https://www.youtube.com/watch?v=rjyoDSjqv8Y
Daniel Mahncke pitches Palantir to his co-host Shawn O'Malley, walks through how the ontology layer actually works, why the company looked like a failed consultancy until 2023, what changed with AIP, and what two self-described value investors do with a business growing 90% at a 60% margin. They end up not buying it, and spend the last twenty minutes explaining why that is a judgment about their own understanding rather than about the company.
🎙️ Hosts: Daniel Mahncke and Shawn O'Malley, analysts at The Investor's Podcast Network
📰 Published: 30 August 2026
🔴 YouTube | 🟢 Spotify | 🟣 Apple Podcasts | 🔗 Show notes | ⏱️ 76 min | ✅ Time saved: 47 min
Key Takeaways
The pitch is a reversal, and Mahncke says so in the first sentence
"I saw Palantir as the poster child of overvalued companies for a long time."
"And last year, Palantir traded at twice the multiple with half the growth. And that has completely shifted now."
Palantir's rule-of-40 score is 155%, against the 40 that marks a healthy software company
"Now for Palantir that number is not 40%, it's 155% — 65% in profit margins and 90% topline growth."
The growth came from existing customers spending more, not from new logos
Net dollar retention went from about 100% in 2023 to almost 160% last quarter
Customer-count growth halved over the same stretch
The slow, expensive onboarding that nearly killed the company is now the moat
Two years to onboard the first client; engineers living in the customer's offices three to four days a week
"before AI Palantir was even seen as a fake SaaS company"
Middle management, not technology, is what blocks a Palantir deployment
"many ex employees described internal politics as one of the toughest hurdles"
Microsoft is building an ontology layer and Mahncke is unimpressed
On Fabric IQ: "I would call it an attempt at best."
ServiceNow is the one competitor he says he cannot rule out
Palantir has no model of its own, and that is deliberate
It resells Claude, ChatGPT, Gemini, Mistral and Llama, so the pricing power stays with Palantir rather than with a lab
Karp has told investors he will grow the whole business at the US commercial rate for 18 months
That implies about 150% growth, and even 100% would mean 17 or $18 billion of revenue at the end of 2027
On Karp's own numbers the multiple collapses from about 60 times sales to about 20 by the end of 2027
The analyst case, at a 22% revenue CAGR, produces a fair value of $90
$9 billion of cash, no debt, no buyback, and stock comp running at 13% of revenue
The repurchase program was terminated in January and Mahncke thinks that was right at triple-digit sales multiples
The capital-allocation decision neither host follows is the Accenture handoff
Accenture consultants are being trained to do the forward-deployed work that built the moat
The reason they pass is a test about themselves, not about Palantir
If growth halved next quarter, neither could explain why — the same problem Mahncke has with The Trade Desk
"And if you see the stock below $100 and nothing changes, count me in."
Palantir Sells a Live Map of an Organization, Not a Database
Mahncke opened by conceding the position he was arguing against. "I saw Palantir as the poster child of overvalued companies for a long time." he said, adding that Michael Burry had recently said the stock is not worth a single dollar. "But after looking into it, I actually believe Palantir has never been more attractive than today." O'Malley set the terms immediately: "Well, to be clear, it's far from a value play."
The case for the change is a change in the facts. "And last year, Palantir traded at twice the multiple with half the growth. And that has completely shifted now."
Asked what the company actually does, Mahncke used an airline. A plane breaks down at Frankfurt airport. Five separate systems hold five pieces of the answer — which part failed and how long the repair takes, how many duty hours the crew has left, who is on board, which of those passengers have connecting flights, and what the replacement seats are worth.
The state of the art is worse than a listener would guess: "So you basically have five different systems and teams that need to communicate and coordinate. And the way this is actually still getting resolved today in 2026 is by just phone calling each other."
Palantir sells a software layer that sits on top of all five and connects what it calls things — the plane, the part, the crew, the passenger, the connection
What separates it from a database is permissions and actions: "what Palantir builds is not only facts", but the relationships between them, the fact that rebooking is an action that exists, who is allowed to approve it, and what else has to happen when it does
That combination is what Palantir calls the ontology, and it is the word the rest of the episode turns on
O'Malley's read on it was cinematic: "it kind of reminds me of something that would be part of a Hollywood movie, like some action movie where the lead roles are out there in the field fighting the bad guys and then there's like this control center that always knows exactly what's going on and is like feeding them advice."
The airport example was not hypothetical. Mahncke flies to New York in two weeks for the show's own conference, and said a breakdown at Frankfurt is "traditionally how it works with my travel." O'Malley reminded him of the four canceled flights on the way to Omaha for the Berkshire Hathaway meeting last time.
PayPal's Fraud Problem and 9/11 Are the Company's Two Origin Stories
Palantir was founded in 2003 by Peter Thiel, Alex Karp — still the CEO — Joe Lonsdale, Steven Cohen and Nathan Gettings. Mahncke traced the idea to PayPal, a company he noted the show lost money on: Thiel's answer to payments fraud was software that surfaced suspicious patterns across a lot of data, with a human making the final judgment call.
Then 9/11. US agencies had the information needed to stop the attack, but it sat in separate agencies that could not bring it together in time — the same ontology problem, applied to intelligence.
Why Thiel never took the CEO job, as Mahncke reads it: he chose Karp early and took a fundraising and advisory role, while keeping the largest individual stake — "he also owns the largest individual stake in Palantir at about 4%", with Karp at two and a half percent
On both men's reputation: "I don't know as well as you what Peter Thiel's reputation is in the US, but I can tell you that it's not that great here in Germany." Mahncke added that Karp is on his way to an equally bad one there, and that Karp lived and studied in Germany for over a decade
O'Malley's own view of Thiel was blunter: "he does give off like super villain vibes, which doesn't pair very well at all for a company that is really generally working to increase government and corporate surveillance in some sense", and he reached for the Empire from Star Wars. He also said Thiel is hugely respected in Silicon Valley and that Zero to One is one of his favorite investing books
Thiel bankrolled an initial cost of $30 million. In-Q-Tel, the CIA's venture arm, put in $2 million — trivial against a company now worth, depending on the day, between $350 and $400 billion.
The money was not the point: "But it was less about the $2 million and more about the access, that you get to the government sector if you're sort of sponsored by the venture arm of the CIA."
For a good two years, Palantir built its only product for that one client
O'Malley stopped him there — "You said it casually there, but there's a venture arm to the CIA." — and said he would be researching it after the episode
The Onboarding Was So Slow the Company Looked Like a Fake SaaS Business
Two years to onboard a customer was not an anomaly. "the ontology layer is basically a digital twin of a customer's entire organization", Mahncke said, so the work is understanding every detail of a corporation before any software is useful.
Palantir ran two kinds of engineers: product engineers who built the software, and forward deployed engineers who flew to the customer and worked out of their offices at least three to four days a week. Deployments no longer take two years, but they still take months.
What that looked like from outside: "before AI Palantir was even seen as a fake SaaS company", because it read as enterprise consulting with some software attached
O'Malley compared it to CoStar, a holding in the show's portfolio, which has spent decades sending people to photograph and map commercial properties: "they have armies of people that are creating this basically unique data that nobody else has because they're going out and physically documenting all the details of these properties" — slow, inefficient, and a moat because nobody else will do it
Mahncke accepted the analogy and raised it: "obviously what Palantir does is 100 times more difficult"
The data is rarely ready. Much of it is only half digitized — handwritten documents scanned to a PDF, digitized but, as he put it, you cannot just search it up
The bigger obstacle is human: "many ex employees described internal politics as one of the toughest hurdles"
AIP and the Free Boot Camps Turned a 12% Grower Into a 90% One
O'Malley put the growth chart to him: revenue growth decelerated from 2019 to about halfway through 2023, when "year-over-year revenue growth was as low as 12%". "But then from there on it's accelerated dramatically like a rocket ship and we're talking about 100% year-over-year growth."
The explanation runs through Palantir's four platforms. Gotham came first, in 2003, built for intelligence work and the problems that led to 9/11, and it creates the digital twins for defense agencies. Foundry arrived in 2016 and does the same for corporate operations. Both sit on Apollo, a deployment engine that ships updates into siloed databases where a one-click update is not possible. And then AIP, the AI platform, which Mahncke called the real game changer.
The sales motion changed with it. O'Malley described the boot camps: CEOs and CIOs invited in with some of their own data — not the sensitive kind — for a three-to-five-day workshop in which Palantir shows them what it could fix
"And it's important to say that Palantir paid for those workshops out of their own pockets." Attendees took no risk beyond handing over data
"I think I heard that they had over a thousand workshops already", each with a CEO or CIO in the room
The workshops were an answer to a scaling problem the company hit in 2023, when befriending the C-suite of every prospect one at a time stopped working
The Growth Is Existing Customers Spending More, and It Came Free
Customer count is not where the growth is. "customer count growth has actually come down from about 50% in 2024 to about half of that lately", Mahncke said. What moved was spending per customer.
Net dollar retention is the metric: "that metric increased from a low of about 100% in 2023, which basically means they spent as much this year as they have last year to almost 160% in the last quarter. So for every dollar spent last year, a customer now spends $1.60."
And none of it cost anything at the margin: "Incredible thing with Palantir is that all this growth, which is pretty much unheard of for a company of this size, did not come with any margin compression." No major marketing spend, and profit margins expanded alongside the top line
O'Malley put the before and after together: in 2023 "the company was barely breaking even with a net profit margin of just about 3 to 4%", and "Today though, net profit margins are 60% or almost there."
The rule of 40 is the cleanest way to see it. Add revenue growth to profit margin; 40 is the mark of a healthy software company. "Now for Palantir that number is not 40%, it's 155% — 65% in profit margins and 90% topline growth."
The deal sizes explain why the customer count looks unimpressive. Palantir only chases large customers with upsell potential, which is why a base of about 1,000 customers and 42 new ones in a quarter understates the business
Mahncke's summary of the mechanism: net dollar retention is high "and why Palantir is growing like a weed"
The Multiple Is Absurd on Trailing Numbers, Which Is the Whole Argument
O'Malley made the bull case for the valuation before the valuation section. The multiples look absurd because they are struck on trailing earnings and ignore the growth ahead — "you had the stock at times trading at like a 100 times sales, which again sounds ridiculous" — and with a large addressable market to roll into, he said it starts to feel a lot less crazy, emphasis on trailing.
Mahncke previewed his conclusion: "Palantir is actually as attractively valued as ever.", "Even though the multiples look kind of scary at first". He also admitted the timing hurt. He first looked at the stock a couple of weeks earlier, thought it looked like a good opportunity, "then earnings came up and suddenly the stock went up 40%. And I got to say that was a bummer."
US Commercial Is Carrying Everything, and Europe Is Not
Palantir does not disclose which platform earns what. The splits investors get are commercial against government, and a geographic breakdown.
Government and commercial growth run roughly in line, which Mahncke likes because it reduces dependence on government contracts, a field he sees as more volatile depending on who is in the White House
Commercial has outpaced government recently; he expects that to hold long term, given a larger universe of customers and much-improved onboarding
Earnings before interest and taxes are similar for both segments, so neither is meaningfully more profitable
The geographic gap is the striking one. O'Malley: US growth has more than nine-xed since 2023, "And it's up 115% year-over-year this past quarter", while "While international growth has definitely been much more volatile, and it's currently sitting at 30%" — impressive anywhere else, unimpressive here
On why Europe lags, Mahncke gave three reasons rather than one: European resistance to depending on American companies for data management of this kind, slower adoption of anything new, and the on-the-ground demands of Palantir's own model
Karp's 18-month promise is the number the model runs on. On the last earnings call, "Alex Karp said that he will grow the business at a rate equal or above what the US commercial business is doing for the next 18 months."
"that would imply a growth rate of about 150%."
"That would take Palantir to 17 or $18 billion in revenue at the end of 2027, which is insane." — and that is on 100% growth, not 150%
Why Nobody Has Copied the Ontology
O'Malley pushed the central objection. "Why is there no one that seems to be able to copy what Palantir is doing?" OpenAI, Anthropic, Salesforce, ServiceNow, Microsoft — all have the distribution advantage the show has called decisive in the age of AI, and none of them is going to take this lying down. "I've seen estimates that Palantir could have as much as a $2 trillion addressable market."
Mahncke called this the most important point of the episode and the hardest to judge from outside, because "a lot of Palantir's advantage seems to come down to execution".
The switching cost is the deployment itself. After a multi-month or year-long ontology build, "I think you will think twice about switching and doing it all over again."
First-mover advantage is still intact: "There's no competition that has come up even remotely in the last two years." — and the client base it locked in first is the biggest corporations and governments, on multi-year contracts
Culture is part of the answer he cannot verify. "this company has a very unique culture" with little visible hierarchy, where people work on what they think creates the most value
The product strategy runs against industry practice. Enterprises used to buy best-in-class point tools — Slack over Teams, Excel over Google Sheets, Zoom over Google Meet, Salesforce for the database — which is lower value per transaction and much higher volume. Palantir sells the opposite shape
Microsoft's Fabric IQ Is "an Attempt at Best," and ServiceNow Is the One He Cannot Rule Out
O'Malley asked why Microsoft or Google could not level up their existing bundle, connect it with AI, and offer their own ontology.
"I think it's safe to say that Palantir has not yet been tested by competition", Mahncke said — but going from good-enough tools to a best-in-class one, and one this comprehensive, is easier said than done
Palantir is not taking their market either, so this is not a defensive investment for anyone
He does not think the incumbents can move: "I think they're way too bloated and just not flexible enough to compete with a company like Palantir on that ontology front."
On Microsoft's own ontology layer, Fabric IQ: "I would call it an attempt at best." He sees it as a sensible way for Microsoft to connect its own products, and a long way from what Palantir builds for its top customers
ServiceNow is the exception he flags against himself — the one company here he has not deep-dived, and the one with the most to lose from sitting out a $2 trillion market
Could a Company Just Build Its Own Ontology With AI Tools?
O'Malley's next challenge was that AI has democratized the capability: a company could use Claude, Codex and similar tools to rebuild the relationships inside its own organization at a fraction of the cost. "And actually this time good enough would actually be good enough" — because a company building for itself knows exactly what standard it needs, where a vendor's idea of good enough is a guess.
Mahncke's answer was that the demonstration effect is real but the execution is not.
Palantir showed companies how inefficient their data management actually was; before that it felt unchangeable. Cost savings are among the loudest things its customers report
"And all that said, I think in-house solutions are to some extent wishful thinking."
The reason is the same politics as before, pointed the other way. Palantir wins by befriending the C-suite first and the project workers second, and routing around middle management — but an in-house build has to be deployed by that same middle management
O'Malley granted the point about incentives and pressed on scale: "you can't befriend the C-suite of every single customer you have if you want to be a trillion dollar company", and something may be lost in translation as the playbook stretches to more kinds of enterprise
Palantir Has No Model of Its Own, and That Is the Point
Asked which large language model Palantir runs on, Mahncke was categorical. "Palantir does not have its own model."
"So Palantir itself offers Claude, ChatGPT but also Gemini, Mistral, Llama and basically whatever model is out there." — the customer picks
"The magic of Palantir really comes from the context that the ontology creates." The ontology feeds the model context, data governance rules and guardrails, which is what stops it hallucinating and makes it useful for the specific job
To make the point about context, the show played a clip from an AI masterclass recorded by a member of its mastermind community, explaining that an agent is a model plus a harness — the access it is given to the data sitting in your systems — and that Palantir's ontology is what supplies that harness
Why the labs would struggle to sell the same thing: if OpenAI or Anthropic built the layer, "you're limited to just one model instead of being model agnostic", and all the pricing power moves to the lab. With Palantir, an expensive Anthropic model can be swapped for ChatGPT, and an expensive ChatGPT for Mistral
On Karp's warnings about the labs: Karp is frequently shooting against Anthropic and OpenAI and telling corporations not to hand over their IP, because the frontier models will eventually take over their business. Mahncke noted Karp has an obvious commercial incentive to stoke that fear, and then said it does not make him wrong
The Frankfurt School, and Why the Company Believes It Has a Mission
O'Malley asked whether Karp and Thiel can be trusted with a customer's data. Mahncke said he does not know, but drew a distinction: "Palantir's customers keep all the rights to their data of course", which is different from feeding a chatbot that owns what it is given, even though Palantir gains deep insight into how the business runs.
Karp has been CEO for 23 years. Steven Cohen, who built the first prototype as a student, is still on the board, as is Thiel; the founding team is involved but not as actively as Karp
Karp lived in Germany for about a decade and was drawn to the Frankfurt School — a philosophy analyzing how modern capitalism, mass culture and fascism shape human psychology and social control
The thesis Mahncke says explains the company: "Karp is thinking that American thinking, whatever that means, cures nations of fascism and that it's a sort of moral obligation for Silicon Valley to return to its roots." Those roots being the US defense system
O'Malley: "It seems like a very ideologically driven company, which is sort of weird." It does explain doing business in Europe while saying the growth is bad — a mission rather than an economic calculation
Mahncke would not go that far: "I doubt that altruism is the primary motive for making business there." Palantir makes real money in Europe
Why any of this belongs in an investment case: the worldview matters when the company is taking some of the most sensitive defense contracts in the world.
ImmigrationOS, France, Germany and the Mayor of London
Mahncke thinks the reputational risk is underpriced relative to how investors treat other controversial names. Few investors care about Palantir's reputation outside the investment world, while people are afraid to own Meta over public perception — and "a company that is building an app called ImmigrationOS for ICE definitely comes with much more reputational and regulatory risks if you ask me".
"And in several countries, courts have actually ruled that the way Palantir software connects and automates data is not in line with law." He expects more problems internationally and eventually in the US
O'Malley split it into two questions: whether owning it fits your values, and whether the reputation actually damages cash flows — the same structure as thinking about Tesla and Elon Musk's political impact
The lost business so far: "So, as you hinted at, France's internal security directorate replaced Palantir. Germany's domestic intelligence service appears to be likely to do the same. You had the mayor of London blocking a police contract that would have been worth 50 million pounds."
Mahncke is skeptical that Germany and France can build a real alternative given their own data laws, and would be surprised not to see more European government deals eventually
A Balance Sheet With No Debt, No Buyback, and Stock Comp at 13% of Revenue
Asked about capital allocation, Mahncke described a balance sheet with nothing on it. "They hold more than $9 billion in cash, and there's no debt on the balance sheet. Not long-term, not short-term."
No dividends, no M&A, and "they terminated the share repurchase program that they had in place in January of this year"
He thinks stopping was right: "I would rather have the opportunity for more opportunistic buybacks in the future instead of buying back stock when it's trading at triple-digit sales multiples as it has in the past"
The cost of that choice is dilution. O'Malley: "The problem of having no buybacks is that stock-based compensation is running at 13% of revenue." The ratio is improving mostly because revenue is growing, not because the grants are shrinking — "Two years ago, the stock-based comp was as high as 20% of revenue."
Mahncke's clarification, unprompted: the bullishness is relative to last year, not absolute. A year or two ago he considered Palantir one of the most expensive companies in the market, and says he was right to
The Accenture Handoff Is the Decision Neither Host Can Follow
"But what I asked myself in terms of capital allocation is why does Palantir not invest more money in their workforce?" Demand exceeds capacity and revenue per user keeps rising, so hiring looks like the obvious use of the cash.
Instead, "it partners with the consultancy firm Accenture and basically they use Accenture employees as forward deployed engineers". Accenture trains its own consultants on Foundry and AIP, those consultants do the deployment work at the customer, and Palantir sells the license.
"On paper, it makes sense because they keep the high margin revenue" — but the on-the-ground work is exactly what Mahncke has spent the episode arguing is the moat
The scale of the substitution is not small: "Palantir has about 4,500 employees. Accenture has close to 800,000 and they already have more than a thousand foundry specialists."
O'Malley's reaction was to the size of Accenture rather than the strategy — he said he cannot fathom overseeing 800,000 people, and that it explains why both sides would want the arrangement
Two Base Cases, and a Multiple That Falls From About 60 Times Sales to About 20
Mahncke built two five-year models, one on Karp's own trajectory and one on analyst estimates.
The starting point is the multiple compressing on its own. If Karp is right that the whole business grows at the US rate through 2027, "well then the price to sales ratio would decline from where it is today, which is about 60 to about 20." Everything after that depends on 2028 and 2029
The comparison that does the work is on operating profit, not sales. "when the average enterprise software company is trading at about 7 time sales on roughly a 20% operating margin that's a 35x on operating profit", and Palantir on assumed 2027 numbers at about 22 times sales with a 60% margin lands at the same 35 times operating profit
The first case uses Karp's roughly 150% growth for next year, then steps growth down by 15 to 20 percentage points a year, ending at 28% in 2031 — a revenue CAGR in the high 50s
The second case is the sell side's: "the second base case built up on analyst estimates works with a revenue CAGR of only 22%"
Both hold operating margins at about 60%. He notes competition from Anthropic or Microsoft could compress that, and calls the possibility completely speculative today
On the analyst case, with a 30x exit multiple, a 10% discount rate and the usual 20% margin of safety, "you would get a fair value of $90." Karp's case, with growth about 10 percentage points a year higher afterward, produces a much higher number — "So, a significant difference and way more than the stock is currently trading at."
Dilution is modeled lightly. O'Malley noticed the model assumes 1% annual share-count growth against double-digit stock comp, with no buybacks or dividends
The model is downloadable and the assumptions are adjustable, with a companion newsletter published the same day
The Test That Kept Them Out: What If Growth Came in at Half?
O'Malley set up the ending by admitting his prior — "I would have dismissed Palantir as being a hopelessly overvalued meme stock" — and then assumed the answer was still no.
It was. Mahncke said he was closer to buying before earnings, and then applied the test he uses on fast-growing companies where everything looks good on the surface: assume the next report cuts growth in half.
"So let's assume Karp's outcome 150% growth. Now suddenly you only get 75%." Then ask: "Would I think I'm confident enough to be able to figure out and explain why the company grew significantly slower than people anticipated?"
The comparison set is what makes the test bite. With Lululemon, a slowdown is straightforward — they sell less clothing, because of a competitor or a recession. With payment companies it is harder but doable. "And then there's a company like The Trade Desk where I have absolutely no clue why the topline growth is declining."
"I think when the business is doing great and it's growing, everyone has this illusion of knowing and understanding why it's growing so great." The illusion only breaks when the number does
"I think the two of us would probably be better off sitting on the sidelines just admiring what Palantir is doing and not investing today." He was explicit that someone working in the sector might reasonably reach the opposite conclusion
O'Malley agreed and generalized it: "Surprise, surprise. We, the value investors here, two guys who are very inspired by Buffett and Munger didn't want to invest in Palantir." His gut is that the company and probably the stock will do well, and that a sell-off would be an interesting entry for anyone willing to speculate
It goes on the watch list rather than the never list. Mahncke said talks with mastermind members who understand the business better could get him there, and that price matters as much as understanding
He closed on Buffett, quoting: "The line separating investment and speculation, which is never bright and clear, becomes blurred further when most market participants have recently enjoyed triumphs." A lot of people have made a lot of money in Palantir — "Again, it's up 2400% from the 2023 lows." — and he takes it as natural that some of them are comfortable with blind spots in the thesis
Palantir's growth and margins are real enough that the multiple is defensible on Karp's numbers, and the reason these two are not buying has nothing to do with the price: they cannot say what they would look at if the growth stopped.
Products, Companies & Tools Mentioned
Palantir (The subject. Four platforms — Gotham for defense digital twins, Foundry for corporate operations, Apollo as the deployment engine, AIP as the AI layer that changed the sales motion — and the ontology that connects a customer's data, permissions and actions)
Accenture (Trains its own consultants on Foundry and AIP and supplies them as forward deployed engineers, with more than a thousand Foundry specialists against Palantir's roughly 4,500 total employees)
Microsoft (Building its own ontology layer, Fabric IQ, which Mahncke calls an attempt at best and a way for Microsoft to connect its own products rather than a rival to Palantir)
ServiceNow (The one competitor he will not rule out — strong at workflows on predefined relationship data, and with the most to lose from sitting out the addressable market)
Anthropic and OpenAI (Model suppliers inside Palantir rather than competitors today; if either built its own ontology layer the customer would lose model-agnosticism and the pricing power would move to the lab)
Claude, ChatGPT, Gemini, Mistral and Llama (The models Palantir offers, with the choice left to the customer)
Salesforce (The contrast on deal size — five- and six-figure enterprise deals against Palantir's million-dollar-plus ones — and O'Malley's reference point for what Palantir is, an AI-native version of it)
Google and Alphabet (The bundle strategy of good-enough tools sold into an existing ecosystem; also the largest position in the show's portfolio, and the company Karp cites for walking away from a Pentagon contract in 2018)
In-Q-Tel (The CIA's venture arm, which put $2 million into Palantir and, more importantly, opened the intelligence agencies to it)
PayPal (Thiel's fraud problem was the prototype for the idea — software surfacing patterns, a human making the call — and a stock the show lost money on)
CoStar (A portfolio holding whose decades of physically documenting commercial properties is O'Malley's analogy for Palantir's boots-on-the-ground moat)
Uber (O'Malley's benchmark for the biggest swing in margin profitability he has seen from revenue scaling)
The Trade Desk (The company whose slowing revenue Mahncke says he cannot explain, and the category he reluctantly puts Palantir in)
Lululemon (The counter-example — a slowdown there is easy to diagnose)
Meta and Tesla (The reputational-risk comparisons: investors avoid Meta over public perception, and Tesla shows how a founder's politics can reach the cash flows)
Slack, Microsoft Teams, Excel, Google Sheets, Zoom and Google Meet (The best-in-class point tools enterprises used to assemble one by one, which is the buying pattern Palantir inverts)
Fiscal.ai (The show's data terminal, named as its sponsor)
Books & Resources Mentioned
Zero to One – Peter Thiel (One of O'Malley's favorite investing books, and in Mahncke's reading a guide to the philosophy behind founding Palantir)
The Frankfurt School (The critical theory Karp studied in Germany, analyzing how capitalism, mass culture and fascism shape psychology and social control — Mahncke's route into why Palantir acts as it does)
Satish Terala's AI Masterclass (The mastermind member's talk the show cut into the episode, on what an agent harness is and how an ontology supplies one)
Future Investing's interview with a Palantir forward deployed engineer (Linked from the episode notes, alongside the ex-employee accounts Mahncke drew on for the onboarding and internal-politics material)
The Intrinsic Value Portfolio (The portfolio the pitch was being made for, tracked publicly)
The Intrinsic Value Newsletter (Carries the Palantir write-up published the same day, with the downloadable model and its assumptions)
Previous intrinsic value breakdowns – Constellation Software, Dell, Alphabet and the Constellation spinoffs (The show's earlier company episodes, linked from the notes)
Get the latest market chatter as it happens: