X-Ray Analyses

A Buffett-lens X-ray of Lowe's at $186, 35% below its February high. How a home-improvement retailer makes money from DIY shoppers and Pros; Marvin Ellison's margin climb and the slide since 2023; the $8.8bn FBM acquisition and the debt behind it; the moat tested against Home Depot, Menards and Amazon; Berkshire's five years as a shareholder; a negative balance sheet explained; the six dividend tests; three feed faults; and a valuation by three methods.

A Buffett-lens X-ray of Qualcomm at $175.50. How a patent toll and a chip business make money side by side; why Apple can stop buying the modem but not stop paying the royalty; memory costs and China; the bet on cars and AI data centres, including Amazon's $60bn warrant deal; the Arm trial; buybacks that now exceed free cash flow; the six dividend tests; three feed faults; and a valuation by three methods.

A Buffett-lens X-ray of Hershey at $162.74, 32% below its high. How a chocolate franchise makes money from checkout shelves and four holidays; what tripling cocoa did to prices, volumes and margins; the moat tested claim by claim against Mars, Ferrero, GLP-1 and private label; the Milton Hershey School Trust, Berkshire's McLane and the share sales; the six dividend tests on 96 years of payments; one feed fault; and a valuation by three methods.

A Buffett-lens X-ray of UPS at $94.57 and a 6.9% yield. How a hub-and-spoke parcel network makes money from density; why UPS cut Amazon by half and what it has done to revenue and cost per package; the moat, claim by claim, now that Amazon sells its own network to shippers; Carol Tomé's record and the 2022 dividend and buyback decisions; the six dividend tests on a payout that consumes all the free cash; four feed faults; and a valuation by three methods.

A Buffett-lens X-ray of Abbott at $96.69, 28% below its high. We sort what broke into what is passing (formula pricing, China, the COVID cliff), what is a fight (Libre against Dexcom) and what is permanent (the $23bn, debt-funded Exact Sciences deal); run the six dividend tests on a 54-year Dividend King; correct our own feed's capex sign; weigh the formula litigation; and value the business on owner earnings.

A Buffett-lens X-ray of ServiceNow at $137.76, two-thirds off its high at the April low. We test the 98% renewal rate as a moat, weigh the case that AI agents need a governed system of record against the case that they shrink the seat, draw the gap between a 29.5% adjusted margin and a 4% GAAP one, judge the $11.7bn acquisition spree, and value the business under three AI futures.

A Buffett-lens X-ray of Uber at $67.88, a third below its high. We explain why the network is wider than ever in human-driven rides and delivery, draw one week of Uber's trips against Waymo's, weigh the aggregator and disintermediation cases, subtract the tax gains, portfolio marks and insurance-reserve build from the profits, judge the $14.8bn Delivery Hero takeover, and value the business under three robotaxi futures.

A Buffett-lens X-ray of the retail investor's favourite stock, at $190. We explain the Ontology and the bootcamps, test the moat (net dollar retention of 157%), weigh the AI labs and cloud giants now selling the same layer, count stock-based pay back into the 'Rule of 40', look hard at a founders' vote of 49.999999% on a tenth of the shares — and run the price backwards: $190 assumes about 32% growth a year for ten years.

A Buffett-lens X-ray of the stock of the year, at $1,788 after a sixteen-fold rise. We show why fiscal 2026 was a price event rather than a volume event (gross margin from −12% to 85% on unchanged costs), test the five claims to a moat, read the new long-term contracts as an owner would, compare SanDisk with Samsung, SK hynix, Micron, Kioxia and YMTC, and value it the way Graham taught for any cyclical — on the earnings of a whole cycle, not the best year in it.

A Buffett-lens X-ray of Enbridge at $47.14 (≈ C$66.7). A 5.8% dividend raised for 31 straight years, covered by distributable cash flow guided at C$5.70–6.10 a share; a C$41bn backlog and ~5% annual growth; but debt above target at 5.1× EBITDA, a C$3bn share sale to fund new acquisitions, and Line 5 back in Michigan's courts. We map the network, read the numbers the pipeline way, and explain why it is a sound income holding with a leveraged balance sheet.

A Buffett-lens X-ray of Disney at $105.52. Parks and cruises earn 54% of segment profit and grew 20%; streaming has swung from a $4 billion loss to a profit; adjusted EPS is guided up ~16% this year and double digits next; a new CEO from the parks business is in charge; and the shares sit at 14 times next year's earnings. We weigh the fading cable channels, ESPN's costs and the $60 billion parks build-out against one of the most valuable story libraries on earth.

A Buffett-lens X-ray of Verizon at $46.62. The 6.1% dividend is covered about 1.8 times by free cash flow guided to grow 9–10% this year; a new CEO is cutting costs hard and growing again; but the company carries $200 billion of debt after buying Frontier, grows its service revenue about 3% a year, and competes in a market where price wars come and go. Why it is a sound income holding and not a compounder.

A Buffett-lens X-ray of ASML at $1,770. The sole maker of EUV lithography machines raised its 2026 sales guidance by a fifth in three months and plans 30% more capacity for 2027 — while China, a third of last year's sales, is being cut back and US lawmakers push to cut it off further. We explain the monopoly, the cycle and the geopolitics, read the numbers in euros, and ask how much of the AI wave is already in a share price that has nearly doubled.

A Buffett-lens X-ray of TSMC at $452, a record. Two-thirds of every revenue dollar is gross profit, revenue is growing over 40%, and 77% of wafer sales come from nodes where it has few rivals. We read the numbers in the right currency, explain why the ADR costs 15% more than the Taipei shares, weigh the $265bn American build-out, and face the one risk no spreadsheet can price. Why Buffett sold, and why we would still own a measured slice.

A Buffett-lens X-ray of Prologis at $133, the world's largest owner of logistics warehouses. We map what it owns market by market from its 10-K, show the rent gap that is still to be collected as leases roll, weigh the data-centre pivot, read the numbers the REIT way (Core FFO, not EPS), and test a dividend raised 5.9% this year. A wonderful business at a fair price.

A Buffett-lens X-ray of Netflix at $71, 43% below its high. We chart how revenue and margins grew, strip out two one-offs from the last four quarters (a $2.8bn Warner Bros. break fee and a $619m Brazilian tax charge), follow the cash into record buybacks, and weigh the analysts' engagement fears against what Netflix actually reported. A great business at a price that assumes the worst about viewing.

A Buffett-lens X-ray of AbbVie at $264, near its record. We chart how Skyrizi and Rinvoq replaced Humira, explain why reported earnings (74×) and 'adjusted' earnings (19×) tell different stories, count what AbbVie spends buying its next drugs — about $30 billion in thirty months — and test a dividend raised every year since 2013. A very good business at a fair price, not a bargain.

A Buffett-lens X-ray of Home Depot at $293, near its 52-week low. Mortgage rates near 7% have frozen home sales and big-ticket projects; earnings have slipped for three years; the company has borrowed $23 billion to buy SRS and GMS and paused buybacks; and its CEO took medical leave in August. We separate the housing cycle from the business, test a dividend raised just 1.3% this year, and explain why 19.5 times depressed earnings is a price to begin at.

A Buffett-lens X-ray of P&G at $146. Seventy years of dividend increases, a new chief executive, organic growth of 1% that came entirely from price, and a fiscal 2027 plan that absorbs about $1 billion of higher raw-material, energy and transport costs. We trace the earnings bridge, show that nearly every dollar of free cash goes back to owners, and explain why 21 times earnings is a price to accumulate at.

A Buffett-lens X-ray of ExxonMobil at $161, with Brent near $100 after the Iran war closed much of the Strait of Hormuz. We separate what Exxon earns in an ordinary year from what it earns in a war — $6.70 a share at $69 oil in 2025, an expected $12 this year — explain why the first quarter looked weak and the second spectacular, show where the 2025 cash went, and test whether the 43-year dividend is safe. It is. The price is another matter.

Our second X-ray of Coca-Cola, three months after 'Accumulate' at $80. Volume returned (+5%), guidance rose, the dividend grew for a 64th year — and we correct two June errors in public: Henrique Braun, not James Quincey, has run the company since March, and a $14 billion IRS dispute we omitted is now before the Eleventh Circuit. We also explain why screens show a dividend bigger than the cash, and why it never was.

Our second X-ray of Alphabet, three months after 'Own, Watchfully' at $357. Search grew 17%, Cloud 82%, and the ad-tech case ended without a break-up. But Alphabet's biggest-ever quarterly profit came mostly from marking up its stakes in SpaceX and Anthropic — $6.26 of $9.11 a share — and capital spending, now guided to $205 billion, overtook operating cash flow. We grade our June call, strip the paper gains out, and explain why a better business at a lower price is one to accumulate.

Our second X-ray of NVIDIA, three months after we called it 'too hard' at $200. The shares are $225 and the business has beaten every forecast — but we grade ourselves in public, including a dividend rise we missed. Then the finding: NVIDIA now owns $99 billion of stakes in AI companies, many of them its own customers, and the gains on those stakes made up about 16% of its profit in the last six months. What happens to a supplier's earnings when it finances its own demand?

Oracle's shares have fallen 57% from their high even as revenue grows 30% and the order book has reached $664 billion. We separate the two Oracles — the database franchise that still throws off cash and the AI cloud that consumes it — and follow the money: capital spending from $2bn to $56bn in five years, free cash flow from +$14bn to −$24bn, debt past $150bn, $20bn of new shares in a single quarter and a credit rating one notch above junk. Then the finding that ties this to Microsoft: more than $500 billion of the two companies' order books rests on OpenAI.

SpaceX went public on 12 June in the largest IPO in history and is now worth about $1.96 trillion — 105 times last year's revenue, with a net loss of $4.9 billion. With so little history, the only honest way to analyse it is to ask what the price assumes and by when. We set out every major plan — Starlink, direct-to-phone, Starship, the Moon, Mars, orbital data centres, xAI — alongside what was promised, what has been delivered and our realistic window for each. Then we do the arithmetic: to return 10% a year to 2035, SpaceX must earn about $235 billion a year, nearly twice what Microsoft earns today.

Our second X-ray of Microsoft, three months after the first, and the share price is 36% higher. We graded ourselves in public: the business thesis held, but we published two days before the low, called the price fair, and then asked for 15% less — and missed a 43% rise. We explain the mistake, then look at what has changed: Azure past $100 billion, a $678 billion order book with OpenAI supplying at least a third of it, capital spending at 35% of revenue with free cash flow falling, a dividend raised by the smallest amount since 2017, and what 29 times earnings means with a 5% Treasury.

The largest bank in America earned $21.2 billion in a single quarter — but $4.2 billion of that was a one-off gain on Visa shares, and much of the rest came from a trading boom its own CEO warned will not last. We take the record apart, explain what a bank is actually worth (its tangible book multiplied by how much it earns on it, divided by what investors demand in return), and show why the Fed's first rate increase since 2023 and a 5% ten-year Treasury matter more to the share price than any quarter's earnings. Plus Dimon's succession, the $1.65 dividend declared nine days ago, and eleven things our own data feed gets wrong about banks.

The most famous dividend stock in America has just had the best year of its modern life. In twelve months J&J absorbed the loss of its biggest drug, settled the talc litigation that had hung over it since 2016 for $5.5 billion — less than the $9–10 billion it offered two years ago — and struck a pricing deal with the White House. The shares rose 54% to a record. We explain why the recovery is real, why at a drug company amortisation is not the harmless accounting charge it is elsewhere, why the dividend is safe while the future has to be bought, and why at 23 times earnings and a 1.98% yield the price has already collected the good news.

Our second X-ray of NKE, nine weeks after the first and eighteen percent lower. A public scorecard of the three tests we set: wholesale passed convincingly, China and margins failed. The quarter's headline 49.2% gross margin was about 900 basis points of one-off tariff recovery; without it the margin fell ten basis points and earnings per share were $0.20 rather than $0.72. On the year that makes clean EPS $1.58 against a $1.64 dividend, which is 104% of earnings and 110% of free cash flow, with the buyback switched off and the twenty-fifth consecutive raise due in November. Set against that: running has grown double digits for five straight quarters, Foot Locker comps turned positive for the first time in four years, and On missed while Hoka slowed.

Our second X-ray of ADBE, seven weeks and one earnings report after the first, at a price nine cents higher. A public scorecard of our own call: six things right, one wrong, and one where we picked the wrong metric to watch. Revenue accelerated to 12.9% and the contracted order book decelerated from 13% to 8%; gross margin broke an eight-quarter band; net new recurring revenue fell by about a third; and the payoff for giving the product away was explicitly deferred to 2027 by the man who will not be there to deliver it. Meanwhile Figma grew 48% into a negative operating margin and Canva was marked down 17% by its own backers — both for the same reason. The question stopped being whether these tools survive. It is what they earn.

Our second X-ray of O, ten weeks after the first: a scorecard of what we said at $62 against what happened — AFFO +3.8% in Q2 and guidance raised, the 136th dividend increase, Fitch's 'A', and a price that fell anyway as the ten-year went to 4.85%. The real news is the funding model: public stock paid for 18% of 2026's $10B of investments against a 47% norm, the Core Plus Fund went from $80M to $3.3B in nine months, Apollo bought half of 492 stores, and a $6B data-centre venture arrived the day our first report went out. We draw where the money now comes from and what it buys, and name the two numbers the transformation cannot hide: 6.4% and 1.9%.

A Buffett-lens X-ray of KSPI at $98 — the first company on our board from outside the U.S., Canada and Britain, and the one Mohnish Pabrai calls a 'heads I win, tails I don't lose much' bet. Kaspi is what a payments network, a marketplace and a consumer bank look like when one company builds all three inside a single app and 75% of a nation uses it daily: a 41% return on equity, take rates rising, cost of risk under 1%. Against that: net income flat for three quarters, a 25% bank tax, a state-run QR system launched in July, a currency that lost 15% in 2024 and gained 17% since, and a Turkish acquisition that cost the dividend for a year and is losing money faster this quarter than last. We test Pabrai's tails.

A Buffett-lens X-ray of AVGO at $362: Hock Tan's toll bridges — RF filters, switch silicon, mainframe and VMware software — now bolted to the fastest-growing business on Earth: custom AI accelerators for Google, Meta, OpenAI, Anthropic and two others. Q3 FY2026 revenue +86%, AI +221%, a 68% operating margin, $13.7B of free cash flow in a quarter. Against it: 45× trailing earnings (19× FY2027, 12× the FY2028 promise), six customers who own the architecture and re-bid every generation, Google formally multi-sourcing, gross margin sliding 78% → 73%, and a $35B customer-financing vehicle Broadcom partly backstops. We separate the bridge from the promise and name the price at which both are worth owning.

A Buffett-lens X-ray of LULU at $96: the purest test of 'See's Candies or Dexter Shoe?' on our board — a single brand, a single category, no wholesale cushion. Q2 FY2026 brought a −9% comp, women's leggings down ~20%, a China engine that stalled in one quarter, and the third guidance cut of 2026 — dressed up as an EPS 'beat' that was entirely a tariff refund. Against that: $1.4B of cash, zero financial debt, a 19% operating margin even now, 8× trailing earnings and a founder-refreshed board around a brand-new CEO. Cheap, not safe. We name the price at which it becomes both.

Our second X-ray of MCD, eight weeks after the first: a scorecard of what we said at $265 against what happened — the Q2 'bad trade' (+0.8% U.S. comps on −4.5% visits), only 60–65% of franchisees following the price list, July comps negative, the 50,000-unit target pushed to 2028 — set against a franchised margin that did not move and a 50th consecutive dividend raise due 22 October. At $253 and a 2.9% yield, the thesis held, the price fell into our add-zone, and one dial came down honestly.

We published British American Tobacco in July and called it income with eyes open. Altria is the American answer to the same question, and it is a more extreme version of it: no international diversification, no growth of any kind, negative shareholders' equity, and one of the highest operating margins of any company we have examined — 55.9 cents of operating profit on every dollar of revenue. The dividend was raised four days before we priced this, to $4.44 a year, which is a 6.47% yield covered 1.12 times by free cash flow. The bear case writes itself and is entirely correct: this company's product kills its customers and fewer of them buy it every year. The interesting question is narrower. Volume declines have moderated for four straight quarters, nicotine pouches are now sixty per cent of the American oral tobacco market, and Altria's own alternatives have been a catastrophe — $12.8 billion paid for JUUL, written down to $450 million, followed by a vape business its rival has had banned from the country on patent grounds.

This is the closest thing to a genuine mispricing we have found in a month of looking. S&P Global earns a 46.3% operating margin, converts more than a hundred per cent of its earnings into free cash flow, and holds a position in the world's financial plumbing that is very close to unassailable: a company issuing a bond needs a rating and there are two names that count, and anyone wanting to run money against the S&P 500 must pay S&P Global for the privilege. The shares have fallen about a fifth from February's high on a fear that has a name — that AI will replace the Capital IQ terminal — and a size, which is roughly six per cent of revenue. We think the fear is real and the discount is disproportionate. What gives us genuine pause is something quieter and almost nobody writes about: since the $44 billion IHS Markit merger in 2022, revenue is up 85% and earnings per share are up seventeen. The company bought scale. It has not yet turned it into value per share.

The insurance instalment of our metric series, and the one that matters most, because float is the concept that made Berkshire Hathaway and we have somehow never explained it. Chubb is the best large underwriter in the world by the only measure that counts: it pays out 83.8 cents of every premium dollar against a US industry average of 92.9, and it has done something like that for two decades under the same chief executive. Book value per share compounded 12.3% in the last year. Earnings per share have gone from $8.87 to $25.73 in nine years while the share count fell 15%. Meanwhile four of the numbers on its screen — a bankruptcy score of 1.08, a return on invested capital of 3.5%, an EV/EBITDA of 12 and a discounted cash flow value of $724 — are all category errors, for reasons that take one paragraph to explain and that almost nobody explains. The problem is none of that. The problem is that at 1.74 times book, against a ten-year median nearer 1.36, the market has worked out how good this is.

Almost every number on this company's screen looks attractive. Eleven times earnings. A 9% free cash flow yield. A 4.63% dividend covered twice over, raised for forty consecutive years. Net cash on the balance sheet and an Altman-Z score of 6.7. And in the second quarter it reported the largest pool of assets in its ninety-year history. The problem is what is underneath: of the $183.7 billion by which assets grew, $190.2 billion came from markets rising and minus $6.5 billion came from clients — and the outflow is concentrated entirely in fundamental active equity, which happens to be 48.6% of everything the firm manages. Management's own guidance is that the second half will be, in its words, meaningfully more challenging. So the question is not whether this is cheap. It is whether a 4.6% dividend and a share count shrinking at five per cent a year can outrun a business that is being slowly taken apart by index funds — and, if it can, whether that is an investment or merely an orderly liquidation.

This is the deliberate counterpart to our NNN REIT analysis, published the same day, and it is an argument about what a multiple is actually for. Agree Realty owns 2,825 buildings in all fifty states, 99.8% of them occupied, leased to Walmart, Tractor Supply, Dollar General, Hobby Lobby and O'Reilly. Two-thirds of its rent comes from tenants with investment-grade credit. It grew adjusted cash flow per share by 7.7% in the first half and raised its investment guidance to $1.8 billion. It costs a quarter more per dollar of cash flow than NNN and yields a full percentage point less — and once you add growth to income, it is the better proposition of the two. What earns it a genuine discount is not the multiple but the machine: Agree's growth is funded by continuously issuing shares. It raised $686 million of forward equity in six months. NNN funds most of its growth out of rent it has already collected. One of those engines needs a share price. The other does not.

This is the most genuinely difficult company we have written about this month, because everything in it points two ways at once. The bear case is that the whole thesis has broken: Rexford commanded a premium multiple because it could re-let warehouses at enormous increases, and it is now re-letting them at a discount. The bull case is that the Southern California industrial market has just posted its first improvement since 2022 — vacancy down, absorption strongly positive, and new construction at half a per cent of existing stock while state legislation makes future supply harder still. What is not in dispute is the governance: Elliott Investment Management built a stake, both founder co-chief executives departed on 31 March, and the new management has put $2 billion of the portfolio up for sale and started a $1 billion buyback. The $625 million impairment that came with it is the price of admitting that not all of it was irreplaceable after all.

We published Realty Income in June. This is its smaller, older, duller sibling — and on every valuation measure, the cheaper one. NNN owns 3,692 buildings averaging 10,700 square feet, in all fifty states, leased to the businesses you cannot order on the internet: car washes, tyre shops, convenience stores, day-care centres. Occupancy is 99.1%, the highest in the company's history. The weighted average debt maturity is 10.1 years, which management says is nearly double the nearest peer. And the whole enterprise is a spread machine: it buys buildings at a 7.3% initial yield against a blended cost of capital of about 6.7%, and the difference, multiplied by volume, is the entire business. Understand that sentence and you understand every net-lease REIT there is — including why this one deliberately grows slower than Realty Income, and why that might be the point.

Animal health is structurally one of the best businesses that exists: no insurer negotiating the price, no patent cliff of the human-pharmaceutical kind, a customer who pays cash because the dog is family, and a distribution channel of veterinarians who are hard to reach and harder to replace. Zoetis compounded earnings per share at 15.5% a year for a decade on those advantages. Then, over sixteen months, the shares halved. This is an attempt to separate what actually broke from what merely frightened people. Along the way we found something we have not seen written anywhere: Librela, the product at the centre of the noise, is an anti-NGF monoclonal antibody — and the human version of that same mechanism was voted down 19 to 1 by an FDA advisory committee in 2021 over accelerated joint destruction, then abandoned. By Pfizer. The company Zoetis was spun out of. That is the most interesting fact here, and it is still not the reason the stock fell.

The cyclicals instalment of our metric series. Deere's screening profile is a museum of category errors: a bankruptcy score in the grey zone, net debt of roughly five times EBITDA, a price of 54 times free cash flow and a discounted-cash-flow value of $199 against a $649 share price. Every one of those is an artefact of consolidating a $54 billion lending book into a tractor company. Strip the bank out and the manufacturer carries about $2.6 billion of net debt against $21 billion of equity. Meanwhile the real story is a cycle that has almost certainly bottomed: combine unit sales across America fell 51% in two years, used high-horsepower inventory is down nearly 40% from a year ago, and on 20 August Deere said its model-year-2027 early-order programmes for planters and sprayers came in up mid-single-digits on last year. The freeze on the dividend is the same freeze the company ran from 2014 to 2017, and it did not cut then either. All of which leaves one question, and it is a question about price rather than about quality.

We put this one on the shortlist saying openly that it looked cheapest and was probably a trap, and promised to test that rather than confirm it. The answer is more interesting than either caricature. Pfizer's own investor deck says it expects roughly $17 billion of annual revenue to disappear by 2030 as patents expire — 27% of last year's sales — with the step arriving in April 2028 when Eliquis goes generic. Add the pipeline up generously and it offsets about $13 billion of a $20 billion hole. Meanwhile the $43bn Seagen acquisition has produced $8.3 billion of write-offs in under three years, the company paid double its own opening bid for an obesity asset four months after writing off its own, and the dividend was not covered by free cash flow last year. And yet: nine times adjusted earnings, a 6.4% yield, leverage of 2.7 times rather than the 4.9 the screens show, and the single most valuable product in the portfolio still growing 14%.

The utilities instalment of our metric series, and the clearest one yet. NextEra's Altman-Z score of 1.08 puts it nominally in distress; its free cash flow is minus $11 billion a year; its return on invested capital is 4%. All three are category errors — the score was built in 1968 on 66 manufacturers, negative cash flow is what a utility's business model looks like when it is working, and a regulated utility does not earn a return, it is awarded one. Florida's regulator authorised 10.95% and Florida Power & Light earned 11.70%. Meanwhile the company is in the middle of the largest utility merger in American history, buying Dominion for around $67 billion in stock — which puts 36% of its rate base in Virginia, the world's densest concentration of data centres. And buried in the guidance is the fact almost nobody has written about: the famous 10% dividend growth rate steps down to 6% from the end of this year.

The first genuinely cheap wonderful business we have found in a while. LVMH is 49% below its April 2023 peak because Chinese luxury demand collapsed and the aspirational customer walked away. But look at what actually happened this summer: Fashion & Leather Goods — Louis Vuitton and Dior, 71% of group profit — returned to organic growth in the second quarter, ending seven straight quarters of decline. The operating margin held at 22.5%. Free cash flow last year was the second-highest in the company's history. Seventeen analysts say buy and none says sell. And the family that controls it took its stake past 50% while the shares were falling. The catch is that profit is still a quarter below its peak, the return on capital has been dragged down by what was paid for Tiffany, and Bernard Arnault is 77 with five children and no announced successor.

A Buffett-lens X-ray of AMAT at $507. Applied Materials makes the deposition, etch, implant and inspection machines that turn silicon into transistors — it touches almost every chip made anywhere, and it has quietly retired 28% of its own shares in nine years while raising the dividend nine years running. Q3 FY2026, reported two days ago, was the best quarter in its history: revenue $9.12bn (+25%), gross margin above 50%, and a Q4 guide of $10.25bn — up 51% on last year. And still the stock fell 5%, because it had already risen 167% in a year and now costs 43× trailing earnings and 64× free cash flow. Three things sit underneath that price: a moat built on breadth rather than monopoly, a customer base whose home countries supply 74% of revenue and whose politics can be switched off by decree, and the oldest trap in cyclical investing — a high multiple on earnings that are themselves at a high.

The cannabis landlord, mapped and counted. IIPR owns 111 buildings across 19 states, and we have drawn where every one of them sits — including the Washington property that collected no rent at all last year. The business exists because of a prohibition: cannabis is federally illegal, banks will not lend, and IIPR stepped into the gap charging rents no ordinary industrial landlord could dream of. It worked spectacularly until the tenants ran out of money. Now revenue has fallen 14% from its peak, the dividend has been frozen for nine quarters and exceeds the cash the business generates, and the company is quietly lending $270 million to a life-science REIT because it can no longer find cannabis deals worth doing. And the catalyst everyone is waiting for — federal rescheduling, which partly arrived in April — may be the worst thing that could happen to it.

A business that is paid to hold other people's money, examined honestly. Paychex's own annual report contains the most revealing table we have found this year: $100 invested five years ago is worth $111 today, against $194 in the S&P 500 — and its peer group did worse still, down 38% in a single year. The entire 'boring compounder with a compliance moat' category has been repriced for AI disruption. Yet Paychex's client retention is identical to last year, revenue per client is rising, it names price realisation as a growth driver, and margins are expanding. The AI effect is visible entirely in the multiple and not at all in the operating results. The catch: the shares have rallied roughly 40% off their low and now trade above every single analyst price target on the street.

Written two days after the results the whole thesis depended on — and the market's answer was a 10% rally. Diageo is still 53% below its January 2022 peak, has cut its dividend in half, and took $2.4bn of exceptional charges. But look at what it actually owns: $8.7 billion of maturing spirit — 81% of all its inventory — of which $5.7 billion is Scotch. You cannot decide today to sell 18-year-old Johnnie Walker; that decision was taken in 2008. No competitor with unlimited capital can close that gap, because the barrier is measured in years. The trouble is that the same law protecting the moat forbids moving production to escape a tariff, the flagship brand lost 6% of its volume last year, and the rally has left barely 3% between the price and the average analyst target. Includes the fact almost nobody knows: Buffett's first foreign stock was Guinness — and he sold rather than take Diageo shares.

The supplier to everyone in drug discovery, examined honestly. Roughly 42% of revenue is genuine razor-blade consumables locked in by regulation — switching a validated cell-culture medium is not a purchasing decision, it is a filing amendment. That business earns 37% margins. But it is 23% of the company. The other 56% is contract research and manufacturing at a 14% margin, and that is where the biggest cheques went: $33.5bn into Patheon, PPD and Clario. Five years on, revenue is still below 2022, GAAP earnings per share are still below 2021, and free cash flow has not grown since 2020. Return on invested capital including goodwill is roughly its own cost of capital.

The twin study to our Coca-Cola report, and the most rigorous look at dividend sustainability we have published. PepsiCo is not really a drinks company: Frito-Lay earns a 22% operating margin and roughly 45% of divisional profit, while beverages in North America have collapsed to a 3.86% margin. The stock sits near its 52-week low on a 4.1% yield with more than fifty consecutive years of increases behind it. But free cash flow has been flat at $7–8 billion for six years while the dividend grew from $5.5bn to $7.6bn — coverage has fallen from 1.20× to exactly 1.00× — and the company cut snack prices by up to 15% only to see volumes stay flat.

Our first Business Development Company, and a teaching piece: how to read a BDC, where the conventional ratios actively mislead, and what you are really buying. Main Street is genuinely the class of its sector — internally managed while nearly every rival pays an external adviser 1.5% of assets plus 20% of profits, a saving worth roughly $90 million a year. Its monthly dividend has never been cut since it listed months before the financial crisis. But it trades at 1.6 times net asset value when most of its peers trade below book, its investment income has been flat for six quarters as rates fall, and most of its book-value growth has come from issuing shares above book rather than from lending.

The purest test on our board of whether generative AI destroys software moats. Adobe owns the industry-standard creative tools and invented PDF, earns an 89% gross margin and a 36% return on invested capital — and trades at fourteen times earnings and ten times free cash flow, cheaper than Salesforce despite fifteen points more operating margin. Revenue has grown 60% since 2021 while the shares fell 64%. But the growth is genuinely decelerating, management has deferred planned price increases, it has begun giving the low end away free, it stopped disclosing the metric everyone was watching, and both its CEO of eighteen years and its CFO left within three months.

The companion to our Boeing report, and the same disease at a different stage. Lockheed is the pure-play defence prime that is supposed to do fixed-price development well — yet it has taken four reach-forward losses in three years, twice on the same classified programme. Its F-35, a quarter of all revenue, delivered 110 aircraft in one year at an average of 238 days late, some with ballast in the nose where the radar should be. Its fighter backlog has fallen three years running. And the stock beat the S&P 500 by nineteen points, because missiles are booming and the market is looking at the arsenal instead of the aeroplane.

The first Industrials analysis on our board, and the definitive case study of management destroying a moat. Boeing owned half a global duopoly protected by certification barriers, 20-year backlogs and switching costs no rival could breach — then subordinated engineering to financial engineering, spent $43.4 billion on buybacks instead of the aeroplane it needed to build, and had to sell shares at $143 to survive. Two crashes, a door plug, a criminal charge and $33.7 billion of cash burn later, the backlog is at an all-time record and the repair is genuinely working. The question is what you pay for a wreck that is being fixed.

The contrarian close to our payments trilogy. PayPal was the future of money; then Apple Pay and Shop Pay ate its checkout button, its digital-wallet share fell from 90% to 40%, its turnaround CEO was fired after 29 months, and the stock lost 82% from its 2021 peak. It now trades at ten times earnings with an 11% free-cash-flow yield, net cash, and a buyback retiring 5% of the shares a year — the textbook 'value trap or opportunity' dilemma. And in the middle of writing this, the tiebreaker arrived: Stripe and Advent bid $53 billion for the whole company, and the board turned it down as too cheap.

The companion to our Visa report, and the second front in the payments war. Mastercard is the #2 rail in the world's most durable duopoly — the same toll booth that takes no credit risk and keeps none of the interchange, clipping a few basis points on $10.6 trillion of volume. It grows faster than Visa (+16% vs +11%), reaches further (57% international), and leans harder on a services business now 41% of revenue. It has always cost more than Visa for that — but by 2026 the premium has all but disappeared, and its two great litigation clouds have cleared. A study in choosing between two nearly identical wonderful businesses at nearly identical full prices.

Our third REIT, and nothing like the first two. VICI owns the dirt beneath Caesars Palace, the Venetian, MGM Grand, Mandalay Bay, Luxor and Excalibur — about 660 acres of Las Vegas Boulevard — and rents it back on triple-net leases averaging almost forty years. It collected 100% of its rent in cash through 2020 and 2021, when comparable landlords collected as little as 15%. It employs roughly twenty-eight people. It now trades at 1.02 times book value on a 6.7% yield, because its largest tenant is being taken private in a debt-heavy buyout and a performance-linked rent reset on 38% of the rent roll arrives in November 2027.

The deliberate companion to our Costco report. Same aisle, opposite model: Costco charges a fee and sells goods near cost; Walmart sells goods for a thin profit and is quietly becoming an advertising business. Advertising and membership now supply about a third of Walmart's operating income — yet across the last ten years revenue grew 47% while operating profit grew just 31%, and the operating margin actually declined. The stock re-rated from ~15× earnings to 40× anyway. A study in what happens when a magnificent operator meets a magnificent multiple.

The companion to our Novo Nordisk report, whose verdict was titled 'Accumulate, Eyes on Lilly.' Here is the other side. Lilly has beaten Novo in every direct comparison that matters — including REDEFINE-4, a trial Novo designed and lost — and became the first healthcare company on Earth worth a trillion dollars. But in Q1 2026 its realised prices fell 13%, it has contractually accepted a price ceiling on every future medicine it will ever launch, and a single pharmacy middleman deleted its clinically superior drug for fifteen months. A study in what happens when you win a war whose spoils are set by the government.

A Buffett-lens X-ray of CVX at $188 — the highest-quality Western oil major: the lowest-cost Western barrels, a Guyana crown jewel that breaks even near $25, and a dividend raised 39 straight years through five oil crashes. But it is a cyclical commodity business earning a price it doesn't control, trading up 28% this year on an Iran-driven war premium, at 32× trough earnings and above fair value — while Berkshire, its marquee owner, quietly sells. A study in a wonderful dividend at a treacherous entry point.

A Buffett-lens X-ray of the master's own company — written, for once, about the man whose voice we borrow. Berkshire at $502: a ~$1.08 trillion fortress with $397 billion in cash, an insurance float of ~$176 billion, and the greatest long-run record in the history of capitalism (19.7%/yr for 60 years). And a new hand on the tiller — Greg Abel wrote the annual letter this year, the first not signed 'Warren' since 1965, and promptly sold Visa, Mastercard, Amazon and UnitedHealth and tripled Alphabet. Is it still Buffett's Berkshire? The most important analysis we have ever attempted.

A Buffett-lens X-ray of COST at $956 — arguably the finest retail business ever built: ~92% membership renewal, a subscription fee that is roughly half of all operating profit, suppliers who finance the shelves, and a culture Charlie Munger called the company he most admired. The catch is the one he named himself — at ~48× earnings it has never been cheap, and a 21% crash last winter didn't change that. The definitive study in a wonderful business at an unwonderful price.

A Buffett-lens X-ray of UNH at $423 — the largest US healthcare company, recovered ~80% from a self-inflicted collapse that halved its earnings, killed its CEO in the street, and drew a federal criminal probe. The Q2 blowout this week proved the operational turnaround is real. But the DCF now says overvalued, analysts see +7%, and Berkshire — which bought the panic in 2025 — sold every share this spring. A wonderful franchise whose fat pitch already passed.

A Buffett-lens X-ray of IBM at $219, six days after the worst single day in its 115-year history — a 25% crash on a software warning that terrified the whole sector. The wreckage exposes a $12 billion free-cash-flow machine at ~14× the cash, a 3% dividend raised for 31 straight years and paid since 1916, a DCF at +85% — and one binary question the market can't answer until the earnings call on the 22nd: was the miss a deferral, or the AI boom eating IBM's own software budget?

A Buffett-lens X-ray of NKE at $45: the most valuable apparel brand on Earth, down ~75% from its 2021 peak to a decade low — profit halved, running-share bleeding to On and Hoka, China falling, and a 32-year veteran CEO promising a turnaround that isn't visible yet. A ~3.7% yield (its highest ever), one raise short of Dividend Aristocrat — and the exact question Buffett got wrong on Dexter Shoe: is a sneaker brand a durable moat or a fashion cycle?

A Buffett-lens X-ray of NVO at $51: the company that made Ozempic and Wegovy, down ~63% from its 2024 peak — 82% gross margins, 66% on equity, a foundation-owned forever-holder, and a DCF that says $89. The catch is real: a guided 2026 revenue decline (its first in the modern era), a patent cliff, and Eli Lilly winning the efficacy race. Weeks after we called GLP-1 the central risk to McDonald's, the molecule-maker trades cheaper than the burger chain.

A Buffett-lens X-ray of MCD at $265: 95% franchised, 46% operating margins, and the secret Wall Street forgets — most of the profit is RENT on corners McDonald's has owned for decades. At a fresh 52-week low, ~20× forward, 2.8% yield with the 50th consecutive raise due in October — and, for the first time on our board, a DCF that sits ABOVE the price. Buffett sold it in 1998 and called the decision 'a very big mistake.'

A Buffett-lens X-ray of META at $681: 3.5 billion people a day feed an ad auction earning 41% margins and 33% on equity — priced at ~21× next year's earnings, cheaper per unit of growth than anything its size on our board. The catches are real: a $125–145B debt-assisted AI bet, paused buybacks, a founder with 61% of the votes, and a $1.4 trillion courtroom demand this August. The machine pays for all of it.

A Buffett-lens X-ray of TSLA at $394: the company that made electric cars real — and a stock at 328× earnings while deliveries decline, margins halve, and the profit engine (regulatory credits) dies by law. The car business earns about a dollar a share; everything above ~$30 is robotaxi and robot promises. We admire the builder. We pass on the price.

A Buffett-lens X-ray of AMZN at $247: the everything store, the toll booth on two million sellers, and the railroad of the AI era — revenue up 5× in a decade, profit up 33×, and free cash flow now negative as a ~$200B capex bet swallows every dollar. A wonderful flywheel that has never paid its owners a cent — and asks for a decade of faith.

A Buffett-lens X-ray of BofA at $58: one of America's Big Four banks, built on a ~$2-trillion low-cost deposit base — banking's version of float. ~14× earnings and 1.4× book with a ~13% return on tangible equity. A fine, cheap-looking bank — but banks are leveraged and cyclical, it sits near its highs, and even Buffett has been trimming.

A Buffett-lens X-ray of Amex at $348: the closed-loop, spend-centric network — it issues, lends, and sees both sides of every swipe. A premium-brand moat around affluent spenders, Berkshire's untouched ~21% stake, and ~21× earnings for double-digit growth — cheaper than Visa, but it carries credit risk Visa never touches.

A Buffett-lens X-ray of Apple at $294: the most valuable company on earth and one of the great consumer franchises — a brand-and-ecosystem moat, a fortress balance sheet, and the largest buyback in history. The catch is the price: ~35× earnings for mid-single-digit growth, and even Buffett trimmed.

A Buffett-lens X-ray of Salesforce at $157: the world's #1 CRM, ~43% below its high on AI-disruption fear — now ~10× free cash flow, ~17× earnings, with a deep switching-cost moat and a fixed-up balance of margins. A wonderful franchise on sale, if the fear is overblown.

A Buffett-lens X-ray of BAT at $60: perhaps the clearest case of a wonderful business in a terrible industry — a penny-to-make, dollar-to-sell cash machine yielding ~5.3%, but with declining volumes, £87B of acquisition goodwill and an ESG stain. The easy money has already been made.

A Buffett-lens X-ray of Visa at $343: perhaps the purest toll booth in public markets — a 51% net margin, a two-sided network no rival can rebuild, and double-digit growth as cash keeps dying. The catch is the price: ~30× earnings leaves a thin margin of safety.

A Buffett-lens X-ray of ARE at $53 — a deep-value special situation. The pioneer of lab real estate trades at ~0.5× book after a 71% crash, a 45% dividend cut and $1.4B of impairments — yet founder Joel Marcus is buying. Contrarian bargain or value trap? A cheap asset with still-falling fundamentals.

A Buffett-lens X-ray of O at $62 — and a lesson in reading a REIT. The headline P/E of 51× and 265% payout are illusions; on the right metric (AFFO) it trades at ~15× with a well-covered 5.2% monthly yield. A dependable, bond-like income compounder at a fair price — own it for the cheque, not the capital gain.

A Buffett-lens X-ray of GOOGL at $357: a toll-booth on the world's information — ~90% search share, YouTube, a top-tier AI lab and Waymo — now at a fair ~27× earnings, but facing its first real test: does AI answer-engine search monetize as well as ten blue links? A wonderful franchise, watchfully owned.

A Buffett-lens X-ray of NVDA at $200: the best economics on the board — 63% net margins, revenue doubling, a real CUDA moat — but a cyclical, concentrated, unknowable decade and a $4.85T price that assumes permanence. A magnificent business that lands in the 'too hard' pile.

A Buffett-lens X-ray of KO at $80: the simplest 5/5 business on the board — a brand moat with 60+ years of dividend raises and Berkshire's 9% stake — but 25× earnings for ~6% growth keeps it a fair, not cheap, forever-hold.

A Buffett-lens X-ray of MSFT at $367: fortress balance sheet and a wide moat, now ~22× earnings — but owner earnings (~30×) and a thin margin of safety say keep watching.