AInvest - Latest Articles

Why Is PLNT Stock Falling Today? Planet Fitness Drops 19.8% Premarket on Q2 Earnings Concerns

Ainvest Pre-Market Radar — Mon, 26 Oct 2026 10:39:33 GMT

Planet Fitness (PLNT) shares fell 19.84% in pre-market trading on Tuesday as investors digested concerns from the company's Q2 2026 earnings report, including significant margin compression and weak same-club traffic growth. Despite beating on both EPS and revenue, the stock tumbled on the underlying operational headwinds.

What Did Planet Fitness Report?

For the quarter ended June 30, Planet Fitness reported total revenue of $365.2 million, up 7.1% year over year and above the $356.6 million analysts had modeled. Adjusted EPS came in at $0.88 versus the $0.85 consensus. Adjusted EBITDA grew 3.5% to $152.8 million.

The headline beats masked deeper issues. Adjusted EBITDA margin compressed 150 basis points year over year to 41.8%, down from 43.3%. System-wide same-club sales growth slowed dramatically to 1.7%, compared with 8.2% a year earlier. Management noted the growth was driven entirely by pricing rather than increased member visits or traffic.

Adjusted net income actually declined 5.7% to $68.4 million. The EPS beat was supported by a sharply lower share count — weighted-average diluted shares fell to 77.5 million from 84.4 million, a reduction of about 8.2% — driven by $200 million in share repurchases during the quarter alone.

Part of the top-line increase came from a mechanical change: the National Advertising Fund contribution rate rose from 2% to 3% in 2026, adding $10.1 million to both revenue and expense. The pass-through increased reported revenue without providing a corresponding profit contribution.

Why Did Investors React?

The market's reaction centered on two core concerns: profitability erosion and the lack of organic member traffic growth.

On the margin front, the 150-basis-point compression in adjusted EBITDA margin signaled that revenue growth was not translating into proportional profit. Free cash flow margin halved year over year, falling from 2.4% to 1.1%. Quarterly interest expense rose to $33.4 million from $26.2 million, and full-year net interest expense guidance was raised by $4 million to approximately $115 million.

On the traffic front, same-club sales growth of 1.7% — driven solely by pricing with no contribution from member visits — pointed to stagnant engagement. The company ended the quarter with 21.5 million members, up 3.6% year over year but flat sequentially versus Q1. CEO Colleen Keating acknowledged the challenge, telling investors the company is working to "reignite sustainable member growth."

Management is testing a new marketing campaign, a redesigned member app, a $10 Classic Card promotion, and AI-driven churn prediction models. The company also introduced an AI churn prediction model in alpha phase and is testing regional pricing strategies. None of these initiatives had quantified results to share.

Planet Fitness reaffirmed most of its full-year 2026 guidance, including approximately 7% revenue growth, 6% adjusted EBITDA growth, and 6% adjusted EPS growth. It projected adjusted net income to decline approximately 3% and plans to open 180–190 new clubs system-wide.

What Comes Next?

Planet Fitness is scheduled to report Q3 2026 results before market open on November 5, 2026. Analysts expect EPS of $0.81 on revenue of approximately $348.4 million. The Q3 print will be closely watched for signs that the company's marketing evolution, pricing tests, and retention initiatives are gaining traction.

The company's stock has faced sustained pressure. Shares trade near $42 in pre-market trading, down sharply from a 52-week high of $114.26. Despite the decline, the analyst consensus remains "Moderate Buy," with an average 12-month price target of approximately $69. A proposed securities class action alleges the company misled investors about its marketing effectiveness and growth outlook.

With full-year guidance intact but member growth in question, the regular trading session will show whether today's reaction reflects a fundamental reassessment of Planet Fitness's growth trajectory or an overreaction that moderates once broader market participants weigh in. With Q3 earnings arriving in November, investors have a short runway to see whether the initiatives management outlined are showing early results.

The stock's magnitude of decline — nearly 20% — is significantly larger than its approximately 11.6% drop on the original Q2 earnings print in August, suggesting investors are now pricing in deeper concerns about the sustainability of Planet Fitness's growth model. Pre-market liquidity can be thinner than during regular trading, so the final level will depend on how the stock is received once full market participation begins.",

Why the Courts Keep Tearing Down Trump's Tariff Wall - and the Foundation That Finally Sticks

Adrian Sava — Mon, 26 Oct 2026 10:34:58 GMT

Jamieson Greer's job isn't strictly trade policy. It's building a tariff wall the courts won't knock down, and he starts that job carrying a number that explains the whole exercise: $165 billion. That's the refund the government now owes companies, by Bloomberg's tally, after the Supreme Court demolished the tariffs the president had imposed using emergency powers. Greer talks about "durable tools" and walls "built to survive those legal assaults," but the honest way to read the past eight months is as a forced migration - from fast, sweeping tariff authority down to slow, narrow, congressionally-blessed statute. The wall that survives is not the wall that was promised. It's smaller, and it's slower, and that's the part investors should actually price.

USTR Greer talking with Senator Bennett at a hearing on USMCA and trade levies (source: Agri-Pulse)

Why so many walls fall

Start with the constitutional fact that keeps sinking them. A tariff is a tax, and the Constitution hands the taxing power to Congress, not the president. The executive can only tax imports when Congress has explicitly delegated that authority, with limits attached. That single sentence is the whole court fight in miniature. This February, the Supreme Court ruled 6-3 in Learning Resources, Inc. v. Trump that the International Emergency Economic Powers Act - an emergency statute that never mentions tariffs - didn't give the president that power. By ABC News' accounting, IEEPA had carried roughly 70 percent of every tariff the administration had imposed worldwide. When the foundation collapsed, most of the wall came down with it.

The United States Trade Representative's answer was a ladder of replacements, each one stepping down in ambition. Under Section 122 of the 1974 Trade Act - the authority born of a 1970s balance-of-payments crisis - the White House put a stopgap global tariff in place. But Section 122 caps such duties at 15 percent and allows them for only 150 days without Congress coming back to authorize an extension. The trade court has already ruled that the way the administration used it was unlawful - the statute is meant for payments problems, not ordinary trade deficits - and the Justice Department is fighting on appeal while customs keeps collecting. That tariff is set to expire, and the administration is scrambling to replace it before there's a gap in collections.

The durable one is the boring one

This is where Greer's legal strategy and an investor's should converge on the same insight: the tariffs that survive are the ones that come with process. Section 301 of the same 1974 law is the workhorse. Before a penny of duty can be imposed, USTR must run a formal investigation, publish findings that a foreign practice is "unreasonable" or "unjustifiable," invite public comment and hold hearings, and lock its conclusions to a documented burden on U.S. commerce. Because Congress wrote that machinery into the statute, courts have treated it as legitimate - it's the same authority behind the $370 billion in China tariffs from the first Trump term that are still on the books today. The president can't just pull it off the shelf; a record has to exist first. That's precisely the durability Greer is betting on.

Two Section 301 tracks are already in motion. The first, finished this July, slaps duties of 10 to 12.5 percent on goods from 60 countries that haven't enacted or enforced bans on forced-labor products. Importers have already sued over it at the Court of International Trade, arguing USTR dodged statutory guardrails written to protect developing economies, so even this foundation is being tested. The second and bigger track is an investigation into manufacturing overcapacity across 16 economies - China and the EU first among them - covering everything from autos and batteries to chemicals, semiconductors, steel and solar panels. Greer has said that probe will probably conclude before the end of the year. If it does, it becomes the template for the durable tariffs that follow.

Why this matters to money: a durable tariff and a temporary one behave differently in a company's accounts. A tariff that's built on Section 301, with national-security duties under Section 232 still standing on steel, aluminum, copper and semiconductors, is a cost a firm can plan around - it shows up in margins, pushes supply chains to relocate, and can survive a change of administration. The stretch authorities are the opposite. IEEPA, Section 122 and brand-new Section 338 retaliation against Canada all carry the same defect the courts keep punishing: they're narrow tools being used as a general tariff power. They get imposed, then reversed, then refunded.

The refund is the tell

That refund is where the whiplash becomes visible in cash flow. The Supreme Court's ruling didn't order refunds automatically; it left the mechanics to the trade court, and the process has been anything but clean. Customs has said its systems can't yet process the volume of affected entries without major manual work, refunds will only flow electronically for importers who set up accounts, and companies are being told not to assume the money arrives anytime soon. For the firms that booked IEEPA duties as a cost, the reversal is a real but messy cash-flow event that lands slowly, in fits, across quarters rather than as one clean line.

Reading the rebuild as an investor

Step back and the pattern is structural, not political. The courts have effectively written the executive a speed limit: the faster and broader a tariff authority is, the more likely it's straining a statute written for something else, and the more likely it dies in litigation. Every time it dies, the same revenue gets re-argued, re-imposed on a new legal basis, and refunded to whoever paid it - a cycle that turns the "tariff wall" into something closer to a recurring fine that companies absorb and occasionally get back.

The inference to carry forward is a narrow one. Treat the broad, emergency-justified tariffs as reversible noise, not as a durable cost - they've lost in court repeatedly and the refund pipeline guarantees reversals come back. Put the analytical weight on the Section 301 and Section 232 duties actually designed to stick: those are the targeted, investigation-backed tariffs on specific countries and sectors, and those are what will shape margins and supply chains for years. A tariff wall is only as high as the law under it survives, and the law that survives is the one Congress wrote fences into.

TokenOS in the "heart of Europe": the gap between Xunce's story and its accounts

Wesley Park — Mon, 26 Oct 2026 10:34:56 GMT

When a small Chinese data company announces that it is "expanding deep into" Germany, the sensible first question is whether any money has changed hands. By that test, Xunce (3317.HK) has mostly delivered a story. The Shenzhen firm says its new flagship product TokenOS is "accelerating its entry into the European market" after attending the Berlin Global Dialogue and the Invest in Germany Summit as a "strategic partner," with negotiations spanning automotive, high-end manufacturing and energy. Mercedes-Benz, Airbus, Bayer and Siemens Healthineers gathered at the same conference; nothing in the release suggests any of them signed anything. The company's entire overseas book in the first half of the year was ¥11.86m, up 264% on a year earlier - still barely 1% of the ¥967m the group booked in the half.

On-site of BGD 2026, via michigan.newsnetmedia.com

The gap matters because of what Xunce actually does. It is a maker of real-time data infrastructure, a business built serving Chinese fund managers, who demand millisecond responses and near-total accuracy. Over five years it pushed the same stack into eleven industries - telecoms, power, energy, advanced manufacturing, biomedicine - accumulating 230 clients and more than 300 standard modules. For most of its life it sold projects and software, a linear model whose ceiling it reached in 2024, when revenue grew just 19%. Then, in May 2026, it relaunched itself around a new idea: data as a metered token.

TokenONE, the product in that relaunch, turns an enterprise's private data into AI-ready "scenario tokens" - compressed industry knowledge a generic large model cannot derive from public text - and charges by usage, at $10–$100 per million tokens, ten to a hundred times the price of an ordinary model token. The revenue formula became customers times modules times price times calls; only calls are unbounded. "Xunce has installed an electricity meter for the client's AI usage," as one Chinese industry account put it. The upgrade is genuine: a project seller that bills when a job is delivered becomes a rentier that bills as long as a customer keeps using AI.

And the meter has begun to spin. In the first half of 2026 Xunce booked revenue of ¥967m, up 389% on a year earlier, and its first-ever half-year profit - attributable net income of ¥72.5m against an ¥89m loss a year before. Token-paid revenue has climbed from about 5% of the total in April, and the company is aiming for 20–30% by year-end. Customer retention sits above 90%.

Germany is where the narrative runs ahead of the accounts. European data sovereignty is both the reason the region wants "token factories" that keep industrial data under EU control and the toll-gate a Chinese vendor must clear. With overseas sales at roughly 1% of revenue, the Berlin trip is a credibility play - a bid to attach the token narrative to marquee European names - rather than diversification that has yet begun to cushion a business earning the overwhelming majority of its revenue at home. To give the bull case its due, that sovereign-data demand is real and growing as AI moves into factories; a verified win there would genuinely matter for a firm this China-concentrated.

The market has already priced in much of the token story. Xunce listed in Hong Kong in December 2025 at HK$48–55 a share, raising up to HK$1.24bn, and the stock had multiplied roughly sevenfold by its April 2026 peak of about HK$360 before autumn's correction - down some 29% over a month and 35% over three at the level of its over-the-counter American depositary receipt. Even corrected, figures compiled by Morningstar and FactSet put the company at a market value of around $2.8bn, roughly eight times trailing enterprise value-to-revenue, on earnings so thin that a price/earnings ratio in the hundreds is more a warning than a metric.

The variable that will decide whether this internationalisation is real is the same one the press release omits. Until "the heart of Europe" shows up as a signed contract, a quantified EU revenue figure or a disclosed token-factory customer, the German trip is a legitimacy campaign for a small, volatile, China-concentrated stock. The measure to watch is not the next conference invitation but whether the meter starts spinning abroad.

Pick n Pay Is Two Businesses. Only One Works

Adrian Sava — Mon, 26 Oct 2026 10:34:54 GMT

Pick n Pay's stock popped as much as 5.7% to 22.32 rand on Monday morning - the highest intraday price in five months - after the South African grocer reported a narrower loss and said its core supermarket brand was growing volumes again. The market read it as the turnaround finally gaining traction. Bloomberg's headline on the numbers said exactly that.

Then look at where the profit actually came from.

In the 26 weeks through August 30, the group's trading profit rose 58% to 490 million rand. A healthy headline. But the namesake Pick n Pay brand alone still lost 516 million rand on a trading basis in the same half, narrowed only 17% from 621 million rand. The group swing is a simple subtraction problem: roughly a billion rand of trading profit at its discount chain Boxer, which absorbs the flagship's half-billion-rand loss and still leaves the group in the black. Strip Boxer out, and the company that headlines as "recovering" is still deep in the red.

This is the one fact that makes the whole story legible. Pick n Pay Stores Ltd. is not one company. It is two businesses stacked together, consolidated on one income statement, and the word "group" hides which one is doing the work.

Interior of a Pick n Pay supermarket in Cape Town (REUTERS/Mike Hutchings)

The profitable half is already priced

Boxer is the value chain - discount grocery aimed at the everyday consumer. It is the serious business. Pick n Pay listed 34.4% of Boxer in a November 2024 IPO and sold an additional 11.5% stake in May 2026, leaving it with roughly 53%. Boxer is so profitable and fast-growing that it trades as its own stock on the Johannesburg exchange. ainvest values it at roughly 38 billion rand on its own, with operating margin widened from under 4% to about 5.6% and revenue up roughly 80% over four years.

The point of the box is not the numbers themselves. It's that Boxer's value is already sitting in the market, separately observable, attached to its own ticker. The market does not need a Pick n Pay turnaround to value Boxer.

The half you're actually betting on

So what is left when you buy Pick n Pay Stores stock? Majority ownership of that already-valued Boxer, plus the loss-making namesake brand - a turnaround bet with no profit yet to capitalize. That is why the company carries no useful earnings multiple. An unprofitable company has no P/E. Investors are forced to price the parts.

And the part that decides whether this stock works is the flagship, which management wants back to break-even by fiscal 2029. That is the entire case, stated as the standard turnaround playbook: close loss-making stores, cut costs, and rebuild volume so that every extra sale of groceries drops to the bottom line.

The mechanism matters here, so follow it. Grocery labor is one of a retailer's largest costs, and most of it is fixed - you staff the store whether or not it's busy. That's why the quietest detail in this release is structural: Pick n Pay withdrew a restructuring consultation it had begun in May after reaching a deal with its main union on a new in-store staffing model. In plain terms, it can now match opening hours and staffing to actual customer traffic. That converts fixed store costs into flexible ones. It is the operating leverage the whole turnaround depends on - the thing that turns a modest like-for-like sales increase into a much bigger swing in profit. It works in both directions.

The volume side is real but early. Turnover across the Pick n Pay segment was flat at 36.4 billion rand, while like-for-like sales - which strip out store openings and closures - grew 2.4% at company-owned South African supermarkets, and the company says the four weeks after the period ended ran slightly ahead of the first half. Online is the genuine standout, up 35.4%, though the company gave no rand figure.

The fight it is walking back into

Here is where the data argues against the timeline. Shoprite, the market leader, has been aggressive on price and expansion. Walmart began rolling out its own-name stores in South Africa last year. And consumers are under renewed pressure - Pick n Pay cited higher fuel prices from May 2026 filtering into supply-chain costs and household budgets. In a price war over squeezed shoppers, the like-for-like growth that operating leverage needs can stall, and the leverage then grinds the other way: if the flagship has to invest in price rather than volume, that 516 million rand loss can widen again before it narrows to zero.

Add a leadership handover on top. Sean Summers, the CEO who returned in 2023 to run this play, hands the job to Woolworths executive Spencer Sonn on February 1, 2027 - just as the plan aims at the fiscal 2029 break-even. Turnarounds are execution businesses, and a leadership change mid-flight is a real execution variable, not a footnote.

The company funds this with roughly 10 billion rand of net cash, so it is not running out of runway. But that is a time cushion, not a proof of concept.

Verdict

Narrower losses confirm direction, not destination. The market priced the recovery as a triumph, yet the flagship is still 516 million rand in the red and the consolidated "profit" is Boxer's - a business you can already own separately and priced on its own merits. A Pick n Pay Stores share is effectively a leveraged call on one unresolved question: whether the namesake brand can reach zero by fiscal 2029 while Shoprite and Walmart keep escalating the fight. The data supports progress toward that question, not the answer itself. Anything that keeps like-for-like volume compounding, holds the cost flex, and spans the CEO change supports the case. A renewed price war with volumes flat would put the 2029 target - and the stock's re-rating - back in doubt.

SK Hynix's ~100 Trillion Won Payday Is a Harvest at the Top, Not a Dividend

Henry Rivers — Mon, 26 Oct 2026 10:34:54 GMT

Every income investor just got handed a headline that sounds like a bell ringing. SK hynix, the South Korean maker of the AI memory chips no data center can skip, is reportedly readying another 50 trillion to 60 trillion won in shareholder returns, on top of the 40 trillion won buyback it already finished - roughly 100 trillion won across 2026 when the reports are added together. The first instinct is to smile and think "they've finally become a dividend company."

Hold that thought and sort the money first. A dividend is a stream you can count on next year and the year after. A windfall is a one-time harvest. Almost everything SK hynix is handing back right now is the second kind - funded by free cash flow at the top of a memory boom and by a single lucky disposal gain - while the actual recurring dividend remains so small it barely clears the decimal point.

The regular dividend is a rounding error

Start with the part that is a genuine, repeatable income promise, because it sets the scale for everything else. SK hynix's recurring dividend is small - a fraction of a percent of yield on the U.S.-listed shares (ticker SKHY), against a market value approaching a trillion dollars. On the TIKR financial-data series, 2025 dividends totaled about $1.16 billion, and the full five years from 2021 to 2025 together came to roughly $4.37 billion.

Burn those numbers in, because they are the reference point for the "record returns" story. When you hear "60 trillion won" or "100 trillion won," you are not hearing about the dividend. That is a different animal.

Where the big number actually comes from

The bulk of the cash is two one-time or cyclical items, not a raised payout.

The first is a 40 trillion won buyback - about $28.6 billion - approved on August 19. Per the company's own disclosure, it covers roughly 24.07 million shares, about 3.3% of the 730.5 million issued, and every purchased share is cancelled. It is, by the company's telling, the largest share-cancellation program among listed South Korean companies, and it was completed on October 21, about a month ahead of schedule. Put that against the engine: over the trailing twelve months SK hynix generated roughly $59 billion of free cash flow, so the buyback alone hands back about half of a year's output. Add the reported special dividend below, and the combined total approaches 70% of trailing free cash flow.

The second is a special dividend of about 16 trillion won ($12 billion) that the Dong-A Ilbo newspaper reported and Bloomberg picked up - roughly 80% of the estimated ~20 trillion won of pre-tax gains from selling down SK hynix's stake in Japan's flash-memory maker Kioxia. The original investment traces to 2018, when a Bain-led consortium that included SK hynix bought Toshiba's former memory business for about $18 billion; Kioxia listed in Tokyo in December 2024, and as the gains crystallized this year SK hynix booked a large one-off investment-asset gain. That is roughly ten times last year's entire dividend.

Important caveat, because this is the kind of detail that separates a disciplined read from a hype read: as of this writing the special dividend is a press report, not a confirmed commitment. On October 26 SK hynix said it was reviewing various shareholder return measures and would announce details after board approval alongside its third-quarter results on October 28. The 40 trillion won buyback is done; the ~100 trillion won tally and the 16 trillion won special dividend are both subjects of the upcoming announcement, not settled facts.

The SK hynix headquarters sign in Icheon-si, Gyeonggi Province. [Yonhap News Agency]

Why the money is flowing now matters as much as the amount

A company returning this much cash this fast does not do it out of pure generosity, and the reasons tell you when you are. Two forces are at work.

One is real and good: this is a genuinely excellent business at a cyclical peak. SK hynix holds more than half of the High Bandwidth Memory market through at least 2026, and HBM is the near-bottleneck you cannot train a frontier AI model without - a real toll road in the AI buildout, with the pricing power that comes with it. The result is boom economics that Ainvest data shows as an operating margin around 68% and a return on invested capital around 60%.

The other force is the reason the pace has become aggressive, and it is governance, not generosity. South Korea has been pushing its "value-up" corporate-governance reform, and activist money is now circling the SK structure specifically. Align Partners disclosed a 1.5% stake in SK hynix's parent, SK Inc, arguing the conglomerate trades at a roughly 75% discount to intrinsic net asset value - a "double discount" in which the value of SK hynix is marked down once at the intermediate holding company and once at the parent. When a company says, as SK hynix did, that its share price "does not fully reflect its intrinsic value," and then immediately launches the largest buyback in Korean history, you are watching management respond to a valuation-discount problem, not leisurely choosing to share surplus.

The trap is in the timing, not the business

Here is the analytical crux, and it runs the other way from the comforting headline. A pledge to return "over 50% of cumulative free cash flow" across 2025 to 2027 - the target was raised from "up to 50%" - sounds generous. But it is pro-cyclical by construction. Memory is the most boom-bust industry there is, and free cash flow is what collapses in the bust. So this policy hands back the most money in the boom years and naturally returns far less in the lean years, because by then there is little free cash flow left to return. Commitment to an FCF percentage is not commitment to a dividend in won.

That is precisely why the valuation can look seductive and be a warning at the same time. Ainvest data puts SK hynix at a trailing P/E of about 9.5 - cheap on paper, until you remember that the "E" is a peak-cycle earnings figure produced at a 68% operating margin. Memory producers historically swing from those boom margins into losses; a low P/E at the top of a memory cycle is the classic cyclical trap, not evidence of value.

Read the whole thing honestly and it is a lesson in classifying cash. The committed promise - the recurring dividend - is negligible. The buyback and the reported special dividend are cyclical and one-time harvests funded by peak free cash flow and a Kioxia windfall. This is not yet an income stock, no matter how many trillions of won the headlines carry.

None of that makes SK hynix a bad business. It is a great one with real pricing power. But for the income investor the correct framing is: you would be buying a high-quality real-economy cyclical near the top of its cycle, and the "shareholder returns" are the market price of that peak - not a durable income stream and not a valuation safety net. The variable that would change the judgment is memory pricing, because if DRAM and NAND roll over, both the boom earnings and, by design, the return program shrink together.

F5 (FFIV) has doubled on an AI hardware boom - Tuesday's print tests whether it lasts

Julian Cruz — Mon, 26 Oct 2026 10:34:46 GMT

F5, Inc. (FFIV) is one of the more intriguing names in tech right now because of what a single earnings season has done to it. The application-delivery and security company - the maker of the BIG-IP appliances that route and protect application traffic across data centers and clouds - reports the quarter that ended September 30 on Tuesday afternoon (F5 calls it fiscal Q4; it is the calendar third quarter). Shares sit near $455, up roughly 80% over the past year and more than double their 52-week low of $223.76, against a ~$25.8 billion market cap.

The setup is the kind I like to look at carefully, because the numbers and the price are pulling in different directions. F5 has been beating estimates consistently - it has topped consensus earnings in 7 of its last 8 quarters and surprised by an average of about 16% over the trailing four. Last quarter was a standout: $4.73 in non-GAAP EPS against a $4.00 consensus, with revenue of $865.1 million up 10.9% year over year. That is a genuine beat machine, and the market has rewarded it.

But here is the catch that gets my attention. For the quarter F5 reports Tuesday, the bar was already set by the company itself: it guided non-GAAP EPS of $4.14 to $4.26 (midpoint $4.20) on revenue of $870 to $890 million. Wall Street is carrying roughly $4.26 in EPS on about $884 million in revenue. Notice what that means next to the year-ago quarter, when F5 earned $4.39 per share on $810 million. We are looking at about 9% revenue growth, yet earnings guided about 3% below the year-ago figure. Top line accelerating, bottom line roughly flat or slightly down - that is the tension worth understanding.

Why revenue grows but EPS doesn't stretch

The reason sits in what is actually driving F5 lately. The recent acceleration has been led by the Systems business - the physical BIG-IP appliances - as customers upgrade their data-center capacity to handle AI and more data-intensive workloads. Zacks's preview of the quarter flags exactly this: demand for F5's Systems products is expected to have been a key contributor, with customers buying higher-performance infrastructure to prepare for AI workloads, while also noting that "the sustainability of hardware demand beyond customer refresh cycles remains an important factor to watch."

That is a refresh cycle - a one-time-ish surge in appliance upgrades, not yet a compounded recurring engine. Hardware carries lower margins than subscription software, which is a big part of why the EPS is guided flat even as revenue climbs. F5's longer-term story, by contrast, is the shift to recurring software and security subscriptions - BIG-IP software, NGINX, and its Distributed Cloud security - which grow more modestly but are the durable, higher-margin part of the base. The near-term AI boom is showing up on the hardware side; the compounding value is on the software side, and the software line is not yet reaccelerating.

This matters because of what a beat streak can hide. F5 keeps clearing the bar, but the quarter being reported is already guided to roughly flat EPS. That means the real news Tuesday is not whether F5 beats - the whisper number of $4.40 sits above both the guidance midpoint and the consensus - but what the beat is made of and what management guides for fiscal Q1 next. A beat delivered against a bar the company itself set, and driven heavily by a lower-margin hardware upgrade cycle, is not the same signal as a beat built on accelerating recurring software.

The street is not agreeing with the stock's move

The valuation is where the debate gets real. At about $455, the stock trades just over 26 times F5's own guided fiscal 2026 non-GAAP EPS of $17.21 to $17.33. That is a premium multiple for a company the street models to grow revenue around 8% to 9% a year going forward, with margins grinding higher rather than stepping up. It is a high-quality, net-cash, profitable business - gross margins above 80% and net margins around 22% - so the premium is not absurd, but it is fully paid for the good news.

The state of analyst opinion underscores that the stock has moved ahead of the average view. The consensus rating is a "Hold" with an average price target around $426 - a level below where shares trade, implying a mid-single-digit downside to the consensus target. Behind that average is a wide spread: Bank of America carries a Sell at $300 (about 34% below the price), Barclays and Morgan Stanley sit on Holds near $397 to $415, while JPMorgan and RBC Capital run Buys at $515 and $525. Targets ranging from $300 to $525 after an 80% run is a street that doesn't know what to do with the name. The bullish camp is betting the AI application-traffic story compounds - and F5's recent Telenor AI Factory win with Nvidia's BlueField-3, announced a week before the report, gives that camp something to point at. The bearish camp is betting the hardware refresh rolls over and leaves a 26-times multiple with nothing growing into it.

Final Thoughts and Recommendation

F5 is a good company reporting a quarter it has already pre-announced the shape of. It will probably beat again, and the AI infrastructure tailwind is real - when Telenor's AI factory picks your boxes to carry its AI traffic on Nvidia silicon, that is a vote for demand. But the stock has doubled while the quarterly EPS bar got flatter, the average price target sits below the market, and the entire near-term boom rests on a hardware refresh cycle whose durability is the one open question.

As of today I'm rating F5 a Hold, and I would not be chasing it into the report at 26 times guided earnings. The condition that would force a reconsideration is specific and lands Tuesday: I need the fiscal Q1 guidance to show the top line holding its growth and the bottom line resuming expansion off this hardware cycle, or the software subscription line to reaccelerate enough to make the multiple defensible on its own. If instead F5 guides to a step-down in EPS growth and leans on the refresh for one more beat, the premium to that $426 average target has a long way to come back. Watch the guide, not the beat.

A Buyback's Receipt Says More Than Its Headline

Dominic Reid — Mon, 26 Oct 2026 10:34:44 GMT

ISS, the Danish facility-services giant that cleans and runs the offices of much of corporate Europe, has spent most of 2026 strolling through its own stock market with a shopping cart, buying a few tens of thousands of its own shares a day. Its latest weekly notice - an EU filing headed "Transactions in connection with share buyback programme" - is unusually honest about why. The program exists, the company says, to "reduce the share capital" and, in the same breath, to "meet obligations arising from ISS' share-based incentive programmes." That is the filing admitting, in its own words, that part of this buyback is a way to hand your money back to you, and part of it is a way to stock a cupboard with shares to pay employees later.

That second clause is the one worth reading. "Buyback" in a headline sounds like one thing - the company believes in itself, wants fewer shares out there, returns cash to shareholders. In practice it's a more general machine, and the classification of what a company is actually doing runs on a single hinge: what happens to the shares once it has bought them. If it cancels them, your slice of the company gets bigger and earnings per share mechanically tick up. That is a genuine return of capital to the owners who remain. If it holds them in a drawer labeled "treasury" and later hands some to executives and employees under incentive plans, then part of the buyback was really payroll in costume: the cash left the company, but the shares come back out to new holders, and your percentage gains less than the headline implied.

ISS doesn't tell you, in this notice, how the split works. What it does tell you is that both purposes are on the table, and that it currently owns 6,864,248 of its own shares - about 4.3% of its share capital, roughly one share in every 23 - sitting in treasury rather than retired. Whether those end up cancelled or reissued is the difference between a buyback that enriches you and a buyback that mostly pre-pays your colleagues' equity. The company left that door open in its own filing.

The other thing the transaction notice teaches is that a buyback announcement is a license, not a promise. ISS authorized up to DKK 3.1 billion over twelve months - a first tranche of about DKK 1.25 billion, plus a second tranche of up to DKK 1.85 billion that started in August. But the money only leaves the company one small lot at a time. In the week to October 16 it bought 25,000 shares on the 12th at 290.16 kroner, then 30,000, then 63,000 at 296.65, then 26,000, then 30,000 - roughly 174,000 shares over the week, adding about DKK 51.5 million. That's a drip, and it's deliberate: the EU "safe harbour" rules the program runs under cap how much a company can buy on any given day, so it can't just bid its own price up. The notice is how outsiders watch the drip happen.

Read that way, one number in the receipts stands out. Those October purchases went through at about 290 to 300 kroner a share, roughly 10–13% above the 264.88-kroner average the program has paid over its life. The stock rose, and ISS kept buying anyway, at the high end of its own cost curve. That isn't a red flag by itself - buyback programs plow ahead by design rather than time the market - but it is exactly the sort of thing the transaction receipts exist to reveal, and a press release announcing the program would never have told you.

Not every filing with this title covers the same machine. The same week, Bavarian Nordic, the Danish vaccine maker, published a "Transactions in Connection with Share Buy-Back Program" notice for a different shape of deal: DKK 250 million, run through a Nordea broker over about eight weeks, already roughly 86% spent, with 5.71% of its shares now in treasury. Same first line of the story, completely different structure underneath - one a slow, dual-purpose, multi-tranche program; the other a short one nearly exhausted. The title is where the similarity ends.

None of this is a scandal, and ISS's structure is normal; plenty of US companies run buybacks that double as compensation funding. It's just that "return your cash" is only one of the things the word "buyback" can mean, and one headline tends to blur them. So when you see a company announce one, the question isn't just "Is it buying back stock?" It's "What happens to the shares afterward?" Read the purpose line in the filing - "reduce share capital," or "secure shares for our incentive plans," or both. Check whether the shares are being cancelled and the share count is actually falling, or whether they're piling up in treasury, where they can be handed back out. And look at the transaction-notice arithmetic to see whether the authorization is actually being spent, and at what price.

A board can authorize a buyback it never runs. A company can run one that quietly feeds its own payroll. A management team can buy every trading day into a rising stock. All three wear the same headline. The receipts are your tell: how many shares actually came back, and what the company did with them.

Lufthansa quit Israel's market. The fuel bill is the bigger story.

12X Valeria — Mon, 26 Oct 2026 10:34:38 GMT

Lufthansa has pulled out of the entire Israeli market until the end of November and is refunding or rebooking passengers whose tickets it can no longer honor. The airline pushed the restart of its Middle East service back again, and warned customers that scammers are impersonating Lufthansa online as people hunt for refund information. It is not alone: after the U.S.-Israeli strike on Iran, most international carriers treat the region as too unstable to serve, leaving state-owned Gulf carriers Emirates and Etihad as the main way in and out of the country. Take the route as one line of flight news, not the whole report. The route is the symptom; the price of fuel is the disease. Walking away from a market removes revenue, yes, but the number that actually moves earnings is Lufthansa's own estimate of this year's extra fuel bill - roughly €1.7 billion, tied to the closure of the Strait of Hormuz, which is why the airline is raising ticket prices and cutting flights to pay for it.

The gap between one quarter's fuel and one quarter's profit

Put the scale on the table. In the second quarter of 2026, higher fuel prices added about €750 million of cost, and Lufthansa still booked an operating profit of €383 million on a record €11.1 billion in revenue, up 8% from a year earlier. A single quarter's fuel surcharge ran about twice the profit that quarter produced. It was absorbed only because demand, yields, and Lufthansa Cargo held up - cargo's adjusted profit rose 42 million euros - and because, for now, the fuel is hedged. Stretched across the year, the math gets harder. The roughly €1.7 billion incremental fuel bill is about the size of the €1.7 billion-to-€2.2 billion operating profit the group still guides for this year. That is why Lufthansa is doing several things at once: raising ticket prices, cutting flights, and leaning on Lufthansa Cargo. None of this is a solvency story. With 10.7 billion euros of liquidity against 8.3 billion euros of net debt including pensions at the end of June, the group can absorb a fuel shock and bow out of a market without scrambling for cash.

The expiry date hidden in the hedges

Here is the caveat with a calendar on it. The cushion is a hedge contract, and hedges expire. Lufthansa's finance chief has said about 86% of its expected 2026 fuel needs are locked in through derivatives, but only just over half of 2027's.

Coverage year	Share of expected fuel needs hedged
2026	86%
2027	50%

That step-down is the expiry date written into the thesis. This year's results are largely pre-priced, so the current fuel shock is buffered. Next year, a much larger slice of the fuel budget floats with whatever jet fuel costs, and that is when a sustained price shows up in margin without a hedge to hide it. The leading indicator is already legible. Booking cycles shortened and advance-ticket payments fell, which helped cut half-year operating cash flow by about 600 million euros to roughly 2.3 billion euros. Advance ticket sales are cash passengers prepay; when that pool shrinks, the airline is selling seats closer to departure - which is exactly what happens when people will not commit to travel that is both expensive and unreliable, the same unreliability that pushed Lufthansa out of Israel.

Two readings, one observable

The bearish reading: the conflict runs into 2027, fuel stays elevated past the hedge cushion, and Lufthansa keeps giving up revenue - cutting routes, refunding tickets, shortening its network. Analysts are already cautious. JPMorgan cut its price target to €7.00 and held a Neutral rating, trimming 2027 airline profit forecasts by about 25% on higher jet fuel, and the shares trade near €7. The market value is around €8.5 billion. The constructive reading: demand stays strong, fares hold, cargo runs hot, and the moment the Strait of Hormuz reopens, the fuel penalty reverses and Lufthansa flies back into Israel and the Gulf. The margin sits near 3.4%, far below the 8–10% medium-term target that is the basis for that upside. Both readings ride on the same observable - not the route headline. The two inputs that settle the debate are where the fuel price sits relative to what is hedged, and whether Lufthansa keeps booking revenue closer to departure.

The condition that retires the read

The route changes with the conflict; the equity moves with fuel and bookings. So this is not a permanent playbook. Before you put weight on Lufthansa's margin turn, re-run two screens: where 2027 hedge coverage and jet fuel actually sit, and whether advance bookings are still growing. When the Strait of Hormuz reopens and airspace normalizes, the calculator resets - the fuel surcharge unwinds and the exited market comes back, and the question becomes whether Lufthansa kept its fares and cost discipline long enough to bank the margin it is chasing. The headline tells you the conflict is on. The fuel number tells you what it costs.

Hyperscalers' $800B+ Capex Winners Sell Off, Watch 10 Earnings for the Next Big Rebound Through Wednesday

Mon, 26 Oct 2026 09:24:55 GMT

Major US hyperscalers are expected to spend more than $800 billion on capital investment this year, yet some of the biggest infrastructure beneficiaries are suddenly under pressure. GE Vernova plunged nearly 12% in one session, Vertiv suffered another sharp decline, and Celestica has lost almost 30% in a month despite strong AI demand. The contradiction creates a potentially attractive opportunity. Are investors abandoning long-term infrastructure winners too early, or are valuations finally catching up with the cost of the AI boom? With power shortages, faster networking and surging storage requirements still supporting investment, this week's biggest earnings surprises may come from stronger contracts, improving margins and an extended growth runway into 2027. 10 reports through Wednesday morning will test those opportunities, alongside potential recoveries in PayPal and Boeing. The real question is which companies can turn the selloff into a buying opportunity—and which still have further to fall.

Where We Would Look First

GE Vernova, Vertiv and Celestica offer three different post-selloff opportunities. GEV needs stronger order conversion and service profitability, VRT could benefit from delayed installations becoming revenue, while CLS faces a major optical-networking product transition after a sharp valuation reset.

PayPal is a separate turnaround candidate if branded checkout finally stabilizes. Boeing's recent MAX 10 certification improves its longer-term recovery prospects. Corning and Robinhood face demanding sentiment after four consecutive negative earnings reactions, while Seagate has risen after four straight reports. Our focus is on future earnings upgrades, not simply stocks with the largest declines.

Monday, October 26 — After the Close

Cadence Design Systems (CDNS)

One-day average return post earnings: +0.8% | Volatility: 5%

Strength: Quality software compounder with recurring EDA revenue, high switching costs and increasing exposure to custom AI-chip design.

What really matters: AI is expanding the market for semiconductor design as hyperscalers develop proprietary processors and chip complexity drives greater verification requirements. The less obvious opportunity is AI-assisted engineering, which could accelerate design cycles and expand customers' workloads. Stronger recurring EDA subscriptions, greater adoption of agentic tools and new 2027 design commitments would provide more durable growth than one-off IP licensing revenue.

Opportunity or risk: A relatively stable way to participate in AI infrastructure growth. Accelerating core EDA demand and longer customer commitments could support another earnings upgrade. An IP-driven beat without stronger software momentum would be less convincing. Historical reactions have been relatively muted, reinforcing the need for a meaningful forward surprise.

Celestica (CLS)

One-day average return: −2.3% | Volatility: 13%

Strength: High-growth AI networking supplier with advanced switching platforms, hyperscaler relationships and expanding co-packaged-optics opportunities.

What really matters: The nearly 29% valuation reset comes ahead of an important product transition. October's launch of the DS6005 co-packaged-optics switch, offering 102.4Tbps capacity with liquid cooling, could move Celestica further into higher-value networking systems. Its $3 billion equity offering also provides funding for expansion but raises the hurdle for future per-share returns.

Tuesday's investor day should clarify the 2027 customer pipeline, optical deployment schedules and whether new networking products can improve margins. The crucial question is whether additional manufacturing capacity will produce enough incremental profit to offset dilution and expansion costs.

Opportunity or risk: One of our most attractive recovery candidates if 2027 commitments strengthen. Firm hyperscaler orders and sustained margin expansion could justify a rerating. Slower CPO adoption or weaker returns on new capital would make the recent selloff more understandable. Strong revenue growth alone is no longer enough.

Tuesday, October 27 — Before the Open

Boeing (BA)

One-day average return: +1.1% | Volatility: 5%

Strength: High-risk industrial turnaround with dominant commercial-aircraft platforms, substantial backlog and leverage to production recovery.

What really matters: The FAA's October 20 certification of the 737 MAX 10 removes a major regulatory obstacle. With 1,555 aircraft ordered, the larger single-aisle model strengthens Boeing's competitive position and supports its longer-term delivery and aftermarket pipeline. The next step is converting that opportunity into reliable production and cash.

Existing 737 and 787 output, supplier stability and defense-program costs remain central. Certification improves future visibility but does not automatically resolve manufacturing constraints or working-capital pressure.

Opportunity or risk: Better production stability and stronger free-cash-flow guidance could accelerate the recovery. Another delivery increase accompanied by weak cash conversion would be less valuable. The biggest potential surprise is sustained positive cash generation rather than a temporary EPS beat.

Corning (GLW)

One-day average return: −7.3% (4 consecutive declines) | Volatility: 4%

Strength: Structural AI optical-infrastructure beneficiary with leading fiber technology, hyperscaler relationships and exposure to growing data-center connectivity demand.

What really matters: The next growth opportunity extends beyond conventional fiber connections. Larger AI clusters require higher-density optical systems, and Corning's photonics expansion could create another substantial revenue stream as networking architecture evolves.

Its Springboard Plan targets a $20 billion annualized sales run rate by year-end 2026. Longer term, management has outlined a $35 billion high-confidence sales target for 2030, alongside a more ambitious $40 billion internal plan. New hyperscaler commitments, optical pricing and segment profitability will determine whether that trajectory becomes more credible.

Opportunity or risk: Four consecutive negative earnings reactions highlight a demanding expectations backdrop. Stronger multiyear optical commitments and improving margins could interrupt that pattern. Another routine beat without better forward visibility may not. The surprise investors need is a longer and more profitable optical expansion cycle.

PayPal (PYPL)

One-day average return: −5.0% | Volatility: 12%

Strength: Value-oriented fintech turnaround with global merchant acceptance, established branded checkout, Venmo and substantial payment-processing scale.

What really matters: The recovery must come from regaining competitive strength, not merely lowering expenses. Recent analyst downgrades reflect worsening checks on branded checkout and Braintree, raising concerns about share losses and processing margins.

The key is whether merchant conversion, checkout engagement and Venmo monetization can improve transaction margin dollars. PayPal's global network remains valuable, but scale offers limited advantage if merchants demand greater pricing concessions or consumers increasingly choose competing payment methods.

Opportunity or risk: Our preferred non-AI turnaround to examine, but operating evidence remains essential. Stabilizing branded checkout and accelerating transaction profitability could challenge bearish expectations. Continued market-share pressure would reinforce the value-trap risk. Given its 12% historical reaction volatility, forward guidance could matter considerably more than a modest EPS beat.

SoFi Technologies (SOFI)

One-day average return: −6.3% (3 consecutive declines) | Volatility: 9%

Strength: High-growth digital banking platform with an expanding deposit franchise, strong cross-selling and long-term fee-based revenue potential.

What really matters: The next earnings surprise may come from a neglected business rather than lending. Technology Platform revenue declined 23% last quarter following the departure of a major client, while contribution profit fell 65%. That weakness challenges SoFi's diversification narrative even as its consumer franchise expands.

The new SoFi Tech Solutions platform creates an opportunity to restore enterprise growth through banking infrastructure, payments and risk management. New contracts, improving account activity and stronger segment margins would demonstrate that recovery. Healthy loan-sale demand and credit quality remain necessary to support the broader business.

Opportunity or risk: A technology-platform rebound combined with stable lending economics could produce a more meaningful earnings upgrade than another strong origination quarter. Continued enterprise weakness would leave SoFi increasingly dependent on consumer lending. Three consecutive negative reactions suggest rapid headline growth has not been enough.

Tuesday, October 27 — After the Close

Robinhood Markets (HOOD)

One-day average return: −9.1% (4 consecutive declines) | Volatility: 4%

Strength: High-growth retail investing platform with strong customer engagement, expanding prediction markets and opportunities for recurring subscription revenue.

What really matters: Prediction markets could become a structural business rather than another short-lived trading catalyst. Broader event-contract distribution creates opportunities beyond options and crypto, but its economic value depends on repeat participation, regulatory stability and customer retention.

The longer-term growth opportunity is increasing assets per customer, subscriptions and financial-product adoption. Net deposits and Gold membership provide better evidence of platform durability than revenue driven by unusually active trading. Nonrecurring investment gains also need to be separated from operating earnings.

Opportunity or risk: Four consecutive negative reactions suggest investors remain demanding despite strong growth. Higher deposits, recurring monetization and stronger customer retention could eventually change that pattern. The biggest upside surprise would be more durable earnings rather than another surge in transaction activity.

Seagate Technology (STX)

One-day average return: +12.9% (4 consecutive gains) | Volatility: 8%

Strength: Cyclical AI storage leader with strong nearline HDD demand, HAMR technology and improving profitability from higher-capacity drives.

What really matters: The next opportunity is extending the storage profit cycle beyond temporary supply tightness. HAMR technology can increase capacity per drive without equivalent manufacturing expansion, while AI data growth supports long-term cloud-storage requirements.

Customer purchase commitments, HAMR adoption and pricing discipline will reveal whether current profitability is sustainable. The danger is mistaking accelerated customer buying during a tight supply cycle for an enduring increase in annual demand.

Opportunity or risk: Four consecutive positive earnings reactions have raised expectations considerably. Firm multiyear commitments and stronger forward margins could support further upside. Slower purchasing schedules after another strong quarter would increase reversal risk. Excellent fundamentals do not guarantee an attractive entry price.

Wednesday, October 28 — Before the Open

GE Vernova (GEV)

One-day average return: +1.5% | Volatility: 9%

Strength: Quality industrial growth leader with gas-turbine technology, extensive service relationships and major exposure to electricity generation and grid investment.

What really matters: Last week's nearly 12% selloff may have overlooked how much of the opportunity extends beyond hyperscaler spending. Grid modernization, industrial electrification and long-term turbine service demand offer additional growth drivers, while constrained global turbine capacity supports pricing.

Gas Power equipment backlog and slot reservations reached 116 GW last quarter, with management targeting at least 125 GW by year-end. The crucial distinction is between reserved capacity and firm contractual commitments. Faster reservation conversion, better pricing and stronger service agreements could materially extend earnings visibility into 2027–2028.

Electrification margins and wind performance also matter. A strong turbine business cannot indefinitely offset underperforming operations elsewhere.

Opportunity or risk: Our leading AI-power recovery candidate. Stronger binding orders, higher margins and better cash-flow guidance could suggest the selloff was excessive. Financing-related delays or weaker conversion would justify investor concerns. Despite the correction, valuation still requires substantial future earnings growth.

Vertiv Holdings (VRT)

One-day average return: +0.8% (2 consecutive declines) | Volatility: 17%

Strength: High-growth AI infrastructure leader in data-center power management, thermal systems and liquid cooling.

What really matters: The most important potential surprise is execution finally catching up with demand. Last quarter's revenue included timing shifts from complex installations and temporary supply-chain congestion, making Q3 a test of whether delayed projects are now converting into sales.

Longer term, higher GPU density and liquid-cooling adoption could expand Vertiv's revenue per data-center deployment and create additional service opportunities. Investors need stronger backlog conversion, credible installation schedules and customer commitments extending into 2027.

Cash conversion is equally important. Rapid growth can absorb substantial working capital if inventory and unfinished installations rise faster than collections.

Opportunity or risk: One of the most compelling post-selloff growth setups if delivery execution improves. Strong bookings, recovered shipments and resilient margins could restore confidence. Another round of installation delays would weaken the thesis. With historical earnings volatility near 17%, the potential reward comes with substantial short-term risk.

Historical statistics use the latest four earnings-related, close-to-close daily returns. Average return is the arithmetic mean of signed returns; volatility is their sample standard deviation calculated using STDEV.S. Premarket reactions use the previous trading close to the reporting-day close; after-hours reactions use the reporting-day close to the next trading close. Figures are rounded. Consecutive gains or declines describe historical earnings reactions, not predictions.

WEEKLY REPRICING: Meta's Spending Shock Spreads AI Funding Fears as Oil Keeps the Fed on Edge. The Real Test Is Next

Mon, 26 Oct 2026 07:28:08 GMT

The AI trade faced a more demanding test last week as financing concerns spread from leveraged infrastructure companies to one of the world's most profitable technology giants. Meta's advertising business remained strong, but record spending pushed free cash flow below zero and sent its shares lower. Lam Research reported exceptional profits only to disappoint investors, while Tesla faced questions over the slower-than-expected commercialization of robotaxis and humanoid robots. Brent crude briefly approached $117 before retreating as the new week began, but continued volatility amid Middle East tensions complicated the Fed's path toward monetary relief. For the week, the S&P 500 lost 0.5%, the Nasdaq fell 0.7% and the Dow gained 0.4%, while the Russell 2000 rose 2.0%. Those modest index changes concealed a sharper repricing. Strong AI demand is no longer enough; investors increasingly want to know who can finance the buildout and earn an acceptable return.

Meta Brings AI Financing Anxiety Into Big Tech

Meta's third-quarter revenue climbed 26% to $64.6 billion, supported by higher advertising impressions and prices, while its fourth-quarter revenue outlook reached $73 billion to $76 billion. Yet shares fell 9% for the whole week. Capital expenditure surged to $34.91 billion, quarterly free cash flow turned negative by $2.12 billion, and management put its 2026 spending forecast at $135 billion to $145 billion. A $9.89 billion litigation charge also weighed on reported operating income; excluding it, operating profit would have been $22.88 billion rather than $12.99 billion. The advertising engine remains profitable. The concern is how quickly its expanding infrastructure commitment is consuming the cash it produces. One negative quarter of free cash flow does not establish a sustained funding shortfall, but it forces investors to reassess how much spending can be carried by advertising profits without reducing flexibility elsewhere.

Meta's decision to create non-voting Class C shares, effective January 2027, added a financing dimension. The new class could support future capital raising, acquisitions or employee compensation, although creating it does not immediately dilute shareholders. Reuters also reported that protection against default on Meta's bonds rose to a record 114.79 basis points. Some analysts attributed part of the move to banks hedging financing exposure rather than deteriorating fundamentals. Meta is not CoreWeave: it still generated $32.79 billion in operating cash flow during the quarter and held $82.45 billion in cash and marketable securities. But after the previous week's CoreWeave bond selloff, Meta demonstrated that even cash-rich hyperscalers face a higher burden of proof over capital allocation.

The spending has a credible commercial rationale. AI is improving Meta's advertising products, while Muse could eventually support transactions, subscriptions and other services. But agent adoption also creates substantial computing costs before monetization is established. Reuters Breakingviews highlighted research suggesting autonomous agents can consume roughly seven times the tokens of conventional human-led chats. That makes the critical question less whether Meta can build useful AI products than whether incremental revenue and margins can keep pace with the next wave of capital spending. Equity investors are beginning to impose the same discipline that creditors have demanded from more leveraged AI operators.

Oil Near $117 Makes Fed Relief Harder

Brent settled Friday at $116.72 a barrel, up about 6% for the week, while WTI finished at $103.27. Houthi attacks on Saudi oil infrastructure and continuing disruption around the Strait of Hormuz and Bab el-Mandeb kept supply risk elevated. Energy shares benefited even as technology stocks struggled. Higher fuel costs squeeze consumers, raise transport and manufacturing expenses and complicate the disinflation the Fed needs before policy can become less restrictive. The pressure also reaches AI infrastructure through construction, backup generation and operating costs, just as investors are demanding more predictable returns. This matters even if oil eventually retreats: contracts, construction schedules and financing decisions made during a period of elevated costs may affect project economics long after the immediate energy shock fades.

Sunday brought a possible diplomatic opening, with Qatari mediators visiting Tehran, but not a supply resolution. Saudi Arabia continued to report missile and drone interceptions. Markets can quickly remove part of oil's risk premium on encouraging headlines, yet physical availability depends on functioning shipping routes and sustained de-escalation. That leaves the Fed caught between weaker risk appetite and persistent inflation pressure. A pause in tightening would not necessarily mean cheaper capital is coming. Oil is raising the required return on investment precisely when the AI sector needs financing to become easier, not harder.

Record Chip Earnings Meet a Higher Valuation Bar

Lam Research reported record quarterly revenue of $8.22 billion, up from $6.72 billion in the preceding quarter, and a non-GAAP operating margin of 40.2%. December-quarter revenue guidance centered on $8.6 billion. Yet the shares fell sharply as investors weighed demanding expectations, supply constraints and uncertainty over China. The Philadelphia Semiconductor Index dropped 6.7% on Thursday, with Intel, AMD, Sandisk and Marvell each losing more than 8%. For the week, semiconductor shares declined about 3.2%, much more than the broader market. An outstanding supplier earnings report was no longer enough to validate every valuation in the sector.

The tension is structural. One company's infrastructure expenditure becomes a chip supplier's revenue immediately, but the buyer must recover that outlay through utilization and cash flow over several years. Reuters Breakingviews columnist Edward Chancellor highlighted this timing mismatch in the AI profit cycle. It helps explain why supplier earnings can be spectacular while the economics of their customers become more uncertain. At the same time, weak chip stocks do not prove demand is collapsing. In a Friday industry discussion, SemiAnalysis founder Dylan Patel pointed to bottlenecks in substrates, memory and optical components, while Bernstein analyst Stacy Rasgon argued that financing limits might prevent capacity from expanding too quickly and thereby support favorable supply-demand conditions for longer.

A shortage can also cut both ways. Limited supply may allow chip manufacturers to hold margins, but it can delay data-center deployment and push project costs above initial budgets. Financing constraints then become another bottleneck layered on top of the physical ones. That produces a more selective investment case, not a blanket bearish one. Scarce components can preserve pricing power for established suppliers, even as their customers struggle with expensive equipment and capital. Companies with strong balance sheets, binding purchase commitments and shorter payback periods are better positioned than operators dependent on repeated fundraising. Record industry sales and attractive shareholder returns remain two different tests.

Tesla Cut Spending as SpaceX Demonstrated Monetization

Tesla and SpaceX offered a revealing contrast to Meta. Tesla lowered its 2026 capital expenditure outlook from more than $25 billion to roughly $20 billion and reported $1.53 billion in third-quarter free cash flow. Quarterly capital spending was $4.7 billion, well below the approximately $7.5 billion expected by analysts. Yet shares declined as the commercialization of robotaxis and Optimus humanoid robots continued to progress more slowly than investors had anticipated. Stronger cash flow offered some reassurance, but uncertainty over when these businesses could generate meaningful revenue continued to weigh on Tesla's valuation. Unlike Meta, which faces questions about excessive spending, Tesla must prove that its ambitious AI and robotics investments can translate into commercial results.

SpaceX reported third-quarter revenue of $14.58 billion, up 161%, including $7.56 billion from AI-related activities, alongside $4.4 billion in net income. That provides more immediate evidence of revenue from computing infrastructure, although quarterly capital expenditure of $19.6 billion shows how expensive the expansion remains. Over the weekend, Elon Musk said SpaceX had closed a $40 billion debt arrangement for GPUs expected in the first half of 2027; Reuters noted that the terms and completion status were not immediately clear. Musk contrasted the company's established launch and satellite businesses with neocloud operators dependent almost entirely on future AI revenue. Reports of committees exploring a potential Tesla-SpaceX combination raised a further question about whether integrating AI, robotics, energy and satellite operations would unlock efficiencies or merely concentrate already substantial capital requirements. No transaction terms have been agreed.

Together, the results show why Wall Street is not simply demanding less spending. Meta was punished for absorbing too much cash, while Tesla's cuts raised execution concerns; SpaceX drew attention because substantial investment was accompanied by reported revenue growth. What matters is the credibility of the path between capital deployed and money earned, not the size of a spending announcement alone.

Falling Treasury Yields Did Not Ease Credit Stress

The 10-year Treasury yield slipped below 5% during Thursday's technology selloff, reaching around 4.95%, and ended Friday near 4.99% despite the oil rally. Investors sought government bonds as equities weakened. Yet corporate borrowing costs include credit spreads, which can widen faster than Treasury yields decline. CoreWeave's earlier bond selloff and Meta's higher credit protection costs illustrate why safe-haven buying does not necessarily translate into cheaper financing for AI projects. A lower risk-free rate cannot fix weak utilization assumptions or make an uncertain borrower safer.

The pressure was not confined to technology or the United States. French 10-year yields exceeded 5% amid fiscal concerns, while U.S. pending home sales fell 7.3% in September to their lowest level since records began in 2001. Mortgage rates near 7.4% were already restricting housing activity. The France-U.S. divergence is important: safe-haven flows can lower the yield on the most sought-after government bonds even as other borrowers face higher risk premiums. The fall in U.S. yields was therefore a sign of investors seeking safety, not evidence that credit markets had recovered.

These signs of strained demand would normally strengthen expectations for monetary relief, but expensive energy complicates that response. The Fed faces weaker interest-sensitive activity without convincing relief from inflation, leaving leveraged AI companies exposed even when Treasury yields retreat.

Beyond AI, Investors Reward More Visible Returns

Elsewhere, earnings showed a market increasingly focused on the quality of returns. GE Aerospace reported a 49% increase in quarterly orders to $19 billion and raised its 2026 earnings outlook, supported by commercial engine demand and recurring service revenue, although free cash flow weakened. General Motors announced a $20 billion multiyear share repurchase plan and improved its free cash flow outlook, but shares still declined amid concerns about rising costs and 2027 profitability. Netflix faced similar scrutiny after lowering its 2026 advertising revenue forecast from approximately $3 billion to $2.7 billion, reflecting a slower rollout of advertising capabilities. Together, these results showed that strong earnings and shareholder returns are no longer sufficient when cash flow weakens or future growth falls short of expectations.

Sector performance also challenged the idea of a broad recession trade. Defensive businesses attracted buyers, and the Russell 2000 gained over the full week. The shift was not an indiscriminate retreat from risk but a rotation away from the most demanding long-duration growth assumptions toward more visible cash generation and returns. That is the meaningful message in a week when the major indexes barely moved: beneath the surface, investors were changing what kind of growth they were prepared to pay for.

The next test comes with Wednesday's Fed decision and Microsoft and Alphabet earnings, followed by Amazon, GDP and PCE data on Thursday. Cloud growth and AI investment returns will show whether Meta's cash burn is company-specific or part of a wider hyperscaler pattern; inflation will determine how much support investors can expect from monetary policy. AI can keep growing rapidly while the market pays less for that growth. What matters now is whether the industry's largest spenders can prove that the returns justify the bill.

AI Is Making Workers Faster. So Where Are the Profits?

Mon, 26 Oct 2026 01:50:42 GMT

In an October 24 note, chief economist Torsten Slok argued that AI had delivered no financial payoff outside technology, pointing to weak revenue per worker and profit-margin trends beyond the Magnificent Seven. Those figures challenge the case for a broad profit boom. Faster work becomes higher profit only when a company converts the extra capacity into revenue or retained savings after the costs of using AI.

A 2025 published study of 5,172 customer-support agents found that AI assistance increased issues resolved per hour by 15% on average. A separate Danish study, revised in March 2026, found no detectable average effect on earnings or recorded hours in data through December 2024. Most chatbot users in that study reported putting the time saved into other tasks. More useful work can therefore coexist with unchanged paid hours; neither measure alone tells an investor how much profit AI adds.

Economists Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied the rollout at one business-software company. Less experienced workers benefited most, while the most skilled workers saw small speed gains and small declines in quality. The authors caution against generalizing across firms, occupations, and AI systems. This software-company example cannot establish a payoff across industries outside technology. It does show why measuring completed work and measuring a company's overall margin answer different questions.

Anders Humlum and Emilie Vestergaard linked adoption surveys to administrative records across 11 highly exposed occupations. Their estimates ruled out average earnings or recorded-hours effects larger than 2% over the period studied. Yet 85% of chatbot users reported reallocating saved time to other work, and new tasks included checking AI output and integrating tools. That is evidence of work changing before average pay or hours changed, not a measurement of U.S. corporate profits in late 2026.

Where the savings go

At JPMorganChase's February 23 company update, finance chief Jeremy Barnum said the firm had identified about $600 million in efficiencies, some AI-related, enabling it to invest more than otherwise. He also described about $1.2 billion of additional spending on major projects that year. These were broad technology and investment figures, not an AI-only profit calculation. They cannot be subtracted to produce a net AI return. They illustrate management's choice to use efficiencies to finance new activity.

That choice can keep expenses growing even when individual processes become cheaper. Barnum said the bank was not managing for short-term improvement in the relationship between revenue and expenses. He also described higher hardware costs as AI-related shortages pushed up memory prices. Lower unit costs, more investment, and higher input prices can operate at once. A flat margin can hide an efficiency gain absorbed elsewhere, just as a faster task can fail to cover the cost of deployment and review.

For Microsoft and Amazon, flat aggregate margins warrant scrutiny of new AI capacity, while their last reported results still support demand. Microsoft said in July that new Azure capacity was quickly monetized; AWS sales grew 37% year over year in the June quarter. Both clouds also serve non-AI workloads; these figures do not establish customer AI project returns. If those returns prove persistently weak, new projects become more vulnerable to budget cuts. Paid usage growing more slowly than installed capacity would then pressure cloud margins and cash generation, even while revenue still rises.

NVIDIA faces that risk through the next equipment budget. Management's August forecast was revenue of $108.0 billion, plus or minus 2%, for the following fiscal quarter. Microsoft finance chief Amy Hood said CPU and GPU purchases could be slowed if demand changed. That flexibility protects the cloud operator while exposing chip suppliers to deferred orders. Amazon's Trainium momentum also means cloud growth need not translate one-for-one into NVIDIA sales. If customer AI returns prove persistently weak, more cautious long-term order-growth assumptions would be warranted for NVIDIA and the memory, networking, and power equipment supply chain. The direct bearish signal would be operators reducing expansion or shifting chip purchases, before AI usage necessarily contracts.

Tech Stocks Tumble, Two Reasons to Consider Buying

Fri, 23 Oct 2026 08:03:21 GMT

Technology stocks are selling off again. The Nasdaq fell 2.6% on October 22, putting a question back in front of investors: is this an opportunity to buy the market's leaders, or the start of a more lasting reversal?

JPMorgan strategist Nikolaos Panigirtzoglou sees a reason to become more constructive: hedge funds have already cut substantial exposure, potentially limiting how much forced selling remains. DataTrek co-founder Jessica Rabe reaches a favorable conclusion from a different direction, arguing that technology's retreat still resembles pauses within a bull market rather than the beginning of a sustained downturn.

JPMorgan Thinks Much of the Forced Selling Is Behind Us

Panigirtzoglou and his team see limited further downside after hedge funds reduced equity exposure. Their reasoning starts with the relationship between semiconductor stocks and equity hedge-fund performance.

During the second-quarter rally, the VanEck Semiconductor ETF, SMH, and an equity hedge-fund performance index moved closely together. As chip stocks reversed, funds operating within strict risk limits had to reduce positions, amplifying the decline. The bank interprets the recent semiconductor retreat as evidence that much of the earlier long positioning has now been unwound, including positions held by momentum-driven commodity trading advisers.

The implication is that the selling process itself may be nearing exhaustion. When losses trigger compulsory reductions, prices can fall faster than changes in earnings expectations would otherwise warrant. Once those reductions have largely taken place, further bad news faces a smaller pool of positions that must be sold simply to meet risk limits.

That is the prospective change JPMorgan identifies. It does not require every concern about AI to disappear, or technology valuations to return immediately to their previous highs. It requires the market to move beyond a phase in which falling prices keep generating additional selling.

The bank acknowledges that near-term headwinds remain. Its argument is about a reduction in downside pressure, rather than a forecast for an immediate return to the earlier rally's pace. Thursday's renewed losses make the next few sessions important: stabilization in chip stocks would support the positioning thesis more convincingly than a one-day bounce that quickly reverses.

History Also Supports the Dip-Buying Case

Rabe's DataTrek analysis favors treating the pullback as a buying opportunity. At the time of her research, the Technology Select Sector SPDR ETF, XLK, was 6.7% below its highs, while its return over the preceding 100 trading days was negative 1.3%.

Those are two different measures of the retreat, but both fit her broader conclusion: technology has lost momentum without yet producing the sort of sustained losses associated with a major external shock.

Since 2015, XLK's rolling 100-trading-day price return has averaged 8.4%, and 81% of the windows studied have been positive. The recent bull market has been stronger still. Average returns over those roughly four-and-a-half-month windows rose from 6.2% in 2015-2019 to 12.3% from 2023 onward; the average since 2020 was 10.0%.

The summer rally was unusually strong even against that record. On August 21, 2026, the trailing 100-day gain reached 43.8%, exceeding the three-standard-deviation threshold of 42.6%. The window began at the depressed March 30 low associated with the U.S.-Iran conflict, so the starting point helped magnify the subsequent advance.

An extreme positive reading has not previously been an automatic sell signal. Rabe identifies two earlier initial crossings of that threshold since 2015. Both were followed by additional gains over the next 100 trading days:

Initial Threshold Crossing

Subsequent 100-Trading-Day Price Return

August 4, 2020

+13.8%

August 28, 2025

+8.7%

Source: DataTrek Research. These are forward price returns after the historical signals.

In both instances, the gains continued but became much smaller. Rabe notes that technology had advanced 2.9% over the 42 trading days after August 21 in her research snapshot. Her bullish argument allows for that slower pace; it does not depend on another surge comparable with the rebound from March.

There is also a precedent for brief negative readings after very strong advances. In May 2023, XLK's trailing 100-day return reached 35.6%, above the two-standard-deviation level of 31.2%. It later spent eight trading days below zero in October. In March 2024, the return reached 31.0%, followed by six days of negative readings that August during the Nikkei selloff and yen carry-trade unwind.

Earlier Rally

Later Negative 100-Day Readings

Trading Days With Negative Readings

May 2023: +35.6%

October 2023: -0.1% to -3.5%

March 2024: +31.0%

August 2024: -0.3% to -6.2%

Source: DataTrek Research. The ranges describe rolling 100-day returns, not cumulative losses over the six- or eight-day stretches.

Rabe places the latest negative 1.3% reading within that pattern. Her study associates the deeper setbacks with broader disruptions: the 2015 growth scare, late-2018 monetary tightening, the 2020 pandemic, the 2022 rate shock, the April 2025 tariff shock and the March 2026 oil shock.

Together, the two research approaches offer complementary reasons to consider buying. JPMorgan sees less positioning left to unwind; DataTrek sees historical room for a rally to survive a temporary loss of momentum.

Chips Matter More Than Their Market Weight Suggests

JPMorgan's second chart helps explain why semiconductor selling has been so disruptive. The bank puts the industry's share of MSCI All Country World Index market capitalization at 13%, against a 21% share of risk capital under its measure.

That gap makes the sector's behavior consequential well beyond its index weight. In JPMorgan's assessment, chipmakers have become a particularly useful gauge of global risk appetite. A recovery would matter not only for semiconductor investors but also for funds whose risk-taking has become closely tied to the industry.

It also makes stabilization in chips a better test of the bullish positioning argument than strength in unrelated defensive sectors. If the companies at the center of the unwind stop making new lows, the claim that forced selling is fading becomes more persuasive.