Bank stocks just had their worst month since the March 2025 selloff, and it isn’t the usual suspect — credit losses or a recession scare — driving the decline. The Invesco KBW Bank ETF (KBWB) is down 5.7% in September, its steepest monthly drop in over a year, with Goldman Sachs and Bank of America both off more than 8% and Morgan Stanley down nearly 8%. Brokerages have been hit even harder: Charles Schwab has fallen 9.8% this month and Interactive Brokers 8.3%.

Rising Treasury yields explain part of the selloff — higher rates squeeze loan growth and bank margins. But Apollo Global Management’s chief economist Torsten Slok has flagged a more structural threat that long-term bank shareholders should understand: what he calls an “agentic bank run.” His thesis is simple and unsettling. AI assistants like Meta’s newly launched Muse are increasingly capable of automatically sweeping idle cash into higher-yielding accounts. Fintech deposit products from firms like SoFi Technologies currently pay around 4.5% APY, compared to a national average of just 0.1% on traditional checking accounts. If millions of households let an AI agent quietly optimize their cash the way a robo-advisor optimizes a portfolio, the cheap deposits that banks rely on to fund loans could erode gradually, not all at once, but persistently.

Slok is careful to note there’s no sign of an actual bank run underway. This isn’t a 2023 Silicon Valley Bank-style crisis. It’s a slower-moving structural risk: a 44x gap between what SoFi pays and what the national average checking account yields is the kind of arbitrage that automation eventually closes. For patient investors holding traditional bank stocks for their dividends, the calculus on “cheap deposits as a durable moat” may need updating — deposit costs could creep higher over time even without a crisis, compressing the net interest margins that have underpinned bank profitability and payout growth for decades.

So what for long-term investors? This is a moat question, not a solvency question. Regional and money-center banks that have historically funded lending with near-free checking deposits may see that advantage erode as agentic AI makes yield-shopping frictionless for ordinary savers. Investors holding bank stocks for dividends should watch deposit cost trends and fee income diversification more closely than usual — and recognize that fintechs offering transparent, high-yield cash products may be quietly gaining a durable structural edge over legacy banks that are slower to compete on rate.

While most investors are fleeing housing stocks, Berkshire Hathaway just did the opposite — and the scale of the move is the real story for patient capital.

Berkshire has nearly doubled its stake in Lennar, the nation’s second-largest homebuilder, since the start of July. As of late last week, the conglomerate held roughly 25.9 million shares (common and Class B combined) worth about $2.1 billion, up 93% from the 13.4 million shares disclosed in its mid-August 13F filing. That pushed Berkshire’s ownership to 10.9% of Lennar’s roughly 238 million shares outstanding — crossing the 10% threshold that triggers mandatory “insider” disclosure under SEC rules, meaning any further trades must be reported within two business days instead of waiting for the next quarterly filing. Regulatory filings show Berkshire spent $212.4 million buying shares between September 17 and 21, then added another $136.4 million over the following three trading days. The stock jumped as much as 6.8% on the news and held a 5.2% four-day gain.

This isn’t a small side bet. It’s likely the work of portfolio manager Ted Weschler, and it comes while the broader housing market remains under pressure from elevated mortgage rates and soft new-home sales — exactly the kind of “hated” sector value investors have historically hunted in. Lennar trades well below its highs, and buying into a slumping cyclical business ahead of a recovery is a classic Berkshire playbook: accumulate quality at a discount when sentiment is worst, not when headlines turn positive.

So what for long-term investors: this filing is a visible, dollar-quantified signal that one of the most disciplined capital allocators in the world sees more upside than risk in housing right now, even with rates elevated. It doesn’t mean buy Lennar tomorrow — but it’s a reminder that cyclical sectors trading near multi-year lows can still harbor durable moats (land positions, scale, balance sheet strength) that patient capital is willing to accumulate aggressively while others wait for confirmation. The next full 13F in mid-November will show whether this was an isolated add or the start of a much larger position.

Tesla trades at roughly 159 times forward earnings, a multiple that would look absurd for almost any other company on the planet. But strip away the sticker shock and a more interesting story emerges: the market has effectively stopped valuing Tesla as an automaker at all. Long-term investors weighing a position here need to understand what they are actually buying — and how much speculative optionality is already baked into the price.

The numbers tell an uncomfortable story on their own. 2025 revenue slipped to $94.8 billion and net income fell to just $3.8 billion, with second-quarter operating margin compressed to a razor-thin 1.4% as pricing cuts and rising costs bit into the core vehicle business. Management is simultaneously funneling more than $25 billion into 2026 capital expenditures — money aimed almost entirely at Robotaxi, Full Self-Driving, and the Optimus humanoid robot. None of those three bets has yet produced a dollar of durable profit. The Cybercab robotaxi launch in Austin remains limited in scope and squarely in regulators’ crosshairs, while Optimus is, by Tesla’s own admission, still years from commercial scale.

That’s the trade long-term holders are making: a shrinking-margin car business subsidizing three unproven, capital-intensive moonshots, any one of which could redefine the company’s economics — or simply burn cash for years. Competition adds another wrinkle. Tesla is not alone in chasing autonomous fleets, and if robotaxi technology becomes commoditized across multiple platforms, the recurring, high-margin revenue stream investors are underwriting today may never fully materialize. Institutional conviction is telling, too: hedge fund ownership dipped to 116 funds at the end of the second quarter, down from 123, even as the dollar value of those combined stakes ticked up to about $23.8 billion — a sign that the funds still in the stock are doubling down, while others are stepping to the sidelines.

So what for long-term investors? Tesla isn’t a value stock, a dividend payer, or a business with a conventional margin of safety — and pretending otherwise invites disappointment. It is closer to a call option on three separate technology bets layered on top of a struggling car company. That can be a legitimate long-term thesis if you have genuine conviction in autonomous driving and robotics timelines, but it demands a much longer time horizon and much higher tolerance for volatility than a typical compounder. Patient investors should size any position accordingly, and watch operating margin trends — not the narrative — for the real signal on whether these bets are starting to pay off.

McDonald’s just told Wall Street it will spend $8.5 billion through 2036 rebuilding its own machine — and the market’s initial answer was to sell the stock down as much as 6.5% intraday. That gap between management’s confidence and investor skepticism is the real story for long-term holders, not the headline number itself.

The mechanics matter more than the marketing. Of the $8.5 billion, roughly $5 billion arrives by 2030, mostly as rent relief and capital support funneled to the 95% of restaurants owned by independent franchisees. Layer on top about $3 billion a year in baseline capital expenditures from 2027-2030, plus $1.5-2 billion in cumulative “capital partnering” support, and McDonald’s is betting productivity — not new units — drives the next leg of growth. Unit expansion is expected to contribute just 2% of systemwide sales growth by 2030, down from nearly 2.5% in 2027, meaning existing restaurants have to do more heavy lifting.

McDonald’s own math claims the investment generates 250 basis points of restaurant-level efficiency gains, worth roughly $100,000 in additional annual cash flow per average U.S. restaurant, with a four-year payback for franchisees after support. By 2030 the company is targeting operating margins in the low-to-mid 50% range, free-cash-flow conversion in the mid-to-high 80% range, and G&A held near 1.9% of systemwide sales — all while gaining 1.5 points of market share in chicken and beverages without ceding its beef lead. A generative-AI system called ArchIQ is meant to squeeze drive-thru throughput as part of the plan.

So what for long-term investors: the targets are specific enough to be checked against, which is the point. CEO Chris Kempczinski pinned some of Wednesday’s stock reaction on persistent inflation pressuring flat traffic in company-owned markets — a real near-term headwind that has nothing to do with the ten-year plan. Investors patient enough to track quarterly progress on margin expansion and free-cash-flow conversion, rather than reacting to a single down day, will know within a few years whether this was disciplined reinvestment in a wide-moat franchise system or an expensive bet on execution that never quite arrives.

The 30-year Treasury yield just touched 5.44% on Thursday, its highest level since 2004, capping a bond-market rout that has pushed yields across nearly every maturity to levels not seen since 2007. The 10-year climbed to 4.897%, its own 19-year high. For long-term investors, this isn’t background noise — it’s a fundamental repricing of the competition for every dollar in a portfolio.

The math is simple and uncomfortable. When the U.S. government pays 5.4% for locking up money for three decades, a dividend stock yielding 3% needs a much stronger growth or moat story to justify itself. Ed Al-Hussainy, a portfolio manager at Columbia Threadneedle, put it bluntly: “People are running out of superlatives for the yield on the 30-year bond. Investors are saying, ‘Look, if we’re going to lock up our money for 30 years, we need much higher compensation.'” That demand for compensation is being driven by a mix of persistent inflation, heavier government borrowing, and rising energy costs — Brent crude’s jump on Thursday was itself a contributing spark. Global debt has now crossed $365 trillion, a backdrop that keeps upward pressure on the term premium investors demand.

Treasury Secretary Scott Bessent expanded the government’s bond buyback program back in August specifically to cap long-term borrowing costs — and the move has had essentially no lasting effect on yields. That’s a signal worth sitting with: even direct policy intervention isn’t fighting the tide of fiscal concerns and inflation expectations right now.

So what for long-term investors? Rising discount rates compress valuations across the board, but they hit richly-priced, low-yielding growth stocks and highly leveraged companies hardest, since refinancing debt at 5%+ eats directly into margins. Meanwhile, quality dividend growers with low payout ratios, pricing power, and clean balance sheets become relatively more attractive — they don’t need a bailout from falling rates to keep compounding. It’s also a reminder that “risk-free” 5%+ yields on long Treasuries are themselves now a legitimate income option for patient investors, not just a hurdle rate for stocks to clear. The bond market is telling a story about the next decade of capital costs — ignoring it because it isn’t a hot stock pick would be a mistake.

AutoZone doesn’t pay a dividend, doesn’t split its stock, and rarely makes headlines for anything flashy. Yet its fourth-quarter results, reported this week, are a reminder of why the auto-parts retailer has quietly become one of the most disciplined capital allocators in the market — and why patient shareholders have been rewarded for decades without a single dividend check.

For the quarter ended August 29, net sales rose 5.6% to $6.6 billion, with total same-store sales up 2.7% and diluted earnings per share climbing to $56.05 from $48.71 a year earlier. Full-year net sales hit a record $20.3 billion, up 7.4%, while annual EPS grew 5.3% to $152.55. Management pointed to a stronger back half of the quarter after a sluggish start, with both domestic and international commercial sales accelerating — commercial has been the growth engine as AutoZone leans into professional mechanics rather than just do-it-yourself shoppers.

The more interesting number, though, is the one AutoZone doesn’t headline: buybacks. The company repurchased $697.5 million of stock in the quarter alone and $2.0 billion for the fiscal year, at an average price near $3,496 a share — a level so high the stock trades in the thousands rather than the tens or hundreds, with $1.6 billion still authorized for more. Since 1998, AutoZone has authorized $40.7 billion in total buybacks and executed $36.3 billion of it, methodically shrinking its share count for over 25 years. That relentless compounding is a big reason a $1,000 investment in AutoZone two decades ago is worth vastly more today than the underlying business growth alone would suggest — fewer shares means every dollar of profit is spread thinner across owners, and thicker per share.

There are real cracks to watch: gross margin was helped this quarter by a 145-basis-point tariff refund and a LIFO accounting quirk, not just organic pricing power, and operating expenses crept up 100 basis points as the company keeps opening new stores and Mega Hubs. Buybacks at nosebleed prices also carry opportunity cost if growth ever stalls.

So what for long-term investors: AutoZone is a case study in capital discipline over dividend optics. No yield, no splits, no drama — just decades of buying back stock and compounding per-share value while building out an 8,031-store footprint across the U.S., Mexico, and Brazil. It’s a model worth understanding even if the ticker itself never makes your watchlist.

3M’s operational turnaround is running ahead of schedule, and for patient investors in the century-old industrial giant, the numbers are hard to ignore. CEO William Brown told investors this week that on-time delivery has climbed from the low-80% range to roughly 90%, new product launches have more than doubled from 125 three years ago to 284 last year — with over 350 expected this year — and new products now account for a rising share of sales, on track to reach roughly 20% by 2027 versus about 11% when Brown took over. Second-quarter adjusted organic sales growth came in at 5.4%, and management is now targeting an operating margin above 25% by 2027.

There’s a genuine growth angle layered on top of the cleanup: 3M’s expanded-beam optical technology has become a data-center standard for Microsoft, opening a market Brown pegs at nearly $2 billion. That’s a small but real diversification story for a conglomerate too often dismissed as legacy industrial ballast.

The catch is a legal overhang that refuses to fully clear. A Montana federal judge recently declined — without prejudice — to dismiss a nationwide PFAS class action tied to firefighter turnout gear, and the Australian government is separately pursuing more than A$2 billion in damages over alleged misrepresentation of PFAS risks. That sits on top of 3M’s existing $10.5–$12.5 billion U.S. settlement covering public water system claims. Management’s framing that legal risk is “better defined” keeps colliding with fresh filings in fresh jurisdictions.

Institutional money seems to be leaning toward the operational story regardless: hedge fund ownership climbed from 63 funds at the end of Q1 2026 to 72 by the end of Q2, while short interest sits low at just 1.79% of float. So what for long-term investors: 3M looks like a case where the operating metrics are healing faster than the narrative admits, but real tail risk from PFAS litigation hasn’t been priced away entirely — meaning the actual margin of safety here depends on how much of that legal uncertainty the market has already discounted into today’s valuation.

Aflac just extended a 43-year streak of dividend increases, and the mixed headline numbers behind that milestone tell long-term investors more about durability than any single quarter could. Net earnings jumped 37.7% to $825 million in the second quarter, while adjusted earnings — the measure management prefers — actually fell 7.7% to $883 million. Both figures are accurate; they’re just answering different questions, and the gap is almost entirely currency, not operations.

Strip out the yen, which averaged 159.45 to the dollar and ran 9.3% weaker than a year ago, and adjusted earnings per share actually rose 4.1% to $3.57. Japan’s business is getting more profitable too, with the pretax adjusted margin widening to 34.3% from 32.0% as claims took a smaller bite out of premiums. On the U.S. side, net earned premiums grew 2.3% to $1.5 billion and sales rose 2.6% to $349 million, led by group voluntary benefits, dental, and vision — steady, unglamorous growth that compounds.

The soft spots are real and worth tracking. Japan’s yen-denominated premiums fell 3.7% on a reinsurance deal and older policies reaching paid-up status, and customer persistency slipped to 92.7% from 93.7%. The U.S. margin narrowed to 20.9% from 22.5% as claims ran hotter, and adjusted book value per share dipped to $41.22 from $42.97. Hedge fund ownership dropped to 39 funds from 46 — institutions are trimming — but short interest sits at just 2.84% of the float, suggesting indifference rather than a bearish bet.

None of that stopped Aflac from returning $1.3 billion to shareholders in a single quarter, $983 million of it via buybacks, alongside a $0.61 dividend the board says it intends to keep raising. At 15.2 times forward earnings, the stock is priced for stability, not growth — which is exactly what a 43-year dividend grower with a genuine moat in supplemental insurance should offer. So what for long-term investors: this is a story about persistence, not fireworks. The currency noise will fade, the Japan reinsurance drag will roll off, and the dividend streak is the more reliable signal than any single quarter’s adjusted EPS line.

Most financial firms watch their margins compress as they scale — more customers usually means more overhead, more compliance, more cost creep. Interactive Brokers is doing the opposite. In its most recent quarter, the brokerage turned 77 cents of every revenue dollar into pretax profit, up from 75 cents a year earlier, even as it added over a million new customer accounts. That is the kind of operating leverage long-term investors dream about but rarely find.

The numbers behind it are striking. Customer accounts grew 34% year-over-year to 5.19 million, but customer equity grew even faster — up 40% to $930.3 billion — meaning the average account is getting richer, not just more numerous. Margin loans jumped 67% to $108.5 billion, far outpacing account growth, and net interest income rose 23% to $1.06 billion, now the single largest line on the income statement. Commission revenue climbed 30% to $673 million, driven by a 17% rise in options volume and 14% in stocks. Earnings per share rose to $0.69 from $0.51 a year earlier.

None of this comes without fine print. Execution and regulatory fees rose in step with trading activity, and the firm’s unusual practice of holding equity in a 10-currency basket — the GLOBAL — cut comprehensive earnings by $36 million this quarter on a mere 0.21% currency move. Hedge fund ownership climbed to 87 funds from 70 last quarter, and short interest sits at a thin 2.64% of float, suggesting little organized skepticism. The stock trades at nearly 28 times forward earnings, a full price that assumes growth in accounts, balances, and borrowing keeps compounding without interruption. The quarterly dividend of $0.0875 per share is modest by design — this is a business reinvesting for scale, not returning cash.

So what for long-term investors: Interactive Brokers has built a genuine moat in low-cost, high-volume global brokerage infrastructure, and its margin trajectory shows real pricing power and operating discipline rather than financial engineering. The risk is that today’s valuation already prices in years of continued account and balance growth — a business this dependent on trading activity and margin borrowing can see its economics shift quickly if markets turn quiet. Patient investors should watch whether interest income and margin loan growth hold up in a slower-trading environment before assuming the current multiple is cheap.

NextEra Energy just reaffirmed a growth plan that matters far more than this week’s stock chart: at least 8% annual adjusted earnings growth through 2032, a dividend rising roughly 10% a year through 2026 and 6% annually after that, and a $67 billion all-stock acquisition of Dominion Energy that would create the largest electricity producer in the United States. For patient, dividend-focused investors, this is a case study in how utility consolidation can compound wealth quietly for a decade while headlines chase AI chip names.

The mechanics matter. NextEra says Dominion is expected to be immediately accretive to earnings, pushing the combined company’s adjusted EPS growth above 9% annually through 2032, extending to 2035 off a 2025 base. Together they would run the country’s largest natural gas generation fleet and the second-largest nuclear fleet — a scale advantage that’s nearly impossible for competitors to replicate given how long it takes to permit and build power plants. That’s the kind of structural moat long-term investors should care about: not a clever product, but physical infrastructure competitors can’t quickly copy.

Smart money is already positioning, though not uniformly. Hedge funds holding NextEra grew to 80 in the second quarter from 74 in the first. Marshall Wace increased its stake 350% to 3.37 million shares, and Balyasny raised its position more than 20-fold to 1.62 million shares — while GQG Partners, the largest holder, trimmed 20%. Short interest sits at a modest 2.39% of float, suggesting the market isn’t betting heavily against the deal, but it isn’t a slam dunk either.

The real risk is regulatory, not financial. The deal hinges on Virginia approving it by the second half of 2027, and NextEra and Dominion have sweetened the offer — doubling residential bill credits to four years and adding $200 million combined for low-income assistance and workforce development — specifically to shield ratepayers from costs tied to Northern Virginia’s data center boom. If regulators impose conditions that erode the deal’s economics, or approval slips, the growth thesis weakens considerably.

So what for long-term investors: this isn’t a trade on a merger closing next quarter. It’s a bet on whether disciplined capital allocation and regulatory patience can turn two already-growing utilities into a compounding dividend machine with a scale moat that took decades to build and can’t be replicated overnight. Watch the Virginia regulatory timeline, not the daily stock print.