App Economy Insights — This Week in Visuals (E06): 36 Q2 prints

Apple had record June-quarter sales and the shares still fell 6%, because the problem is no longer demand — it's that Apple cannot build enough. Demand for iPhones and Macs came in higher than planned, and Apple had already committed its supply of the most advanced chips, so it has nothing left to flex. That shortage hits Macs now and spreads to iPhone and iPad next quarter, which is why the sales forecast came in light despite a strong product cycle. The other squeeze is memory chips: Tim Cook called the price surge a "hundred-year flood," Apple has already raised some Mac and iPad prices, and profit margins are set to step down again once the cheaper chips it bought earlier run out. This was Cook's last earnings call — John Ternus takes over in September and inherits both the strong cycle and the supply and margin pressure. A recap, not a call.

Free cash flow is what a company has left after paying for the buildings and equipment it needs. Meta earned $31.9 billion of cash from operations this quarter and spent $31.1 billion on AI data centres — leaving $784 million, a 91% collapse. It also borrowed $24.9 billion and bought back no shares at all. Nothing is broken; Meta can easily afford this. But it is the first quarter where the AI buildout ate the entire quarter's spare cash, and the spending floor keeps rising (now $130–145 billion for the year). Watch how it is now funding the next stage: for a new one-gigawatt campus in El Paso, BlackRock puts up the money and owns 80% while Meta owns 20% and rents the whole site — the commitment is the same, the upfront cash outlay is smaller. The consolation is that AI is already earning its keep inside advertising: 8% more ad clicks, 16% more conversions on Facebook, and its automated-ads product past a $75 billion annual pace. A recap, not a call.

Samsung just posted the best quarter in its history — revenue more than doubled and operating profit rose eighteenfold — almost entirely from memory chips, whose prices are soaring because AI systems need enormous amounts of them. The stock fell anyway, and is down over 40% from its June peak. That gap is the whole story of this issue. Management said the shortage now lasts through 2028 and has locked 60–70% of its most advanced memory capacity into multi-year contracts, so the near-term earnings look assured. But memory is a famously boom-and-bust business, and investors are refusing to pay up for peak-cycle profits they assume will reverse. Note the irony inside Samsung itself: its phone division swung to a loss, squeezed by the very component prices making its chip division rich. A recap, not a call.

Visa doesn't lend money; it runs the rails that card payments travel on and takes a small cut of everything that passes through. So the numbers that matter are volumes, and every one of them sped up: total payment volume passed $4 trillion in a single quarter for the first time, cross-border spending (the most profitable kind) grew 13%, and transaction counts accelerated. Two things to notice. Visa pays partner banks "incentives" to win their business, and those payments grew 18% — faster than revenue — which is the price of the volume growth. And on the same day it announced a stablecoin platform and an AI-shopping partnership with OpenAI, it cut about 2,600 jobs, 7% of staff, mostly in technology and product. Management trimmed the full-year outlook to the low end, blaming currency swings it expects to persist. A recap, not a call.

Last quarter investors worried that Middle East conflict was starting to dent cross-border travel spending — the highest-margin thing Mastercard processes — and that Mastercard, with more exposure there than Visa, would suffer most. It didn't: the impact moderated, cross-border growth held at 12%, and the extra exposure became an advantage. The more durable shift is that Mastercard is becoming less of a pure payments network and more of a services business — fraud prevention, cybersecurity and data products grew 20% and are now the faster half, helped by its Recorded Future acquisition. Margins widened to 61%, guidance went up, and it bought back $4.9 billion of stock. A recap, not a call.

AbbVie's old blockbuster, Humira, lost patent protection and is shrinking fast (down another 36%). The whole investment question is whether its two replacements grow faster than Humira falls. This quarter they did comfortably — Skyrizi up 24% and Rinvoq up 25%, and notably Skyrizi has not lost momentum even after Johnson & Johnson launched a rival pill. Brain and nervous-system drugs added a second engine, up 20%. The odd detail: AbbVie raised its sales forecast but cut its profit forecast, because it is buying Apogee Therapeutics for $10.9 billion and the deal dilutes near-term earnings. That's a deliberate trade — spending today's profit to extend the immunology franchise into the 2030s. A recap, not a call.

Lam sells the machines chipmakers use to build chips, so its orders lead the industry by a year or more. While most chip-related stocks sold off this season on fears that AI spending is peaking, Lam rose — because it guided next quarter to $8.1 billion against roughly $7.1 billion expected, more than 20% growth in a single quarter. That is a company telling you the buildout is accelerating, not fading. The demand is broadening too: sales into NAND flash memory doubled in one quarter, and advanced packaging (the technique used to stack memory next to AI processors) is guided to grow more than 70%. Management even raised its long-run profitability targets. The only constraint left is physical — whether customers can install tools as fast as Lam can ship them. A recap, not a call.

For years consumer-goods companies grew by raising prices while selling fewer units. Coca-Cola is deliberately doing the opposite: it held price increases to just 2% and let volume do the work — 5% more drinks sold, growth in every region, led by India, China, the US and Brazil. That's a healthier kind of growth, because it means people are actually buying more rather than paying more for less. Remarkably, profit margins still widened, and the World Cup sponsorship lifted core Coke and Powerade volumes. The stock hit a new high. Two known drags are coming: six fewer shipping days in the fourth quarter, and the sale of its African bottling arm, worth a 2–3% hit to reported revenue. With price growth capped on purpose, volume has to keep carrying it. A recap, not a call.

P&G's underlying sales were completely flat — no growth from volume, price, or mix — a clear step backwards from last quarter, with every division except beauty falling short. But the guidance is where the damage sits. The Iran war has pushed up the cost of raw materials, energy and shipping, and P&G says that will cost it about $1 billion after tax next year — roughly an 8% bite out of profit growth — with the forecast assuming oil stays near $90 a barrel, so any escalation makes it worse. The bull argument is that this is a macro problem, not a competitive one: P&G just gained share in Greater China for the first time in fifteen quarters. The test is the first half of next fiscal year, since management has promised sales will accelerate while it absorbs the worst of the cost hit up front. A recap, not a call.

Arm doesn't make chips. It designs the blueprints and collects a royalty every time someone else's chip using its design is sold — which historically meant smartphones. Phones are slowing, and that is why the stock fell 7–8% despite beating on every line. The bull case rests on the shift into data centres, and that part is working: royalties from data-centre chips more than doubled for the second quarter running, its server designs have shipped 1.5 billion processor cores, and Nvidia, Amazon, Google and Microsoft are all leaning in further. The number to watch is a mismatch of Arm's own making — it says the pipeline of demand for its AI-server processors now exceeds $2 billion, but its official forecast still counts only $1 billion of it, because that's all the manufacturing capacity it has secured. If it raises that figure, the data-centre story is real enough to outrun the phone slowdown. A recap, not a call.

KLA makes the inspection equipment that finds defects on chips during manufacturing — a small slice of the industry's spending, but one that grows when chips get harder to make, which AI chips certainly are. Its most useful disclosure is its estimate of what the whole industry will spend on equipment: it just raised that for 2026 to about $150 billion, from "$140 billion plus" three months ago, and said AI demand signals have strengthened materially since March. A caveat worth understanding: KLA buys memory chips for its own tools, those prices are rising, and it cannot reprice orders customers already placed — so roughly a 1% margin drag is locked in until 2027. The question is whether serving a bigger market comes at a structurally lower margin. Note the split figures reflect a ten-for-one stock split in June, so per-share numbers aren't comparable to older ones. A recap, not a call.

A drug company is worth what its pipeline will eventually sell, so a high-profile trial failure raises the question of whether everything behind it is also shakier. AstraZeneca's heart drug Wainua missed its main goal — rare for a company that claims a 75% trial success rate — and the stock is down this year as a result. This quarter was the rebuttal. Cancer drugs carried the business (Imfinzi for lung cancer up 27%, Enhertu for breast cancer up 31%), offsetting older drugs going off-patent, and the CEO pointed to positive results from six late-stage programmes while insisting the $80 billion 2030 revenue goal already assumes failures like Wainua and needs no acquisitions to reach. More than twenty trial readouts land over the next eighteen months; two — Datroway in lung cancer and camizestrant in breast cancer — will decide whether that target is credible. A recap, not a call.

Airbus has cut its delivery target late in the year three times in four years, so the only question that mattered was whether it would do it again. It didn't: it delivered 351 aircraft in the first half against 306 last year, with the acceleration concentrated in the second quarter after a dismal first, and profit jumped 54%. That was enough to make the full-year goal of about 870 look reachable. Two things to keep watching. Cash flow was negative €1.2 billion because Airbus is deliberately building inventory ahead of the ramp, and a shortage of Pratt & Whitney engines still limits how fast it can go. And the arithmetic is demanding: roughly 519 more aircraft have to ship in the second half to hit 870, so either the monthly rate holds or the back-half crush strains suppliers again. A recap, not a call.

Qualcomm has one enormous known problem: Apple is replacing Qualcomm's modem chips with its own, and this quarter Qualcomm admitted the loss is coming faster than it had said — its share of the new iPhone will be "materially lower" than the 20% previously guided. The company's whole answer is diversification, and that part is genuinely working: car chips grew 61%, and two custom data-centre silicon wins start generating revenue in December, letting it raise its 2029 non-phone target to $40 billion. The near-term pinch is the same memory shortage hurting everyone — it compresses Qualcomm's chip margins, which is why next quarter's profit guidance came in well below expectations, and why Qualcomm is pushing through double-digit price increases from September 1. The diversification is ahead of plan; it just has to outrun a modem business disappearing faster than expected. A recap, not a call.

For Boeing, the single number that matters is cash: after years of burning it, the company generated $631 million this quarter versus losing $200 million a year ago. Deliveries hit 171 aircraft, the most since 2018, and the order backlog reached a record $715 billion — meaning demand is not the issue; the factory is. Boeing is lifting 737 output from 42 a month toward 47 and then 52, and that ramp is what the full-year cash target depends on. Two honest asterisks. The quarter's cash was helped by roughly $1.5 billion of customer prepayments, and a $700 million legal payment to the Justice Department hits next quarter. And the fixed-price Air Force One contract took another $280 million charge — it is four years late with more than $1 billion of overruns that Boeing, not the government, absorbs. A recap, not a call.

"Comparable sales" measures growth at stores open more than a year — it strips out new openings, so it shows whether the existing business is actually getting better. Starbucks grew 8% globally and 8% in the US, with a quarter of that from more customers walking in rather than higher prices. That's the fourth straight positive quarter after seven negative ones, and it accelerated. Revenue fell 1% only because Starbucks handed 60% of its China business to a partner and no longer consolidates it. The caution is on margins: the headline 4.3-point improvement leaned on tariff refunds covering three quarters of duties plus a favourable tax comparison, and management said the year-to-date view is the fairer read — North America's own margin barely moved, from 13.3% to 13.6%. The test is whether that keeps expanding once the refunds stop. A recap, not a call.

Fortinet sells network security — traditionally firewall hardware, increasingly a cloud-delivered service called SASE that protects employees wherever they work. Every number accelerated from an already-strong quarter: revenue up 26%, new business signed ("billings") up 33%, cash flow more than tripled, and the striking one — 90% of its large enterprise customers now use its SASE product, up from 18% three months ago. The thing to watch is the mix. Hardware revenue grew 52% while the recurring subscription half grew only 14%. Hardware is lumpy, one-off revenue; subscriptions are the durable kind. So the beat was front-loaded with the less repeatable part, and management has also assumed a price increase will help second-half bookings. Subscription growth is the number that tells you whether this is a genuine platform win. A recap, not a call.

UPS spent 18 months deliberately shrinking — walking away from about 2 million parcels a day of low-margin Amazon volume, cutting $4.5 billion of costs and a large slice of its workforce. This is the first quarter where revenue and operating profit both grew again, which suggests the reset worked. Amazon is now 9% of revenue, down from over 13% at the peak. What replaces it is higher-value freight: healthcare shipping passed $3 billion for the second quarter running, helped by refrigerated weight-loss drugs, and automated buildings now handle 68.5% of US volume at roughly 28% lower cost per parcel. The reported profit still looks awful because of an $891 million charge for the job cuts — that's a one-off. With the shrinking done and the costs booked, the second half is where the efficiency should finally show up in profit. A recap, not a call.

Cadence makes the software chip designers use to design chips. The most interesting idea in this issue comes from its CEO: when customers point AI agents at their design work, the agents call Cadence's simulation engines far more often than a human would. So AI shows up not as new software licences but as heavier usage of software customers already pay for — a consumption business rather than a per-seat one, which is exactly the model investors have been penalising elsewhere. Backlog hit a record $8.1 billion and the company made the largest single-quarter forecast increase in its history. Two honest notes: margins in the second half are guided lower because Cadence is deliberately investing, and management still refuses to model any sudden jump from AI agents — it can't even separate out how much of today's growth they represent. A recap, not a call.

The bear case on Robinhood was simple: it's a crypto brokerage in disguise, so when crypto falls, it falls. This quarter tested that directly — crypto revenue dropped 38% — and the platform grew 33% anyway, because options rose 29%, stock trading rose 95%, and event contracts (bets on real-world outcomes) rose more than tenfold to $156 million. Deposits hit a record $22 billion and revenue per customer recovered sharply. New businesses are scaling faster than expected: government-backed "Trump Accounts" pulled in $1.5 billion from 7 million children, and its prediction-market venture with Susquehanna became a top-three US exchange within a month, answering the worry about what happens when football season ends. Management even lowered its expense forecast. And the stock is still down more than 20% this year — which the article reads as a verdict on crypto sentiment rather than on Robinhood's execution. A recap, not a call.

Mondelez (Oreo, Cadbury, Milka) spent the past year absorbing a cocoa price crisis that crushed its margins. That has flipped: its chief operating officer says cocoa is now "a very different place," with a half-million-tonne surplus, and gross margin rose instead of falling. The bigger surprise was North America, which went from crawling at 0.5% growth to 3.4%, with management claiming share gains in every category, while emerging markets kept compounding on sheer distribution — 100,000 new stores added in India, a million now in Brazil. Europe is the new weak spot. The revealing detail is that Mondelez raised its sales forecast but left its profit forecast alone, saying it will reinvest any upside — companies do that when they still see enough risk ahead that they'd rather buy insurance than book the profit. A recap, not a call.

The standout detail in this issue comes from an unlikely place. Hilton's biggest improvement was in mid-priced, limited-service US hotels — a segment that was shrinking last year — and its CEO attributed it to the AI data-centre construction boom: the contractors and engineers building those sites stay in mid-range hotels, not luxury ones. It's a rare, concrete second-order read on where AI capital spending actually lands in the real economy, and it implies that demand persists as long as the buildout does. The offsetting drags are geopolitical and known: Middle East and Africa revenue per room fell about 30% on the Iran conflict, and China fell on government limits to group travel. Hilton raised its full-year outlook, though next quarter's profit guidance came in below expectations. A recap, not a call.

Ferrari's business model is the opposite of every other carmaker: it deliberately sells fewer cars than people want, so the waiting list stays long and prices stay high. That's why it can cut deliveries during a model changeover and still grow profit — the mix got richer, led by the €3.6 million F80 hybrid supercar limited to 799 units, and by personalisation (bespoke paint, trim and options) which now exceeds 20% of revenue. The order book is full through 2027 and Ferrari raised its guidance in a quarter when it almost never does, which is its way of signalling confidence. The open risk is the Luce, its first electric car — a €550,000 four-door designed with Jony Ive's studio that got a cool reception in May. It hit its ~500-unit target for the year, but once the F80 finishes its run, three years of volume rest largely on an EV the market isn't yet sure about. A recap, not a call.

Ford's headline earnings look far better than the business: a one-time $1.3 billion tariff refund flattered them, and stripping the one-offs out leaves $0.37 a share rather than $0.66. Underneath, the pattern is familiar — trucks and SUVs pay for everything else. US sales fell 10%, but Ford sold more high-margin Broncos and Explorers to offset fewer F-150s, and its commercial arm stayed the profit engine, while electric vehicles lost money at an extraordinary rate (a negative 90% margin, roughly $4 billion of losses expected this year). Two second-half risks are worth tracking: a fire at the Novelis aluminium mill drove most of Ford's $1 billion-plus cost hit, and the North American trade agreement (USMCA) is being renegotiated after the US declined to renew it, which matters given Ford's Mexican and Canadian plants. Ford still raised its full-year profit forecast by $1 billion, beating GM's raise the week before. A recap, not a call.

PayPal is in play. In July, Stripe (a private payments company) teamed with private-equity firm Advent to bid about $60 a share, valuing PayPal above $53 billion — and the board reportedly thinks that undervalues it, which is a meaningful judgement given the shares had already run up 35% in a month. Management would only say it will "carefully consider" any path that creates superior value. Underneath the drama the business is steadier than the headlines: Braintree, the behind-the-scenes processing arm, grew 13% for a ninth straight profitable quarter, cash flow funded $1.5 billion of buybacks, and PayPal raised its full-year profit forecast. The weak spot is the part everyone recognises — the branded PayPal checkout button — still growing only 2%. The genuinely unanswerable question is whether that raised forecast is a real turnaround signal or a negotiating position. A recap, not a call.

Coinbase is doing everything right competitively and it isn't enough. Its share of all crypto trading hit a record 10.3%, up from 7.1% a year ago; derivatives volume nearly matched an all-time high; prediction markets more than doubled; and stablecoin balances on the platform reached $20 billion. But revenue still fell 19% and it posted a large loss, because the market it dominates is shrinking — crypto prices are weak and retail traders have gone quiet. Management's answer is the "Everything Exchange," broadening into products beyond Bitcoin trading (Bitcoin is now just 12% of revenue), and it has cut its expense outlook. The honest summary from the article: the diversification is real, it just isn't big enough yet to break Coinbase's dependence on the crypto cycle. Next quarter's subscription revenue is guided well below expectations. A recap, not a call.

Live Nation promotes concerts, owns venues and runs Ticketmaster, so it earns from the same fan three times. A record 49 million people attended its shows, up 10%, and almost all the incremental growth came from outside the US — international markets provided 70% of Ticketmaster's profit growth and 80% of sponsorship's, with attendance at overseas stadiums and festivals up more than 20%. US stadium attendance fell simply because more shows are scheduled later this year. One worry got answered: the crackdown on resale (secondary) ticketing hasn't hurt, because resale is only about 5% of transaction value and Ticketmaster's fee-bearing volumes still grew. Deferred revenue — money already collected for future events — hit a record $6.4 billion, and cancellations fell to 1.1%, both signs demand is real rather than promotional. The antitrust overhang hasn't gone away. A recap, not a call.

Restaurant sales grow either because more people come in (traffic) or because each person spends more (ticket). Chipotle's recovery is the better kind: same-store sales accelerated to 2.2% from 0.5%, and transactions actually rose — so people are coming back, not just paying more. The shares jumped nearly 14%. The trade-off is on the other side of the income statement. Beef, freight and labour are inflating at 3–3.5%, and the CEO is deliberately not passing that on to a stretched customer, so restaurant-level margin fell from 27.4% to 25.2%. Management also warned that trends "softened in recent weeks" and guided next quarter to only about 1% growth against a tough comparison, partly on a food-safety news story Chipotle says doesn't involve it. The top-line recovery is credible; whether margins stop falling is the open question. A recap, not a call.

Yum agreed to sell Pizza Hut for $2.7 billion, removing its weakest brand. The maths of that is flattering: excluding Pizza Hut, system sales grew 7%, units grew 6% and same-store sales grew 4% — a visibly better company centred on Taco Bell and KFC. But concentration cuts both ways, and the risk showed up immediately. After the quarter closed, a cyclospora outbreak traced to supplier lettuce hit Taco Bell, and US same-store sales fell 2% through late July, though management says the last ten days improved sharply. Before that, Taco Bell was exceptional — system sales up 9%, restaurant margins up 1.7 points, nearly half of orders digital. So the next quarter answers a simple question: was the food-safety hit a blip, or the first reminder of what it means to depend this heavily on one brand? A recap, not a call.

Hershey's profit soared 57% and gross margin went from 30.5% to 45.3%, which sounds spectacular until you see how: twelve percentage points of price increases against an eight-point decline in volume. In other words, Hershey is protecting profit by charging more, and shoppers are buying less chocolate. Confectionery sales in North America grew 4.2% purely on price, with double-digit volume declines underneath. Management's bet is that cocoa costs are now falling while the higher prices stick, and that volumes recover as shoppers adjust — helped by an early Halloween. The article's caution is the right one: pricing this far ahead of volume isn't durable growth, and if shoppers don't come back as inflation eases, that twelve-point price lever becomes a ceiling rather than an engine. Watch North American confectionery volume next quarter. A recap, not a call.

Roblox changed what its recommendation algorithm optimises for — from short-term spending to whether a player is still around 28 days later — and it worked exactly as designed, in both directions. Players, especially under-13s in the US and Canada, drifted from last year's heavily monetised viral hits toward newer, longer-lasting games that generate less spending per hour. So "bookings" (money actually spent, the number that matters more than reported revenue) grew only 8% and missed, daily users fell sequentially, and the stock dropped 14%. Management argues better retention eventually more than pays for the lost spending, but wouldn't say when that crossover arrives — and then withdrew full-year guidance entirely while guiding next quarter's bookings to decline 14–18%. Last quarter it was safety changes hurting new-user growth; this quarter the fix brought a different cost. A recap, not a call.

Reddit is squeezing far more money out of each user than almost any consumer platform — revenue up 61%, average revenue per user up 36% to a record $6.18, and profit margins reaching 43%. The worry is the denominator. US daily users actually fell slightly, because a large share of Reddit's audience arrives via Google searches, and Google's AI summaries and chatbots increasingly answer the question on the results page using Reddit's own content — so fewer people click through and become regular users. Reddit's response is that app users are worth several times more than search visitors, so it's investing in feed quality, onboarding and retention instead of chasing search traffic. Next quarter's revenue guidance is well above expectations, so monetisation isn't the issue. Whether the audience keeps growing as AI reshapes search is. A recap, not a call.

Rivian's whole future rests on the R2, a cheaper SUV meant to take it from niche to volume, and this quarter it started delivering them to outside customers — with more reservation-holders converting to actual orders than management expected. Demand, in other words, is not the problem. The economics also improved sharply: the company made an 11% gross profit versus a large loss a year ago, and the car-making segment's loss shrank from $335 million to $36 million, helped by a fast-growing software and services business. Deliveries beat its own forecast and full-year guidance went up. What's left is execution and cash: production is limited by its slowest suppliers, it burned $849 million in the quarter, and it raised another $1.3 billion by selling shares, with $1.25 billion more expected from Volkswagen and Uber. Scaling without letting launch costs swamp the improving unit economics is the test. A recap, not a call.

SoFi set records on nearly every measure — loans originated (up 69% to $14.8 billion), members, products, book value — and the stock fell 10%, the same as last quarter, leaving it down over 40% this year. The most encouraging number isn't the headline: 51% of new products were opened by existing members, up from 35% a year ago, and SoFi added twice as many products as members for the first time. That's the "one-stop financial app" thesis actually working. Two things weigh on it. The technology-platform division, which SoFi has pitched as its future, fell 23% as it laps the loss of the Chime contract. And a short-seller report from Muddy Waters alleging hidden credit problems still hangs over the shares — though default rates actually fell this quarter across both personal and student loans, which cuts against that claim. Guidance raised revenue but left profit flat, which reads as reinvestment or margin pressure. A recap, not a call.

Align makes Invisalign clear braces and the iTero 3D scanners dentists use to fit them. Its scanner revenue fell 11% — deliberately. Align is pushing cheaper scanners and leasing deals to get more machines into more practices, on the logic that every installed scanner generates years of aligner cases afterwards. It's trading a one-off hardware sale for recurring treatment volume, and it will depress reported revenue while the transition runs. Meanwhile the activist investor Elliott has engaged: Align is adding three independent directors, launched a strategic and operating-model review, and raised its buyback to $400–500 million. Aligner volumes hit a record and grew 7%, carried by Asia, Europe and Latin America while North America was flat. The number that matters next quarter is aligner volume — that's the engine paying for the scanner transition. A recap, not a call.

BetterHelp, Teladoc's online-therapy business, historically charged patients directly out of pocket. It is switching to billing insurance instead — because 70–80% of prospective patients would rather use insurance — which should build a much stickier business but costs revenue immediately. That's a defensible strategy; the execution is what broke the stock, which fell about 29%. Demand for insurance-covered therapy outran the number of therapists Teladoc has, so it cut advertising to stop generating demand it can't serve — and that same ad cut is what's accelerating the decline of the old cash-pay business. Teladoc cut its revenue forecast by about 5% while keeping its profit forecast intact on lower spending, which looks like discipline but is the very thing starving the funnel. The arithmetic problem: a $110 million insurance run rate is replacing a much larger cash-pay base that's falling faster. A recap, not a call.