Yet Another Value Podcast

Nubank ($NU) has 140 million customers, roughly 60% of Brazil's adult population, an efficiency ratio around 20% versus 40-60% at the legacy banks, and ROEs in the 30s. Evan Vanderveer of Vanshap Capital has owned it for four years and thinks the market is still treating it like a risky EM bank instead of what he thinks it is: a tech company that happens to hold deposits, with a founder (David Vélez) who controls it and a runway that runs through Brazil's $100 billion banking profit pool, Mexico, Colombia, and eventually the US.
My pushback is the Capital One question. Capital One was the smartest data-science lender in the room, IPO'd in 1994, went up 13x in 12 years, and then spent the next 20 as a mature bank that lagged the market. Nubank was built by ex-Capital One people, so is this 1994 or 2006? We also get into what the right cost of equity is for a Brazilian bank trading at high-teens earnings with a 30% ROE, whether MELI and Kaspi tell you EM fintech never gets a big multiple, the 13,000-customers-per-employee stat, Brazil NPLs at 15-year highs, the wave of senior departures, whether any banking fintech has ever expanded across borders, Vélez joining OpenAI's board, and my bigger worry that AI eventually commoditizes every financial product and competes away the 30% ROE.
This episode is sponsored by Trata: https://trata.com/nu. Trata is two sharp buy-siders hopping on an anonymized call to talk through the risks and upside of a stock, and it's the closest thing to this podcast in written form. Go to trata.com/nu for a free Trata transcript on Nubank that I read and used heavily prepping for this call.
Chapters:
(00:00) Intro and Trata sponsor read
(01:55) Evan Vanderveer / Vanshap Capital joins
(02:50) What is Nubank: 140M customers, 60% of Brazil, 20% efficiency ratio
(06:11) What the market is missing: deepening relationships, Mexico's ARPAC
(08:05) The Capital One DNA: QED, Nigel Morris, data science
(10:38) My pushback: is this Capital One in 2006, not 1994?
(13:00) Brazil's $100B profit pool, payroll loans, David Vélez's control
(15:09) Valuation: 30% ROE, high-teens P/E, and the right cost of equity for a Brazilian bank
(18:47) MELI and Kaspi: does EM fintech ever earn alpha?
(21:59) Fintech or bank? SoFi 2021, lending competition, too big to fail
(23:55) 13,000 customers per employee vs 1,300 at legacy banks
(26:15) Brazil risks: NPLs at 15-year highs, the Selic, October's election
(27:45) How much of the value is Brazil vs Mexico, Colombia, and the US
(29:42) Can a banking fintech expand across borders? The Citibank precedent
(31:03) Senior departures, the new CFO from Visa, capping US investment
(33:47) Buybacks in the low $12s and the risk of losing local expertise
(36:32) Valuation bet, business bet, or jockey bet?
(38:52) David Vélez joining OpenAI's board
(40:46) AI inside Nubank: 60% of inquiries, Devin agents, faster credit models
(42:44) Does AI commoditize banking and compete away the 30% ROE?
(46:54) The US expansion: God kings or a real niche?
(50:35) Closing thoughts
(52:07) Disclaimer
Links:
Yet Another Value Blog - https://www.yetanothervalueblog.com
See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer
Production and editing by The Podcast Consultant - https://thepodcastconsultant.com/

DNOW spun out of National Oilwell Varco at $35 in late 2014. A year later it was $13. Today it is around $16. Steve Gorelik's argument is that ten years of that chart is one long headwind rather than a broken business: 1,800 US rigs at the spin, under 600 now, global oil and gas investment 40% below 2014 in real dollars, and DNOW still grew margins and bought companies at 4 to 5x EBITDA the whole way through. Rigs have started ticking back up. The MRC Global merger brings $75m of synergies to two businesses that earned $325m of EBITDA apart in 2024. Management has soft-targeted $350m of EBITDA for 2027 against roughly a $3.5B enterprise value, which Steve gets to about $300m of free cash flow on a $3B market cap.
My pushback is that 10x is not deep value, and the double comes almost entirely from multiple expansion back to the 5 to 6% free cash flow yield the market used to pay. Why is 10x the wrong number and not 12 or 14? We also get into the acquisitive compounder paradox, whether the incremental drilling actually shows up in US shale or somewhere else, the Oracle implementation they inherited from MRC and why they are now running it alongside SAP on purpose, the $50m of stock they bought back in the middle of that mess, and whether a business private equity would happily lever to four or six turns belongs in the public market at all. Steve's 2029 case is $30 to $32 per share.
This episode is sponsored by Fiscal.ai: https://fiscal.ai/yav. I am a customer and I pay for their API with my own money. Two things I use it for constantly. First, they have a huge database of fund letters wired into the API, so when I am prepping a podcast or looking at an event my agent pulls every recent letter on the name and tells me what the bull and bear cases actually are. Second, financials with sourcing attached: I ask for a model and every line links back to the company specific KPI, segment, or ratio it came from, so I can click through and see exactly where the number is from. Use my link, fiscal.ai/yav, for 15% off their AI connector.
Chapters:
(00:00) Nobody gets excited about a distributor
(03:48) What DNOW actually sells
(05:29) The roll-up playbook, without the leverage
(07:13) Why the 2014 spin never worked
(12:47) My pushback: does the drilling come back in the US?
(14:23) Shale payback periods and rigs getting less efficient
(16:50) The MRC Global deal
(18:04) Upstream plus downstream: what the combination buys you
(21:24) The ERP implementation they inherited
(24:39) Why 2027 guidance sits below what the two did apart
(28:16) Free cash flow yield as the North Star
(32:40) Buying growth at 4 to 5x while trading at 8 or 9
(34:42) Paying down debt and buying back stock at the same time
(35:38) $50m of buybacks in the middle of the mess
(37:15) Running SAP and Oracle side by side on purpose
(39:49) 1,907 rigs at the spin, 571 today
(40:40) The 2029 case: $30 to $32 per share
(41:01) Should this company even be public?
(42:56) Would private equity lever it up?
(43:31) Water, utilities and data centers
(45:21) Why boring distributors compound
Steve Gorelik / Firebird Management: https://fbird.com
Links:
Yet Another Value Blog - https://www.yetanothervalueblog.com
See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer
Production and editing by The Podcast Consultant - https://thepodcastconsultant.com/

Paul Cerro was long Hims & Hers in 2024, short it through the compounded GLP-1 unwind, and covered when the stock broke $14 after Q1. He's long again, and his thesis has almost nothing to do with peptides, testosterone, or the international launches everyone else is excited about. Those, he says, are table stakes. Hims has never had a problem acquiring customers. It has a problem keeping them, and subscriber counts have barely moved in three quarters. His argument is that labs and patient data are what push retention and LTV up, and that is the part the market isn't paying for.
I push back in three places. The data play doesn't look unique to me: Whoop and Oura own a wearable and a daily interaction, Hims owns commoditized blood work, and if Hims does unlock it, Apple or Amazon can walk in on top of them. The 2030 targets ask you to double trust management, once on 4x-ing EBITDA and again on a very heavily adjusted EBITDA number, from a CEO Paul says he wouldn't trust to walk his dog. And when peptides go legal, I think a hundred Instagram churn-and-burn startups compete away the customer acquisition edge. Paul's answers are worth the hour, especially the balance-sheet argument for a price war and the Ro story. We close on what to watch in the August 10 print.
Paul's Hims & Hers write-up: https://www.cedargroveresearch.com/p/hims-whoever-controls-the-data-controls-the-industry
This episode is sponsored by Trata: https://trata.com/hims. Trata is two investors who hop on and talk about a stock they're both in, sometimes one long and one short, sometimes both on the same side, but always about what actually drives the stock up or down. Trata now has an MCP, so you can point your AI agent at a company and pull the transcript, which is one of the first things I do when I start looking at a name. They have four HIMS calls, all less than a year old and one about a month old, and if you follow the link you can get their most recent HIMS coverage as a free trial.
Chapters:
(00:00) Long it, shorted it, now long again
(02:57) Paul on the setup right now
(04:17) What he learned building Ro
(05:19) How cash-pay healthcare actually works
(11:54) The original 2024 Hims thesis
(13:26) The compounding loophole and its expiration date
(15:58) Covering the short and going long again
(18:54) Acquisition was never the problem, retention is
(20:35) Why the money in healthcare is chasing data
(22:33) My pushback: what is unique about Hims' data?
(26:32) Hims versus Whoop, Oura and the Apple Watch
(29:21) Valuation: 30x 2026 EBITDA, 6x 2030
(32:56) Why international makes the targets conservative
(34:12) Double trusting a heavily adjusted number
(36:16) Icarus, Napoleon and the Teflon Don
(40:05) On putting too much faith in regulators
(43:49) Peptides and the market nobody has priced
(45:36) Chinese peptides and what is in the vial
(50:12) Can a hundred Instagram startups undercut Hims?
(54:40) Why the balance sheet decides a price war
(56:13) What to watch in the August 10 print
(57:40) CVS, Walgreens, Walmart and Amazon
(1:00:49) Closing thoughts
Paul Cerro / Cedar Grove: https://www.cedargroveresearch.com
Links:
Yet Another Value Blog - https://www.yetanothervalueblog.com
See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer
Production and editing by The Podcast Consultant - https://thepodcastconsultant.com/

Strategy filed an 8K this morning: they sold $300 million of stock, sold $100 million of bitcoin, and put the proceeds into buying back roughly $80 million of their own preferred at a discount. So you have a company trading over NAV, diluting shareholders, and selling the one asset it exists to hold, and it is still trading at a premium. Management is presenting the move from "one-way capital issuer" to "multi-way capital issuer" like it is a financial engineering breakthrough. It is just normal capital allocation, and it took them five years to get there.
That is the through line for this whole ramble: selling. Strategy made a genuinely good call in 2020 and then never sold a thing. Situational Awareness made a generational call, long AI winners and short AI losers, went up something like 10x on it, never rebalanced, kept pressing a trade that naturally degrosses, and blew up when software went from 20 back to 30 and semis went from 400 back to 370. I do a miniature version of the same thing every time I decide in advance that I will start trimming at 15 a stock I bought at 10, and I am not sure that plan is as smart as it feels. I also get into why thematic trades are so hard to sell when there is no price to anchor to, whether the crossover funds actually had an AI information edge or just conviction, why I cannot make the memory valuations work under any assumption I am willing to make, and where I think the real opportunity is: the beaten-up AI power names Situational owned in size, several of which are not trading far above the DCF of the contracts they already have.
Situational Awareness, crossover funds, and the AI edge: https://www.yetanothervalueblog.com/p/crossover-funds-ai-edge-situational-awareness
This episode is sponsored by Trata: https://www.trata.com/glxy. Trata is two buysiders hopping on a call and talking through a stock they are actually working on, which is the fastest way I know to get up to speed on a name. They also have an MCP now that connects to Claude and ChatGPT, so the first thing I do on a new name is run the Trata search and see what buysiders are really saying. That link is a preview of the Galaxy call I mention on the episode.
Chapters:
(00:00) Cold open: three things on my mind
(02:11) Sponsor: Trata
(03:55) Why I am recording a bonus ramble
(04:49) Strategy's new 8K: sell stock, sell bitcoin, buy back the preferred
(06:05) "Multi-way capital issuer" is just normal capital allocation
(06:58) A great call in 2020, and then they never sold
(08:21) Investors are good at buying and bad at selling
(09:37) Is my own sell plan its own trap?
(11:37) The Situational Awareness blowup
(13:19) The trouble with thematics: there is no price
(14:44) Micro versus macro, and the software buy signal I missed
(16:22) Do the crossover funds have an AI information edge?
(18:41) Why I cannot get to the memory valuations
(19:04) The opportunity in the beaten-up AI power names
(22:05) Wrapping up
Links:
Yet Another Value Blog - https://www.yetanothervalueblog.com
See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer
Production and editing by The Podcast Consultant - https://thepodcastconsultant.com/
Disclosures: I am short $MSTR and long $GLXY. Nothing on this podcast is investment advice. Please do your own work.

Management interviews might be the most underdeveloped skill in fundamental investing. Ross O'Toole has been investing for 25 years and read over 500 investment books, and he couldn't find a single one on how to actually conduct an investor-management interview.... so he wrote Breaking the Script, a field guide to getting management teams off their rehearsed talking points. We get into why "what" questions beat "why" questions, whether you should grade a plastic surgery CEO and a coal company CEO on the same curve, the case for recording your management meetings, and why asking for examples is one of the biggest double edged swords in investing.
I push back with my standing worry: management teams are really, really good salesmen, and I always walk out of these meetings wondering if I'm the patsy at the poker table. Ross's answers (build a longitudinal baseline over repeat interviews, ask for the negative example every time you get a positive one, and save the hard questions for the crescendo) are why this book went straight to the top of my "hand it to an intern" list.
Grab Ross's book, Breaking the Script: https://amzn.to/4fLg8RO
This episode is sponsored by Fiscal.ai: https://fiscal.ai/yav. Fiscal.ai is a modern financial data platform for global equities. In addition to their web-based terminal, they offer API access to real-time fundamental data: 20+ years of financial statements, ratios, segments, and KPIs, with data updating within minutes of earnings reports, not days. I'm not just an advertiser; I signed up with my own money to plug their API into the AI tools I've been building. Use my link, fiscal.ai/yav, for 15% off.
Chapters:
(00:00:00) Introduction
(00:02:00) Why management interviews are a double edged sword
(00:03:47) Why Ross wrote Breaking the Script
(00:07:15) Are we deluding ourselves? Testing management credibility
(00:11:53) Should you grade CEOs on a curve across industries?
(00:14:44) "What" vs "why": framing contentious questions
(00:18:27) Are interviews actually an alpha source?
(00:19:39) Where management matters most: deep value vs tech
(00:22:32) Would interviewing 2008 Zuckerberg have helped?
(00:25:14) Preparation and repeat interviews
(00:29:20) Should you record management meetings?
(00:31:54) Asking for examples: conviction builder or sales pitch?
(00:34:00) Always ask for the negative example
(00:36:00) Making management grade their own execution
(00:38:20) Interviewing companies under activist pressure
(00:41:30) Buffett's silver bullet question, reframed
(00:46:02) Question order: crescendo to the hard stuff
(00:48:29) Closing thoughts
Ross O'Toole / Breaking the Script: https://amzn.to/4fLg8RO
Links:
Yet Another Value Blog - https://www.yetanothervalueblog.com
See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer
Production and editing by The Podcast Consultant - https://thepodcastconsultant.com/