing in the reported results explains the share price. Revenue grew 25.7% in the June quarter, operating margin widened to 53.8% from 48.9%, and diluted earnings were $10.45 against $7.40. The stock fell 16.7% on 4 September and 29.2% on 29 September, closed at $719.63 on 14 October, 56.4% below a year earlier, then rose 9.2% on 12 October and 6.75% on 13 October as Needham cut its target to $1,100 from $1,650 and Raymond James to $1,230 from $1,750 without cutting ratings.

So the de-rating is about the price list rather than the business as reported, and that is also what makes the valuation conditional. At 20.7x trailing earnings against roughly 62x in January, the shares have lost two-thirds of their multiple; the 14.5x forward figure is computed against a fiscal-2027 consensus of $49.65 a share that embeds the $10 score and has not been revised. Senator Josh Hawley asked the Federal Trade Commission on 23 March to investigate the price increases, a second claim on the same lever.

The fiscal fourth-quarter and full-year report is scheduled for 11 November. It is the first occasion on which Fair Isaac can publish a price list for next year, and the first look at score units since the largest retail mortgage lender in the country said it would prefer somebody else's model.

Gas engines for new compressor packages now take about 195 weeks to arrive, and the orders going in are for 2029 delivery. Brad Childers, Archrock's president and chief executive, told investors as much on the company's August call: "For equipment required, lead times are now at 195 weeks with ordering for 2029." The cause sits outside the oilfield. Hyperscale data-centre demand has pushed engine manufacturers to full production, limiting engine availability for the gas compression industry, where waits traditionally ran 40 to 60 weeks.

That one constraint splits a group the market prices as a single shelf. Archrock and Kodiak Gas Services lease horsepower by the month and have been read as geared to the gas price and the drill bit; both inputs moved the wrong way for that reading. Henry Hub was $3.03 per million British thermal units on 6 October, down 8.7% from a year earlier, while the US rig count reached 603, some 56 rigs above last year, with the Permian at 274. Meanwhile the numbers that actually set a lease dollar went up.

A compression dollar is a monthly fee, and the fee rose

Revenue in contract compression is three numbers multiplied: horsepower on revenue-generating contracts, a monthly rate per horsepower, and utilisation. Kodiak, the large-horsepower operator concentrated in the Permian Basin, reported average pricing of $23.80 per revenue-generating horsepower per month in the June quarter, up 4.5% year on year, with compression fleet utilisation at 98.2% against 97.2% and compression adjusted gross margin 1.7 points wider at 70%. Adjusted EBITDA of $216.8m was a company record, up 22%, and Kodiak raised full-year EBITDA and discretionary cash flow guidance in the same release.

Archrock, which operates roughly 35% of the US outsourced compression market, showed the same shape. Contract operations revenue rose 3% to $329.3m, period-end utilisation was 94.4%, and adjusted gross margin was 71% against 70%; the operating fleet shrank from 4.7m horsepower to 4.5m because Archrock sold about 165,000 non-strategic horsepower. The same release disclosed an agreement with an existing strategic customer covering approximately 665,000 horsepower on an eight-year base term with a two-year extension. Compression demand follows the producing well count and the rising gas-to-oil ratio of ageing Permian wells, which calls for more horsepower per barrel even when the oil rig count flattens.

Archrock's shortfall sat in parts and overhauls

Where Archrock did disappoint, the mechanism lives away from the lease. Aftermarket services, the parts and maintenance book billed on customer-owned units, took $42.0m against $64.8m a year earlier, about 11% of group revenue; the company attributed the fall to the absence of non-recurring sales of overhauled engines and to customers skipping the usual mid-year maintenance uptick. Full-year aftermarket revenue is guided to $175m-$185m. The top of EBITDA guidance came down to $885m from $915m on four items: make-ready cost to put idle units back to work, second-half lube-oil pressure, the weaker aftermarket and higher long-term incentive compensation. Contract operations margin guidance held at 71.0%-71.5%. Childers called 2026 "a bit like the calm before the storm".

Enerflex sells what the lessors cannot get

Enerflex earns its money the other way: it manufactures compression and gas-processing packages, recognising Engineered Systems revenue as projects are built, and separately owns build-own-operate infrastructure carrying about $1.2bn of contracted revenue. Engineered Systems bookings were $488m in the June quarter against a trailing eight-quarter average of $363m, a 1.5x book-to-bill for the first half, with backlog up to $1.5bn from $1.1bn at the end of 2025. Then came the power awards: 450 MW of gas-fired generation for a North American data-centre developer on 1 October, and on 13 October more than a gigawatt, taking behind-the-meter backlog above 1.75 GW with deliveries through 2028. Enerflex also authorised roughly $85m of extra Engineered Systems facility investment, mostly for 2027 — a manufacturer adding factory capacity while the lessors stand in the engine queue.

The shares have moved accordingly. Over the three months to 14 October Archrock fell 22.7% and Kodiak 27.5%, while Enerflex rose 6.2% and has been in an uptrend since 7 October. Archrock's turn came at the earnings event itself, from $35.74 on 4 August to $33.08 on 7 August. Kodiak sits 34% below its twelve-month high and 29.5% below the $71.00 at which it sold 10.56m shares for about $758m in May, a raise that lifted diluted shares 10.6% in one quarter to 96.8m.

Cheaper than May, and one of them cheaper still

Archrock at $29.79 trades at 5.97x trailing gross profit, against 7.12x in May, on 9.4x trailing EV/EBITDA and a 5.6% free cash flow yield. Kodiak's forward price-to-earnings is 25.3x, against roughly 29x in May; the honest offset is a 0.10% trailing free cash flow yield and credit-agreement leverage of 3.2x, the cost of a fleet in growth mode.

Enerflex, up 137.5% over twelve months, is the cheapest of the three on price-to-gross-profit at 4.05x forward. That is the comparable measure because the margins differ by a factor of three: 23.9% gross margin at Enerflex against Archrock's disclosed 71% adjusted figure makes price-to-sales meaningless across them.

So the lessors' autumn is a change in what the market will pay for horsepower rather than a change in what horsepower earns. What the results do explain is narrower and real: Archrock's second, thinner dollar shrank by a third, and Kodiak's fastest-growing line is its least profitable, with Other Services revenue up 47% to $43.1m and guided at a 13%-16% adjusted gross margin against 70% on the lease, so mix alone can move reported margin. Kodiak's power optionality is also unsigned. Of the West Texas data-centre project, chief executive Mickey McKee said on the August call: "We've invoiced them for an initial deposit to reserve power equipment for the project while we negotiate a long-term contract." Neither lessor has reported since; Archrock's quarter is expected around 27 October and Kodiak's around 3 November.

Archrock says the industry needs roughly a million more horsepower to meet forecast gas demand growth by the end of the decade, and has set out growth capital predominantly for large-horsepower and electric-motor-drive units. The engines for it are already on order. Whoever holds a place in that queue sets the next lease rate.

Texas Pacific Land owns roughly 880,000 acres of West Texas surface and a perpetual, cost-free royalty on much of the oil and gas produced beneath it. In August the company reported the best quarter in its history on every operating measure it discloses, and the shares fell 12.1% across the two sessions that followed.

The gap was in the top line. Revenue of $246.1m came in under a consensus near $255.5m because water-services revenue slipped and a land-sale gain that had lifted the prior quarter did not repeat. Underneath it, net income was $153.9m, free cash flow $155.5m, and royalty production averaged a record 39,700 barrels of oil equivalent a day, up 20% from a year earlier.

The de-rating is longer than one print. From a $531.95 close on 24 February the stock reached $354.57 on 14 October, a third lower, and it has been in a downtrend since mid-August. February's run — 85% in under two months from the end of 2025 — followed the company's $50m commitment to Bolt Data & Energy, co-founded by former Google chief executive Eric Schmidt, for which it took equity, warrants, the right to supply land, and first refusal on water for Bolt's data centers; KeyBanc initiated at Overweight with a $1,050 target. What the past eight months have removed is that un-contracted land option, while the royalty business accelerated beneath it.

Three unlike dollars in one line

The revenue line cannot settle the question because three different businesses report into it. The royalty is a fractional interest — about 85,000 acres at a 1/128th rate and 371,000 acres at 1/16th — with no vote over when a well gets completed. Water is a separate dollar: brackish sales ran 663,000 barrels a day in the quarter, down 19% sequentially but up 38% year on year, and produced-water royalties on disposal volumes reached a record 4.9m barrels a day, generating $37.1m. Across the first half, water was $165.5m against $317.4m of land and resource revenue.

The contracted land dollar barely registers at all. Easements and surface income doubled to $5.4m. The June agreement to supply land and water to Chevron's Project Kilby, a 2.67 GW gas plant in Reeves County built with Microsoft, carries aggregate consideration of $42.5m — a nominal cash payment plus annual payments running through 2046. A twenty-year annuity is close to invisible in any single quarter.

Fewer rigs, more barrels

The obvious volume worry does not hold either. The Permian rig count was 242 on 30 January, down 20% from 303 a year earlier, and Texas permit applications fell to 570 in April, the fewest since February 2021. Royalty production on Texas Pacific Land's acreage rose 20% anyway. Nor is this the oil tape: WTI traded near $94 a barrel on 14 October after strikes on Saudi facilities and tankers in Hormuz. "Our unhedged royalty position allowed us to benefit fully from the strong oil price environment," chief executive Tyler Glover said on the 6 August call.

The same dirt, bought at a different price

LandBridge, which manages over 325,000 surface acres in and around the Delaware Basin and holds only a small royalty collection, reported the same day and rose 6.0% the next session. Its quarter: revenue $66.8m, up 41%, with surface-use royalties and revenue of $52.2m against $34.2m a year earlier — 78% of the top line, driven mainly by produced-water handling volumes. Free cash flow of $40.2m grew 11%. Growth is partly bought: full-year 2025 revenue of $199.1m was up 81%, and the company agreed to pay $20m for 560 acres under a Lea County landfill. Five Point's holding vehicle still controls about 63.1% of the vote after a November secondary at $71.00, and a Class A reporting change makes its own earnings-per-share series non-comparable quarter to quarter.

Both run near-pure land margins, 88.5% gross at Texas Pacific Land and 96.3% at LandBridge, so price against gross profit is the comparison that survives the structural differences: 31.8x and 30.6x trailing. The quarry operator Vulcan Materials sits at 14.1x. Eight months of trading have brought the royalty owner and the surface owner to the same price for a dollar of land profit, having taken it from the one that collects barrels and given it to the one that does not.

That is where the results stop explaining the move. Earnings grew while the trailing multiple fell from 61.8x in early May to 45.2x, and market value from $29.7bn to $24.5bn; consensus still carries $1,010m of 2026 revenue and $9.30 of earnings per share. A business compounding at 31% with zero debt, $247.6m of cash and a 12.5% dividend increase is not the thing being marked down. The asset being marked down is a gigawatt pipeline no one has signed — and LandBridge's version of it, a long-term option on up to 3,400 Reeves County acres with PowerBridge, is being paid for in full.

Glover told investors on 6 August: "I would be disappointed if we don't announce at least one or more major definitive agreements in the near term." The acreage either converts into signed leases or it goes back to being what it has been since 1888 — ground other people's capital pays to cross.

SkyWest flies 517 regional jets in four other airlines' liveries, and on the contracts that produce most of its revenue the carrier whose name is on the fuselage buys the fuel. That arrangement is why a company sitting inside an energy shock has been the calmest airline on the American board. In the three months to October 14, SkyWest shares fell 5.6%, while Delta fell 10.8%, United 18.1% and American 24.8%.

The outperformance has been read as insulation. It is mostly real, and the fifth of the business the contract does not cover is where SkyWest's earnings are going.

How a regional dollar actually gets paid

SkyWest Airlines, the St. George, Utah operator that flies passenger and cargo service under the Delta, United, American and Alaska codes, is paid under capacity purchase agreements: a fixed rate per contracted aircraft each month plus variable rates per block hour and departure. The major owns the ticket and the route decision. In the June quarter, contract revenue was $864m of $1,102.8m in total, or 78.4%. A fare war does not reach that line, and neither does $4.34-a-gallon jet fuel, nearly double the $2.19 of a year earlier, which took Delta's full-year adjusted earnings guidance down to $5.10–$5.60 a share from $6.50–$7.50.

What SkyWest sells instead is hours. Block hours rose 5.4% year on year and pre-tax income of $139m ran 29% above the March quarter on a 9% sequential gain in hours. "Our block hour production outlook remains solid, as we prepare to invest in owning and operating 34 more E175s through the end of 2028," chief executive Chip Childs said in the July 23 release. Capacity is also coming from reseating metal SkyWest already owns: CRJ550 revenue hours reached 28,102 in the first half against 16,268 a year earlier, and a 41-seat CRJ450 conversion is due in scheduled service by the end of 2026.

The slice that buys its own gas

Prorate and charter flying, where SkyWest takes the fare and the fuel risk, rose $55.4m to $201m, up 38.1%, lifting that share of revenue to 18.2% from roughly 13%. On the second-quarter call management said prorate fuel expense nearly doubled and that strong demand let the company recover about 60% of the fuel increase in fares. The uncovered remainder, plus higher hiring and training cost, is why revenue growth decelerated to 6.5% from 15.0% in the September 2025 quarter while operating income fell 8.4% and net income 16.3%. Operating margin, at 14.1%, is about 2.5 points below the year-ago quarter.

The customer list is the other exposure. Of aircraft in scheduled service or under contract at June 30, 46.2% flew for United, 26.9% for Delta, 18.6% for American and 8.3% for Alaska. The contraction already disclosed is at Delta: 19 Delta-owned CRJ900s that SkyWest plans to return over the next two years, each taking its fixed rate with it. Against that, a Q2 agreement to buy and operate 11 new E175s for American, deliveries committed into 2028, debt down to $2.3bn from $2.4bn, and a $250m addition to the buyback.

Republic raised guidance and fell anyway

Republic Airways, the Indianapolis operator paid on the same per-departure and per-block-hour meter, raised full-year adjusted EBITDAR guidance to $395m–$405m and block hours to about 880,000, and its shares are down 27% over twelve months. Its reported growth is arithmetic from the Mesa combination that closed on November 25, 2025: diluted shares were 46.2m last quarter against 2.78m a year earlier, and the block-hour jump came with 60 additional United E175s. Trailing earnings are unusable; the only anchor is 0.52x book, against SkyWest's 1.33x. The third name on this shelf is gone — Allegiant completed its roughly $1.5bn purchase of Sun Country on May 13, 2026, taking the charter and Amazon freighter operations with it.

What the contract does and does not buy

SkyWest's fixed fee explains the share-price gap against its own customers almost entirely, and explains nothing about its earnings direction. At 8.7x forward earnings, 5.9x trailing EV/EBITDA and an 8.8% trailing free-cash-flow yield, it is the cheapest partner-linked carrier here — Delta trades at 14.0x forward, Allegiant at 11.9x. But a year ago SkyWest traded near 9.6x trailing earnings and now trades near 9.1x on the same calculation, so the 9% decline over twelve months has tracked earnings rather than compressed the multiple. Consensus has full-year 2027 earnings per share at $12.12 against $10.57 this year, which requires the 2026 margin squeeze to reverse; the reported trend is still the other way. Sell-side targets average $113 with a hold rating across five brokers, with TD Cowen trimming to $116 on October 2 and Goldman Sachs moving to neutral at $115.

The contract spared SkyWest the fuel bill and also fixed its rate per hour; it cannot ask United for more because crude went up. What it can do is fly more hours — and the Delta CRJ900s it has agreed to hand back are hours already scheduled to leave the book, on a decision that was never SkyWest's to make.

Applied Materials, which sells the tools chipmakers use to process wafers, presented to investors at SEMICON West in San Francisco on 13 October and stopped short of giving a specific forecast for wafer-fabrication-equipment spending, the industry's headline demand number. Its shares closed down 6.7% at $471.66, Lam Research and KLA each fell about 4%, and by the close on 14 October the equipment group had given up between 9% and 19% of its 2 October value, Applied Materials itself down 17.3% and ASML down 9.1%.

The deepest marks in that group landed on three companies whose revenue a tool budget does not describe. FormFactor sells probe cards, the custom contactors that touch a wafer to test chips before they are cut. Onto Innovation sells inspection and metrology steps qualified into a customer's process, plus the software that runs them. Camtek sells automated optical inspection and 3D metrology systems qualified at manufacturers of high-bandwidth memory, the stacked DRAM that sits beside an AI accelerator. The eight sessions were read as a verdict on capital spending.

The one of the three that published numbers after the scare published better ones.

Billed per design, not per tool

FormFactor's probe cards are engineered around a specific chip design, which is why its own filing attributes the quarter to designs: foundry and logic growth came from stronger demand for networking and high-performance compute microprocessor designs. Inside one capital-spending cycle the end markets moved in opposite directions. In the quarter ended 27 June, foundry and logic probe cards were $121.8m, 47% of revenue and up 22% year on year; DRAM probe cards were $85.0m and up 49%; flash probe cards were $2.9m, down 48%.

Group revenue of $258.2m rose 31.9%, the fourth straight quarter of acceleration, and gross margin reached 50.7% against 37.2% a year earlier. "FormFactor's second quarter revenue, gross profit, and earnings per share set all-time records," chief executive Mike Slessor told investors on 29 July.

The guidance carries the mechanism. FormFactor expects third-quarter revenue of $270m give or take $10m at a 54% non-GAAP gross margin, with DRAM revenue comparable to the record second quarter but containing "a significant underlying shift from HBM to DDR, as customers are shifting their wafer start mix to DDR designs to capture profit opportunity from significant DDR price increases". High-bandwidth memory was roughly two-thirds of second-quarter DRAM revenue. A flat line, produced by customers re-mixing wafer starts between designs, with nothing said about tools at all.

What a reader cannot check is the granularity: the filing discloses that three customers were 35.3% of revenue, and nowhere discloses probe-card unit volume or average selling price. For a business whose output is consumed design by design, the two most natural meters do not exist in public data.

The shares are up 216% over twelve months, yet the trailing multiple fell from 199.3x on 3 May to 80.9x while the market capitalization dropped 11%: earnings did that, not price. Forward, it is 39.3x consensus 2026 earnings of $3.05 and about 31.5x 2027, against the roughly 30x forward anchor the business carried in May.

Onto crossed $1bn of backlog, then lost a fifth

Onto Innovation's June quarter revenue of $343.1m grew 35.3% with operating income up 97%, and it raised its second-half growth outlook to 25% or more over the first half from 15%, lifting expected advanced-packaging growth for the year to about 80%. "Visibility continues to be strong with customers maintaining investments in their multi-year growth plans, resulting in a backlog exceeding $1 billion for the first time in company history," chief executive Michael Plisinski said on 6 August. Backlog later exceeded $1.1bn, 30% to 40% of it covering 2027.

The stock is down 19.6% over three months. It trades at 29.7x consensus 2027 earnings of $8.12, against the 101.9x trailing multiple it carried in May.

Camtek answered with a number

On 14 October, the session its shares fell 4.6%, Camtek pre-announced third-quarter revenue of $160-161m, slightly above its own guidance and 27% higher year on year, guided the fourth quarter to about $180m, and gave a first 2027 outlook above $800m. Consensus for 2027 was $756m. More than $800m of orders have come in since the start of the year.

Camtek is also the one with real deterioration behind it: GAAP operating margin fell to 20.4% in the June quarter from 25.9%, operating income declining 14.8% on revenue up 8%. Its order book rests on two memory manufacturers for more than $260m of expected revenue across 2026 and 2027, and it expects China to be roughly 45% of 2026 revenue, down from 49%, under export controls its filings warn could reduce demand across the Chinese sector. At 37.6x 2026 and 27.7x 2027 forward earnings it is priced off numbers the company has just raised; its 169.7x trailing multiple is an artifact of a one-off $53.2m loss booked in the third quarter of 2025.

What the eight sessions actually priced

None of the three has reported into the de-rating. The most recent calls were 29 July, 6 August and 10 August, so every word of management commentary predates the markdown by two months, and the only fresh disclosure, Camtek's, went the other way. On the evidence available, the reported businesses account for almost none of the move; what moved was the price paid for each dollar of forward earnings, from triple-digit trailing multiples in May to high-twenties and thirties on 2027 numbers.

Two real risks sit underneath. FormFactor's third quarter is guided flat in DRAM because customers are moving wafer starts to commodity double-data-rate designs, and a card count is only as good as the designs running against it. And the same investor day that cut the group named optical inspection as an opportunity for Applied Materials, which is the step Onto and Camtek are paid for; KLA, already the larger rival, expects nearly $1bn of advanced-packaging revenue in 2026 against roughly $635m in 2025, more incremental revenue in one year than Camtek's entire base. Applied Materials also said about 75% of incremental equipment growth should come from leading-edge foundry logic, DRAM and advanced packaging, the three sockets these companies serve.

The next word from inside any of these businesses comes on 28 October, when FormFactor reports the quarter it guided to $270m and says whether commodity memory designs replaced high-bandwidth ones card for card. Whatever Applied Materials would not forecast about 2027 will not be in that answer.

Oxford Lane Capital, a closed-end fund that buys the equity tranche of collateralised loan obligations — the first-loss slice beneath every rated note in a securitisation of leveraged corporate loans — reports investment income that nobody writes it a cheque for. The number is accreted off a modelled effective yield, which is an assumption about future defaults, recoveries and reinvestment. Burford Capital, which advances capital against the value of litigation and arbitration claims, has the same accounting in a different asset: its revenue line is dominated by write-ups on claims it still owns.

Both share prices have been falling for a year, and the natural reading is a market refusing a mark that management would not change. Both marks were changed by management. In the same quarters, the cash moved the other way.

Oxford Lane's accrual sits below its cheque

In the year to 31 March 2026 Oxford Lane booked total investment income of $391.8m and reported a net loss of $585.1m, or -$5.99 a diluted share; the difference is marks. Net asset value per share fell from $15.51 at the end of December to $10.56 at 31 March and recovered to $10.74 at 30 June. The company's unaudited estimate for 31 August was $10.34 to $10.64, figures it says did not go through its usual quarter-end closing procedures and were not approved by the board.

The cash disagrees with the accrual, in the fund's favor. For the June quarter Oxford Lane reported net investment income under generally accepted accounting principles of $0.51 a share against core net investment income of $0.95, and that $0.44 difference is the timing gap between effective-yield accrual and distributions actually received. The weighted-average cash distribution yield on CLO equity was 16.3% at 30 June against a modelled effective yield of 11.1%, so the surplus cash amortises cost basis instead of passing through income. Declared distributions of $0.20 a month, halved from $0.40 on 30 January 2026, come to $0.60 a quarter: above the accounting income, roughly two-thirds of the cash measure.

The collateral improved as the shares fell. The twelve-month trailing default rate for the loan index was 0.97% at 30 June against 1.44% three months earlier. Chief executive Jonathan Cohen called the CLO refinancing and reset market "a very natural offset to spread compression in the U.S. syndicated corporate loan market" on the 28 July 2026 call.

At $8.32 the shares sit about a fifth below the $10.49 midpoint of the August estimate. Eagle Point Credit, the nearest listed competitor for the same CLO-equity dollar, trades at a 27.6% discount against a three-year average premium of 2.11%, so the repricing is the class's, not one manager's. Oxford Lane's board responded on 14 May 2026 by authorising up to $150.0m of buybacks through May 2027 — the reverse of the machine that sold shares above book.

Burford's mark went to court and lost

On 27 March 2026 the Second Circuit reversed the $16.1bn judgment against Argentina, holding the investors' claims "not cognizable under Argentina's civil codes or under the public law governing expropriation". Burford took a $2.4bn capital-provision loss, cutting those assets to about $160m. Tangible book value per share fell from $10.57 at 31 December to $3.12 at 30 June, with debt at 3.5x tangible book against 0.9x.

Collections ran the opposite way. Second-quarter cash receipts of $157m were the firm's highest since the first quarter of 2025, with realisations up to $94m from $62m. "While the YPF loss was disappointing and caused a very large non-cash charge to our quarterly earnings, there is no cash impact from that loss," chief executive Christopher Bogart said with the 6 August results. The remaining book carries roughly $4.1bn of fair value against $1.9bn of deployed cost, about a 22% markup, alongside $733m of cash and marketable securities. At $3.32 the shares are near 1.06x tangible book.

The utilities fund on the same shelf is a different fact

BlackRock Utilities, Infrastructure & Power Opportunities Trust, a closed-end fund holding listed utility and power equities, fell under 2% over thirty days, and almost none of that is the discount: BlackRock disclosed an average daily discount of 1.39% for the nine months to 30 September. Its portfolio returned 19.99% at net asset value in the six months to 30 June. What moved it was the discount rate: utilities fell about 6% on the month as the ten-year Treasury paid above 5.2%.

Twelve months explain these falls; thirty days do not

Across a year the arithmetic holds. Oxford Lane's shares are down 49.4% against net assets cut by a third in one quarter; Burford's are down 70.6% against tangible book cut by 70%. Across the last month it breaks: Oxford Lane fell 12.8% while its net asset value held and defaults under it improved, and Burford fell 24.4% in the stretch after its strongest collections in five quarters.

The two have different reasons for that last leg. At Oxford Lane the likelier reading is the closed-end fund class itself, where the average discount reached 8.4% at the end of the third quarter and taxable-bond funds widened from 6.6% to 9.98% amid tax-loss selling. At Burford it is one courtroom: the certiorari petition filed on 30 September 2026 is the only company-specific news in the window.

That is the awkward part of valuing either one. The anchor a buyer would use at Oxford Lane is itself a Level 3 estimate the market is marking down by a fifth, and the anchor at Burford is a portfolio whose single largest asset is now a residual claim waiting on nine justices to decide whether to look at it at all.

Bel Fuse, which sells power magnetics, circuit protection and interconnect hardware into data infrastructure, defense programs and industrial customers, reported in its June quarter the sixth consecutive period in which it took more orders than it shipped. In October its shares fell with every company that has a semiconductor anywhere in its supply chain, and none of the reasons given had anything to do with Bel Fuse.

The 800-volt direct-current rack architecture that hyperscale operators are moving toward is usually traded through the specialists: the gallium-nitride and silicon-carbide chip designers, and the module makers who convert high voltage down to the processor. The assumption is that the diversified component houses are too broad to participate. Bel Fuse's disclosure says the content is already arriving, and it is the one member of this group whose accelerating business is attached to a forward earnings multiple of 22.7x.

$66.5m of defense and $58m of data solutions

June-quarter revenue reached $210.7m, up 25.2% year on year, with gross margin widening to 39.9% from 38.7% and operating income up 47.4% to $38.4m — profit growing at nearly twice the rate of sales. On the quarter's call, the two reorganized segments both grew: Aerospace, Defense and Rugged Solutions rose 20.6% to $110.5m on defense sales of $66.5m, and Industrial Technology and Data Solutions rose 31.1% to $100.2m, within which data solutions reached roughly $58m against $38m a year earlier.

That is not a one-quarter artifact. Full-year 2025 revenue grew 26.3% to $675.5m with operating income up 67.5%, and the company guided third-quarter sales to $205m-$225m at a 39%-41% gross margin. The run of bookings above sales now covers a year and a half.

Six sessions that hit everything with a chip in it

From 6 to 14 October the shares fell 13.4%, from $260.26 to $225.46. Over the identical span Wolfspeed fell 21.1%, Power Integrations 16.2%, onsemi 15.1%, Monolithic Power 14.9%, Microchip 12.4% and Nvidia 7.1%, against the S&P 500 tracker's 1.2%. The Philadelphia Semiconductor Index fell 3.1% on 14 October alone, a sixth straight down session leaving it off roughly 10%, with the decline attributed to China's ChangXin Memory announcing progress on a higher-density chip design, alongside higher rates and the cost of building AI infrastructure.

Bel Fuse's own uptrend had already ended earlier, in mid-September, and the trend turned decisively downward on 13 October. No lost socket, guidance cut, financing or adverse ruling landed in between.

The cheapest claim on the same rack

Vicor, which builds brick-format power conversion modules at its own fab in Andover, Massachusetts, has risen 366% over twelve months and trades at 62.3x forward earnings and 31.0x forward gross profit. Its incremental dollar increasingly arrives by contract rather than by shipment: royalty revenue hit $30.4m in the June quarter against $10.4m a year earlier, and founder and chief executive Patrizio Vinciarelli described licensing on the first-quarter call as "nearly 100% margin in terms of profitability." Navitas, which sells gallium-nitride and silicon-carbide power chips, is the other extreme — June-quarter revenue of $10.5m, down 27.3%, a gross margin of minus 9.5%, diluted shares up 18.6% year on year, and 51.2x forward sales.

Bel Fuse's 55.9x trailing earnings compresses to 22.7x forward on consensus 2026 earnings of $9.86 a share against $4.64 actually earned in 2025, and 8.4x forward gross profit is roughly a quarter of what Vicor commands. The cash meters agree: 18.7x enterprise value to EBITDA and a 2.71% free-cash-flow yield, against Vicor's 80.3x and 0.46%.

None of the October decline is explained by anything the three companies reported. It is sector contagion, and it fell hardest on the names with the least earnings underneath them — which is why the most expensive member of this group gave back the least and the loss-making one gave back the most. What the selloff did do is widen an already wide gap between a business compounding operating profit at 47% and the price assigned to it. The available disclosure does not separate how much of Bel Fuse's growth is organic from how much arrived with acquisitions, and that question stays open.

Vicor's audit of its twice-raised royalty guidance comes on 20 October and Navitas reports on 3 November. Bel Fuse, the only one of the three whose orders have outrun its shipments for six quarters running, has the least to prove and the smallest multiple to defend.

Methanex, the world's largest supplier of methanol, gave up a plant in Trinidad and Tobago this summer because it could not agree a price for natural gas. The 860,000-tonne Titan unit was indefinitely idled on 29 June, with a $115m non-cash after-tax impairment and a $12m restructuring accrual, in the same months the methanol price was climbing hard. The existing contract was linked to methanol prices, and the company judged the economics unfavourable and any renewal likely to be worse.

The shares are widely read as a bet on that methanol price, and the price has obliged: Methanex lifted its October posted contract price for Asian markets to $700 a tonne, up $150 from September. But the revenue line cannot tell a reader whether the business got better. Realised price is struck against posted price after discounts that average about 15%, and the tonnes sold are partly bought in from other producers. What decides the quarter is how many tonnes its gas contracts let it make.

The tonne that never got made

In the second quarter, realised price reached $529 a tonne against $351 in the first, revenue rose 75.2% to $1,395m and gross margin went from 20.9% to 45.7%. The mix did not deteriorate: of 2,555,000 tonnes sold, 2,151,000 were Methanex's own product and only 247,000 were purchased third-party material. Production, though, fell to 2,213,000 tonnes from 2,391,000 on seasonal gas constraints in Chile, a planned winter outage in New Zealand and unplanned Trinidad outages, even as North America set a record 1.6m tonnes with Geismar, Louisiana contributing 1.1m.

The price step itself is a supply accident elsewhere. Iran, the second-largest methanol supplier at roughly 10m tonnes a year, has seen exports effectively cease since February, and assessments for cargoes delivered into China rose 64% between 5 August and 8 October. Methanex estimates 15m to 20m tonnes of annualised supply must transit the Strait of Hormuz. "Under current conditions, demand rationalization will increasingly be required until a resolution allowing Middle East production to resume and reach the market on a more normalized basis is found," chief executive Rich Sumner told investors on the 29 July call. For July and August the company guided realised price to roughly $460 to $485 and told investors to expect lower adjusted earnings in the third quarter.

At 6.7x forward earnings against 56.7x trailing, the trailing figure is mostly the Trinidad write-down and a first-quarter loss rather than a judgment on the business. Consensus has earnings per share falling 39% next year, to $5.79 from $9.49, which still leaves 11.0x. Measured against gross profit the stock has got cheaper while rising, 4.0x now against 4.7x in May, with market capitalisation up from $4.67bn to $4.91bn. The share count is 14.3% higher year on year from the equity issued for OCI Global's methanol business, and the remaining $290m of a term loan was repaid in the quarter.

The shares have held an uptrend since mid-September and sit about 3% below the May high of $65.72, after a 33% drawdown into early July.

LSB's feedstock is cheap and its benchmark is falling

LSB Industries, an Oklahoma City producer of ammonia, urea ammonium nitrate and ammonium nitrate for farmers and industrial buyers, buys the same inexpensive gas, around $3.20 per million British thermal units in the third quarter to date, against European nitrogen costs near $700 a tonne on $19 to $20 gas, management told investors in July. Its problem is the output price. The Tampa ammonia benchmark that sets its next realisations fell from $825 a tonne in May to roughly $555 in September. The reported quarter looked worse than the business was: ammonia realised $658 a short ton, up 78%, volumes fell 46% on El Dorado and a pulled-forward Pryor turnaround costing $35m to $40m, and adjusted EBITDA still rose to $53.1m from $38.3m. El Dorado has since run at 1,380 tons a day against 1,150 of nameplate.

So one de-rating is rational and the other is not happening. LSB, down 40.8% from its March peak, trades at 16.8x forward earnings where CF Industries, twenty-five times its size and lower-cost, trades at 8.0x; price-to-gross-profit has fallen to 6.4x from 10.2x in May, which surrenders roughly two-thirds of the excess and not all of it. Methanex's slide ended with the quarter that explained it, and the posted price has since stepped up 27% in a month. Neither name moved on a single session: LSB's worst day in the past month was -4.6%, and the sell side raised Methanex targets to $74 and $75 in the first week of October.

Both report within two days at the end of October, and the figure that settles the Methanex question is tonnage. Full-year equity production guidance of about 9m tonnes depends on gas that Chilean, New Zealand and Trinidadian fields have to deliver at a price the company will pay, and Titan is what happens when they stop agreeing.

Six Electrons flew and the launch bill got smaller

Rocket Lab flew six Electron rockets in the quarter ended 30 June and billed less for launch than it had three months earlier. Launch Services recorded $44.6m, a 30% sequential decline, against $189.5m from Space Systems — solar cells, reaction wheels, satellite platforms and separation systems sold to prime contractors and government programs — out of record revenue of $234.1m. Two of the six were HASTE suborbital hypersonic-target missions, and the company's quarterly filing attributes the decline to revenue recognized over time and already booked in earlier periods. Flight rate itself is rising: the 16th Electron of 2026 went up on 11 September, against 21 in all of 2025.

So the segment that names the business was 19% of its sales, and group revenue grew 62% year on year with gross margin at 36.1% against 32.1% a year earlier. What the share price has been paying for is the part that bills almost nothing: Neutron, the 8-ton-class medium-lift vehicle built to contest the roughly $10bn market Falcon 9 holds. On 14 October Rocket Lab disclosed that Neutron's inaugural flight slips from the fourth quarter into 2027 after inspections found voids in a newly cured carbon-composite forward propellant dome. The shares fell 7.85% that session.

"They don't necessarily mean that the dome would fail on flight, but our practice is to proceed only when we believe the system has the appropriate margins. Possible success isn't our standard," chief executive Peter Beck said in a video update on 14 October. It is the vehicle's third public slip: to late 2026 in February after a January tank failure, then an August recasting of the Q4 commitment as reaching the pad.

Firefly's revenue arrives when a milestone is certified

Firefly Aerospace, which sells the Alpha small-lift rocket, the Blue Ghost lunar lander, Elytra servicing vehicles and the SciTec defense-software line, is the same shape in sharper relief. Spacecraft Solutions produced $108.3m of Q2 revenue of $117.7m, leaving roughly $9.4m from selling rides. Chief financial officer Darren Ma told the 11 August call that most contracts, particularly on the spacecraft side, are milestone-driven and release cash as progress is made — which is why gross margin swings with cost incurred, falling to 20.3% from 25.7% a year earlier.

Alpha has flown once in eighteen months, returning to flight on 11 March 2026 with a Lockheed Martin demonstrator after a 2025 in-flight breakup and a destroyed ground-test stage. The Block 2 debut was pushed to the fourth quarter, and medium-lift Eclipse is now no earlier than 2027. Backlog is a record $1.5bn, from $1.3bn a quarter earlier, and liquidity of $940.3m covers about eight quarters at a free-cash outflow that widened to $106.3m.

How much of the fall the businesses explain

Rocket Lab's reported results argue against the decline; its one unbilled asset argues for it. Both names broke down in July, long before the 8-K, and the complex fell with them — Planet Labs 43.9% and AST SpaceMobile 37.9% over three months against Rocket Lab's 23.1% — while Lockheed Martin rose 0.8%. The anchor itself moved: SpaceX closed its 11 June debut at $161.11 above $2tn, then gave back roughly $1.2tn from its peak after an August lock-up expiry, removing the reason to own a proxy.

At Rocket Lab the de-rating is doing rational work. Trailing price-to-sales of 44.5x compares with 75.5x in early May, and the 35.7x forward figure still prices consensus revenue of $959m this year with no positive net income expected before 2028 — against 629.7m diluted shares, up 22% in a year, plus the 29.3m issued in a $1.944bn program to fund the $8bn Iridium purchase. Third-quarter gross margin is guided to 29–31%.

Firefly's case is thinner and cheaper. Its 51.5x trailing price-to-gross-profit compares with 110x in mid-May, and at 33.7x forward it sits near a third of Rocket Lab's 95.8x on faster growth — with its margin sixteen points lower, gross profit is the only fair cross-read. Nothing company-specific landed on the day it fell 7.2% beyond a 10 November earnings date, and its SciTec unit expanded an Air Force contract on 8 October. The likelier reading is sector sympathy and burn.

Beck described Neutron's first-stage tank as the largest pressurized structure ever built from carbon composite, too large for an autoclave, forcing new curing processes — the three-step remediation finishes in the first quarter of 2027, and only then does a launch date come back. Until it does, both companies are judged on the one thing neither sells much of.

Ralph Lauren sold its June quarter at close to full ticket, and that single disclosure separates it from the two companies sitting beside it on the premium apparel shelf. Revenue at the owner of Polo, Purple Label and RLX rose 14% to $1.96bn in the three months to 27 June, with average unit retail, the price actually collected per garment, up 15% across its own stores and sites on what the company called "continued elevation and strong full-price selling trends, with lower than planned promotions".

That line is the whole argument of a premium apparel quarter. A revenue number cannot say whether a brand got stronger: a company holding units flat while pushing price runs the opposite business from one clearing the same garments at 40% off, and both print the same sales figure. August and September were read across the shelf as the affluent apparel consumer cracking, and the group de-rated together. US consumer spending on clothing rose 3.9% year on year in August, ahead of overall spending growth of 2.6%.

What divided these three was who sets the number on the ticket.

Fewer promotions, and the margin to show for them

Ralph Lauren's gross margin reached 73.7% in the quarter, 1.4 points above a year earlier, the company crediting average unit retail growth plus channel and geographic mix for more than offsetting incremental pressure from tariffs and other product costs. Inventory finished at $1.2bn, down 5%: a higher price per unit against falling stock is the opposite of a markdown quarter being staged. Roughly 69% of last year's dollars came through retail against 31% wholesale, per the FY2026 annual report, so the company mostly rings its own prices at its own tills.

The duty bill is modelled to arrive later. On the 6 August call Ralph Lauren held an assumption of about 10% tariff rates through the first half of its year and reciprocal rates in the high teens in the second. "Brand elevation is not a destination," chief financial officer Justin Picicci told investors that day. "It's an ongoing journey, and we're still in the early innings." Guidance went up, to constant-currency revenue growth of 5% to 6%. The shares trade at 19.8x forward earnings against 23.3x trailing, and holding consensus constant that forward figure was 21.2x at the 30 June peak and 17.8x at the 2 September low: a round trip, with no net compression.

The $350m PVH collects on clothes it does not sew

PVH, which owns Calvin Klein and Tommy Hilfiger and licenses both to third parties for categories it does not manufacture, takes in more than $350m a year of licensing revenue supporting over $3bn of licensee sales. How profitable that stream is shows up when it stops: bringing previously licensed women's categories in house lifted net revenue about 1% in 2025 while cutting gross margin roughly 50 basis points, because a royalty dollar carries almost no cost of goods.

Revenue fell 3.2% to $2.097bn in the quarter to 2 August and Calvin Klein revenue dropped 7%, yet Calvin Klein denim sold direct grew double digits on double-digit gains in price per unit. Strip out $107m of tariff refunds, worth 510 basis points, and gross margin rose about 20. The constraint sits in a channel: wholesale revenue fell 6% on European weakness and the Spring 2027 order book, taken months ahead, is down mid-single digits as European wholesalers turn cautious. Inventory fell 3% to $1.738bn, so there is no glut behind it yet.

"We are really leaning into cost as a real driver of EBIT margin expansion," chief executive Stefan Larsson said on 3 September. The diluted share count has gone from 56.7m to 46.1m in two years, so per-share earnings can rise on an unchanged profit pool. At 6.7x forward earnings and 0.72x gross profit against Ralph Lauren's 3.90x, with a 19.0% trailing free-cash-flow yield, PVH is priced off that order book rather than the royalty stream; its trailing earnings multiple says nothing after a $439m goodwill write-off.

Oxford moved its price ladder and the customer stayed home

Oxford Industries has no royalty cushion and little wholesale buffer: direct-to-consumer was 82% of FY2025 sales, with Tommy Bahama 56% of the total. Comparable sales fell 1% in the August quarter, retail down 3%, and Lilly Pulitzer posted mid-single-digit negative comps after entry-level dresses under $200 fell from about half of its styles to 35%.

The cause was self-inflicted. "What we did this year is I think we just went too far too fast in shifting up the pricing tiers," chief executive Thomas Chubb said on 3 September, dating the remedy beyond this year: "Spring '27 is the first season where we were able to really incorporate what we realized was wrong about the assortment in spring '26."

Guidance followed, to sales of $1.43-1.47bn and adjusted earnings of $1.60-$2.00 from $2.30-$2.70. Oxford's forward multiple fell to 14.6x from about 20.8x on 3 September, but trailing EV/EBITDA is still 15.8x, because EBITDA fell faster than the price. Long-term debt came down $70m sequentially to $73m and capital spending is guided to $60m from $108m as the Lyons distribution center finishes, so the balance sheet is being repaired while the brands are.

What the autumn slide actually was

Over the three months to 14 October Levi Strauss fell 18.3%, G-III 20.0%, Kontoor 17.1% and Tapestry 15.4%, with Oxford down 30.8%, while Ralph Lauren (+1.2%) and PVH (+5.7%) were the group's only gainers besides Abercrombie. Over twelve months the three span 49 points, from Ralph Lauren up 17.5% to Oxford down 31.3%, which is why a category average near +6% describes none of them.

The businesses account for one of the three moves. Ralph Lauren collects its price increase at its own registers and has been allowed to keep it; Oxford owns its registers too but set the ladder wrong; PVH's largest region is governed by a department-store buyer's inventory plan agreed two seasons ahead.

The one input all three share is landed cost. The average applied US tariff on apparel reached 22.7% in August, and a garment ordered two seasons ahead carries that duty into a gross margin reported months later, which puts it in the holiday quarter now being traded, the one PwC's survey says consumers enter reporting net-negative spending intent across nearly every discretionary category. When Ralph Lauren next reports, the revenue line will again be the least informative number in it.