Buffett & Munger, 1994: Diversify Only If You Don't Understand
Buffett and Munger argue investors should concentrate in businesses they understand and diversify widely only when they lack conviction.
Saturday, October 3
Everything we published on this day.
35 stories — the front page that day →
Buffett and Munger argue investors should concentrate in businesses they understand and diversify widely only when they lack conviction.
Muse / Instinct form factor has been a big unlock for “normies” while more technical people got agentic capabilities with Claude Code and OpenClaw. I think your $MSFT autopilot will deliver a similar unlock to non technical enterprise workers. Are you using it internally? When can we expect it GA?
$GOOG TPU cost per million tokens vs $NVDA Blackwell. Advantage: Google Will be interesting to see the 3P TPU ecosystem evolve now that GCP is focused on it
$AMPL moving off Claude over this absurd behavior.
800VDC power delivery could curb AI rack currents above 10,000A, creating semiconductor opportunities in sidecar conversion, switching,…
The ransomware software of Nubeva Technologies $NBVAF (TSXV: $NBVA) passed every test in a month-long third-party lab evaluation in 2023, and its biggest customer, Kroll, walked away anyway. In January 2022, Nubeva launched Ransomware Reversal, or NuRR, the product the company is now built around. It’s a lightweight agent installed on servers, desktops, and laptops that watches for the file encryption a ransomware attack produces, captures copies of the encryption keys as the malware generates them, and stores those keys on the customer’s own network. After the attack, Nubeva uses the captured keys to build a decryptor (a program that reverses the encryption) and hands the customer its data back, without the ransom. The only outside test of NuRR I found was a month-long evaluation in 2023 at DreamPort, a lab in Maryland (the test doesn’t appear to have been commissioned by Nubeva). DreamPort is run by the Maryland Innovation and Security Institute (MISI), a nonprofit that partners with U.S. Cyber Command to connect the military with outside cybersecurity products. Per Nubeva’s July 20, 2023 release, NuRR passed all 17 tests in MISI’s test plan and captured the encryption keys in every attack, including ones using the LockBit, Ragnar Locker, and BlackBasta ransomware strains. But less than two months after the MISI results, Nubeva lost its biggest customer: → January 2023: Nubeva named Kroll, the risk advisory firm that runs more than 3,000 incident-response investigations a year, as a strategic partner using the product in its ransomware cases. → September 2023: Kroll ended the relationship and paid $180,000 to walk away. It had been 86% of FY2023 revenue (~$1.78M of $2.07M). Neither company gave a reason in a filing or press release. But in a July 2024 video on Nubeva’s YouTube channel, CEO Randy Chou said Kroll sometimes brought in 10 ransomware cases on the same weekend, and “neither the partner or us as a complete solution could handle the emergency services.” In an October 2025 video, he explained the constraint further. Each new ransomware variant needs its own decryptor, which Nubeva’s engineers built by hand. With 10 to 30 attacks running at once, “you don’t have enough engineers to go figure out the new decryptor.” Chou added that Nubeva “simply cannot afford to scale” its support, “and that’s still the reality,” so it’s waiting on automation before it grows again. A longtime investor I spoke with, who has followed Nubeva for years and has spoken with people who’ve used NuRR in the field, described the same constraint. He also says Kroll’s exit is why Nubeva stopped trying to grow on its own and bet on AI to write the decryptors.
McCreedy West had “a few weeks” of development (the tunnels that reach new ore) ahead of its miners and no stopes (the blocks of ore being mined) ready to go when Magna Mining $MGMNF (TSX: $NICU) took it over in March 2025. So Magna spent 2025 catching up and missed its H2 2025 grade guidance, shipping ore at 2.64% copper equivalent (CuEq, which converts every metal into copper at set prices) in Q3 and 3.41% in Q4, below the 2.9-3.4% and 3.8-4.4% it guided. 2026 has gone better. In Q2, McCreedy West set Magna-era records for tonnage (98,446 tons) and cash margin, and production costs fell to US$144/ton processed from US$156 in Q1. Through H1 2026, McCreedy West produced 8.62M payable pounds of CuEq (the metal the mill and smelter pay for after recoveries). It needs 7.4-9.4M lbs in H2 to land in the 16-18M lbs guidance range, so repeating H1 would put the year inside guidance. Costs are inside guidance too. H1 cash costs (operating costs plus smelting and refining charges, per payable pound) of US$3.63/lb sit within the US$3.40-3.80 range. All-in sustaining costs (AISC, which add sustaining capital and corporate overhead) of US$4.39/lb also sit within US$4.20-4.70. Both exclude the precious metals stream held by Franco-Nevada $FNV, which buys half the mine’s gold, platinum, and palladium at a discount. Magna says the stream adds US$0.78-0.92/lb at its 2026 budget prices. Magna’s first reserve estimate for McCreedy West, effective December 31, 2025, came to 987,000 tonnes of probable reserves, which Magna says supports three years of production at its 2026 mining rate. Three years sounds short, but CEO Jason Jessup says it’s “in line with the reserves that McCreedy West has operated with since being restarted in 2003 by FNX Mining,” and that Magna “could have extended” the reserves. McCreedy West’s Intermain nickel zone isn’t in the reserves or the guidance. Jessup wants nickel to hold “north of $8” per pound “for at least a quarter” before restarting it, and says Magna could be shipping Intermain nickel to Vale $VALE “within 8 to 12 weeks” of a decision. At ~$7.48/lb nickel, the Intermain restart waits.
Magna Mining $MGMNF (TSX: $NICU) paid US$6.5M for a package of assets from KGHM $KGHPF, the Polish copper giant, including McCreedy West, now its only producing mine, and Levack, its flagship restart project. It paid ~US$3.7M in cash and ~US$1.4M in shares at closing, with ~US$1.4M due on December 31, 2026. It could also owe up to US$16.9M in milestone payments as the mines reach commercial production, including US$4.2M for Levack. Counting the milestones at their US$4.7M fair value, Magna valued the consideration at US$11.2M against US$25.0M of net assets. That’s why its 2025 results include a US$13.8M “bargain purchase” gain (US$25.0M − US$11.2M). It also took on reclamation liabilities with a fair value of US$9.8M and a precious metals stream held by Franco-Nevada $FNV, the royalty and streaming company, which it booked as a US$22.6M liability at the acquisition. Here’s Magna’s pipeline, from production to exploration: → McCreedy West (producing): Run since March 2025. All of its 2026 guidance comes from the 700 Footwall Copper Zone. → Levack (restart study): Connected underground to McCreedy West and idle but maintained since KGHM stopped mining in 2019. → Crean Hill (permitted): Past-producing nickel, copper, and precious metals mine southwest of Sudbury, Ontario, with a pre-feasibility study (PFS) due in the first half of October. → Podolsky and Shakespeare (later): CEO Jason Jessup pencils in Podolsky as the fourth mine “probably sometime in 2029,” and says Shakespeare, which would need its own mill, is “probably 5 years down the road.” → Exploration: 584 km² of claims across the Sudbury Basin. Copper pays most of the bills. In H1 2026, copper made up 68.4% of McCreedy West’s metal revenue, nickel 10.3%, and platinum, palladium, gold, and silver 21.2%. So despite the ticker (NICU, for nickel and copper), Magna today is a copper producer with nickel upside. Magna doesn’t own a mill. It trucks McCreedy West’s crushed ore to the Clarabelle mill run by Vale $VALE, pays Vale a fixed price per ton for trucking and milling, and sells the ore to Vale on delivery. It also has an ore selling agreement with Glencore $GLNCY for Glencore’s Strathcona mill. Magna’s financial statements report just two significant customers, and with McCreedy West’s ore going to Vale, changes in Vale’s terms hit revenue directly. In Q2 2026, a retroactive adjustment to the Vale contract roughly doubled the smelting and refining charges taken out of Magna’s revenue, to US$2.9M from US$1.4M. Incoming CFO Greg Huffman expects the charges to normalize in H2 2026.
@mkfilko How you be walking with $NKE shoes
Bernstein expects copper to run short from 2027, and the IEA says a new copper mine takes ~17 years to go from discovery to production. So I found a miner restarting old ones, with a CEO who restarted its flagship once before and says it could produce 2-3x today’s output. Its restart study is due in the first half of October, and a much richer discovery at the same mine isn’t in the study yet. That company is Magna Mining $MGMNF (TSX: $NICU), a copper miner in Sudbury, Ontario. After more than a century of mining, Sudbury has plenty of past-producing mines to restart. Per Magna’s September deck, Brazil’s Vale $VALE and Switzerland’s Glencore $GLNCY are the only other companies with significant property holdings in the Sudbury Basin, and they run the district’s two mills and two smelters. Magna has agreements to sell its ore to both, so its next restarts don’t need a new mill. Magna’s stock trades at CAD$2.45/share, down 9% y/y and 38% below its CAD$3.94 intraday high on January 29, 2026. Counting its convertible notes, in-the-money options, and share units as shares, and netting out its cash plus the ~US$5.1M from exercising the options, enterprise value is ~US$454M. That buys you McCreedy West, the producing mine that carried Magna to its first quarter of positive free cash flow (FCF) in Q2 2026. You also get Levack and Crean Hill, two past-producing mines with a preliminary economic assessment (PEA) and a pre-feasibility study (PFS), respectively, due in the first half of October. Notably, Magna’s leaders have run McCreedy West and Levack before. Both mines belonged to FNX Mining, a Canadian miner, before they passed to KGHM $KGHPF, the Polish copper giant that sold them to Magna. CEO Jason Jessup led the FNX team that restarted Levack, and chair Vern Baker was FNX’s VP of Operations. Jessup also says Magna doesn’t plan any more financings after Alpayana, a Peruvian mining group, invested US$99M in August. So Magna could triple or quadruple its output (McCreedy West plus 2-3x from Levack) without raising more money, and R2, the high-grade zone it found at Levack, would come on top of whatever the PEA shows. The question is whether CAD$2.45 is an early price for a multi-mine Sudbury producer, or a price that already assumes Levack restarts well before its numbers are public.
Elmet $ELMT won a ~$36M U.S. defense stockpile contract for metal its plants already make, while deliveries under its $2B stockpile contract wait on new supply. On September 30, Elmet announced the award from the Defense Logistics Agency (DLA), which manages the U.S. National Defense Stockpile. It covers TZM (titanium-zirconium-molybdenum alloy) billets, bar stock, and ingots, plus molybdenum and tungsten wire, materials used to make jet engine components. Elmet's $2B stockpile contract, announced September 14, is separate and covers tungsten concentrate, the upgraded ore a mine ships. The $36M is 27.4% of Elmet's $131.5M total backlog at July 3 ($36M / $131.5M), and the release doesn't say when deliveries start or how long the contract runs. On the slide for the $2B contract, Elmet's October 1 deck says Elmet "does not intend to begin delivering material into the National Defense Stockpile until sufficient incremental new supply is available." Elmet's supply agreement with Tungsten West (AIM: TUN) puts concentrate from Hemerdon, its tungsten mine in Devon, England, first in line for Elmet's DLA deliveries until Elmet's Nevada refinery is commissioned (targeted for 1H 2029). Elmet expects >1,000 tonnes a year of Hemerdon concentrate, and the mine targets full production by the end of Q1 2027. In September I treated $1.85B of the $2B ceiling as an upper bound, not orders Elmet has in hand, and the $36M sits outside the $2B entirely. So the stockpile's newest award goes to metal from Elmet's existing U.S. plants, while the $2B contract still waits on a mine in England that isn't at full production yet.
M&T Bank's 10.4% five-year revenue CAGR, improving net interest margin and commercial loan growth support its earnings resilience despite a…
Adding long exposure (through $IWM and $RSP mainly) makes me feel sick, which backs up my belief that the timing is good to add long exposure.
@TSOH_Investing this is what $nke tried to do unsuccessfully
@coffeesnob @DELAESPRIELLAE Fun because this is 1. my local airport and 2. an $ASR property.
@UsefulSupplyCo @grok for $ARTNA respond with the correct figures about what I mentioned and is there any current analyst coverage? If so how many, who and what price target
@whatsgoingoninm $HHH is a little different. The fee structure is weird and they have a weirder collection of assets, not just a bunch of publically listed stocks anyone can buy. Though no capital returns. Still not a great structure but not as egregious as PSH.
$FEIM (defense) was a great example of this. Not only was backlog growing, it was replacing sub 30% gross margin backlog with over 40% GM. You had all day to buy the stock between $8 & $15 in 2024. Now it’s pushing $100. $SIF (defense) might be approaching that type of set-up.
@realroseceline Kind of crazy that $APP was at a quarter trillion market cap
@SayNoToTrading @YehiaElgohary Would need to do a stock split, $1000 share price ruins a price-weighted index. $TXN has to get in on one of these Dow additions eventually.
@CompoundingEV There is no semi comprehensive tweet on $PAHC, there is no where for people to see a snapshot of it in such a pleasing manner unless they have value line. To me at least it’s nicer reading a quick pitch complied by a human than asking AI
You can't find the Switch Two Zelda Edition anywhere. It is sold out at GameStop $GME. This 40th Anniversary game for Ocarina of Time is going to sell so much hardware and accessories. And then you have GTA 6 coming out too. Insiders at GameStop are loading up on shares like crazy. They see internal sales and how much money they are about to print.
$PAHC: 9x cash flow. 26% ROE Animals have to be healthy whether the economy is good or bad, and livestock producers have an economic incentive to spend money on products that reduce mortality, improve feed conversion, increase yield or prevent disease. 22% Adj EBITDA margin. JPM $55 PT. The company operates through the following business divisions: Animal Health division manufactures and sells animal health products such as antibacterial, anticoccidials, nutritional specialty products and vaccines used in the animal health industry to treat or to prevent bacterial diseases to improve the animal’s health. Mineral Nutrition division manufactures and markets formulations and concentrations of trace minerals, such as zinc, manganese, copper, iron, and other compounds. Performance Products division manufactures and markets various specialty ingredients for use in the personal care, industrial chemical and chemical catalyst industries. The most tangible new initiative is the planned closure of Chicago Heights, Illinois. Phibro acquired the facility as part of the Zoetis deal, reviewed its manufacturing network and decided it no longer made economic sense to keep the facility open. Production is expected to cease in summer 2027, affecting roughly 100 employees. Production will be shifted to other Phibro facilities and selected third-party manufacturers. Phibro Animal Health Corporation Management says the point is straightforward: align manufacturing capacity with future needs, improve efficiency and strengthen competitiveness. Phibro Animal Health Corporation The company expects approximately $10m of one-time cash costs and about $10m of capex, with eventual annual EBITDA benefit estimated around $15–20m, largely from FY2028 onward Regulatory optionality: Brazil virginiamycin One of the most interesting hidden catalysts is virginiamycin in Brazil. FY2027 guidance assumes only minimal sales because Phibro has not yet obtained the required therapeutic indications. FY2026 virginiamycin sales in Brazil were roughly $27m. Management explicitly said a favorable regulatory result could therefore provide modest upside to guidance.
$FSBW: Bought Pacific West Bank in August for 30m. Exp +7.5% EPS growth in 2027 due to this. PWBK has a 92% deposit mix NOT CDs. (attractive b/c lower cost). First Bancorp a bit cyclical due to home lending, consistent double digit ROE, conservative way to play eventual housing rebound. The problem was not that Pacific West had no franchise. It had a recognizable Portland commercial-banking niche, a relatively attractive deposit base, and specialized relationships with SMBs, nonprofits, professional-services businesses and digital partners. The problem was scale. Management acknowledged that Pacific West's size and market position made both organic and strategic growth difficult, while the cost and complexity of running a regulated community bank had risen materially after 2008 and 2020.
$RENT Unique business model Being investigated for fraud 98% of customers rent a brand never owned, this is valuable marketing to the brands Lot of consumer data for past 18 years New CEO (Sept 14) focusing on core competencies Near 100% retention of brands since 2009 Acquisition target? The concise thesis RENT has spent years proving that women will rent clothes and almost as long proving that customer enthusiasm doesn't automatically produce shareholder cash. The turnaround is now more credible than it has been historically: revenue is growing rapidly, Q2 gross margins expanded 600+ bps, add-ons are booming, inventory sourcing is becoming less capital-intensive, debt was materially restructured, management is eliminating side projects, and AI-driven discovery is producing measurable engagement improvements. Rent The Runway, Inc RENT has never yet produced positive full-year owner earnings on this conservative cash definition. Rent The Runway, Inc So the thesis isn't “this is already a great business.” It is much narrower: If RENT can retain the revenue and customer improvements created by last year's inventory rebuild while reducing annual inventory investment toward ~$54m, shifting more merchandise onto brand-funded/revenue-share structures, and maintaining pricing/add-on gains, the company may finally cross from Adjusted-EBITDA profitability into genuine owner earnings.
$ARTNA: microcap water utility at lowest P/E ever, 4% dividend. 9x cashflow. Family run for 115 years. Sussex county development expands regulated customer base. Wastewater growth engine. Its principal subsidiary, Artesian Water, is the oldest and largest regulated water utility on the Delmarva Peninsula. The company generates stable, recurring revenue by distributing and selling water and wastewater services to residential, commercial, industrial, and municipal customers under regulatory frameworks. In addition to its core regulated utility operations, Artesian Resources earns supplementary income from non-regulated activities including contract operations, infrastructure development, and service line protection plans, creating a business model that combines predictable utility income with service-based revenue streams.
$LULU peaked at 14x EV/S and 82x NTM PE in 2020. $NKE EBITDA margin has collapsed from 18% in 2021 to 10.7% today.
Chipmakers are backstopping AI compute financing with guarantees, equity and loans, delaying a potential neocloud debt squeeze.
Howard Hughes plans to raise nearly $4 billion by selling real estate assets and condos, accelerating Vantage’s insurance expansion while…
$FJET prospectively and retrospectively more than doubling CEO pay. While it is true that $350k is not a huge CEO salary, it is $350k more than Starfighter's revenue under his tenure.
AI could shrink $MNDY’s human seats while multiplying the work flowing through its platform. Our latest SaaS Survival Scorecard examines its AI harness, competition from Claude and the key question: can growing agent revenue more than offset lost seat revenue?
AI infrastructure’s $6.3 trillion financing gap is driving structured debt and asset-backed solutions, starting with Meta’s planned 5GW…
Walmart's FY28 consensus requires 15–18% operating-income growth above plan despite diesel costs, price investments and a non-recurring…
Magnolia Oil & Gas closed its $4 billion WildFire acquisition, adding Eagle Ford acreage with cost synergies and funding the deal with…
Phillips 66 expects a structural refining-capacity gap to sustain crack spreads, driving $27.3 billion of after-tax earnings over four…