Too late to buy Palantir stock (PLTR)? The math behind $38 billion

0:00 What has to happen for that price to pay off?

Palantir is worth four hundred and forty-seven billion dollars. That's more than three companies that build America's weapons: Lockheed, Northrop and General Dynamics. Combined. Those three outsell Palantir twenty-eight to one. So the price counts on something that hasn't happened yet. What has to happen for that price to pay off? One way to measure a price: years of profit. An S&P 500 index fund costs twenty-six years. Microsoft, twenty-nine.

Palantir? Hold that thought. By the end, you'll know what has to be true for you to make money on it. Three stops: what the price counts on, whether the boss can deliver it, and where the growth has to come from. Cartoon, yeah. But I searched Palantir's reports to the SEC. All twenty-four. And read two court rulings. And its federal contracts: I counted four hundred and thirty-nine.

Plus six earnings calls. I checked thirteen quotes against the audio. Read it at your pace: the links are on shortcircuitcreations.com.

Palantir joins the Hype Fund when this video ends. That's my experiment: the stocks I roast, against the S&P 500.

1:13 Burry's bet against it

On the other side: Michael Burry, from The Big Short. In April, he was betting the price falls, with puts that run into 2027. A put pays off if the stock falls under a set price. His was fifty dollars. Today it's a hundred and eighty-six. So Burry wins if Palantir loses about three quarters. If Burry is right, the Hype Fund takes the hit.

Then again, Burry has predicted seven of the last two crashes. Ha ha ha. He got both.

1:43 What you'd be buying

Before the price: what you'd be buying. Palantir doesn't collect data. Its customers already have it, in systems that don't talk to each other. So Palantir sends its own engineers. They connect the systems, so one program can reach every database. On top, Palantir's software pulls it all together and looks for patterns.

Since 2023, there's AI on top of that: AIP. You type a question, and it answers from all of it. People or software: which one does the work? Hold that thought. That's stop two. The customers: armies, spy agencies, hospitals, shipyards. And ICE. Its CEO, Alex Karp, has a goal: ten times the sales in America. In his view, within five years.

From three point three billion dollars in America last year to thirty-three billion. Keep that number. You'll need it in stop one.

2:38 Stop one: what the price counts on

Stop one: what the price counts on. Those years of profit have a name: P/E. The price divided by one year of profit. Palantir today: a hundred and forty-eight years. Almost six times the index fund. Hope you're patient. And that's with almost no tax. On its US profit, one point six billion dollars before tax in 2025, federal income tax: zero.

It's legal. Deductions for paying staff in stock, and old losses. Paying tax in cash comes later. The old losses run out around 2027, as profit grows. Stock deductions can keep it lower for years, if the stock stays high. With a normal tax, a hundred and forty-eight becomes about a hundred and eighty-four. So the price is higher than it looks. From here, I count with normal tax.

One quarter will look cheap. Palantir writes there's a 'reasonable possibility' it will book those saved-up deductions as profit. The quarter it does, call it the tax day, profit jumps once. On paper. No new cash comes in. ServiceNow, Workday, Palo Alto and Tesla did the same. Profit jumped. None of the four stocks went up.

So if you see record profit that quarter, check for the tax day before you count it. To see what the price pays for, compare a company that sells about as much. Snowflake, another data software company, sells almost as much. The stock market values Palantir at about four times Snowflake. Palantir grew ninety-three percent last quarter, Snowflake thirty-five. Palantir makes money. Snowflake loses it.

So the price pays for growth.

4:17 The price is profit times the P/E

Turn the P/E around, and the price is one year of profit times the P/E. Today, that's four hundred and forty-seven billion dollars. Split over its 2.4 billion shares, that's a hundred and eighty-six dollars a share. And about a dollar of profit. A P/E this high comes down when the profit grows into the price. Or when the price falls. Or both. How much growth? Remember Karp's goal.

At Karp's goal, US sales are thirty-three billion. Add the rest of the world, growing about a third a year like now: about thirty-eight billion. Keep today's profit per dollar of sales, with normal tax: about fifteen billion dollars. Keep today's price, and divide it by that profit. That fills the empty frame: a P/E of about thirty.

Palantir pays its staff in new shares, so the profit is split more ways: about fourteen billion for today's shares. Counting that, about thirty-two. About what Microsoft costs today. A low P/E is good when you buy. In 2030, your share is worth what the next buyer would pay. Per share, the profit at Karp's goal is about five dollars eighty.

So in 2030, at thirty-two years of profit, like Microsoft today, a share is still a hundred and eighty-six dollars. If you bought today, you'd pay for Karp's goal up front. Four years, for nothing. Since 1928, the S&P 500 has averaged about ten percent a year. At that pace, a thousand dollars in an index fund today is about fifteen hundred in 2030.

At thirty-two, hitting the goal only gets your money back. Anything you make comes from beyond the goal: more profit than it brings, or growth that keeps going after 2030. In 2000, at the top of the dot-com bubble, Cisco cost over two hundred years of profit. The stock fell first. Then its profit per share grew seven to eight times. The stock took almost twenty-six years to get back.

By then, it cost about thirty-one years of profit. Nvidia, in 2023, also cost over two hundred years of profit. Then its profit per share grew forty-one times, far more than the price counted on. The stock: five point six times. Even there, the P/E fell, to twenty-seven. Two famous cases, both picked after the fact.

The other way: growth that keeps going after 2030. To keep up with an index fund, a share has to be worth about two hundred and eighty dollars.

With the profit at Karp's goal, the next buyer has to pay a P/E of about forty-eight. So in 2030, buyers still have to see years of fast growth ahead. Do fast growers get that price? Two that grow fast today. Fortinet, a firewall maker, grew twenty-six percent last quarter. Its P/E is fifty-nine. AppLovin, an app ad company, grew fifty-three percent. Its P/E is twenty.

So growth alone doesn't set the price. At AppLovin's twenty, a share in 2030 is worth about a hundred and seventeen dollars. And at Fortinet's fifty-nine, about three hundred and forty-five. For you, the first thing that has to be true: in 2030, the profit per share times the P/E beats about two hundred and eighty dollars.

At Karp's goal, that takes a P/E of forty-eight. Thirty-two only gives you your hundred and eighty-six back. Fall short on one, and you need more of the other. Can Karp deliver it? That's stop two.

7:46 Stop two: can Karp deliver it?

Stop two: can Karp deliver it? First, has he done this before? In 2022, Karp aimed for four and a half billion dollars in sales in 2025. He got within twenty-five million. And beat his own forecast eighteen quarters in a row. So his goals have held up. That goal was about two point four times in three years. This one is ten times in five.

So his record is a good sign. It isn't proof. One check: contracts customers can't cancel, for work Palantir hasn't delivered yet. Together they're worth about ten months of its sales. Is that a lot? Three big software companies whose contracts can't be cancelled have eighteen to thirty-two months. Three that sell by the year, or let customers cancel, have about ten. Like Palantir.

Palantir lets many customers cancel on less than a year's notice. Those contracts don't count here. And the US government can't lock in a contract's later years more than a year ahead. So ten months is low, but normal for contracts like these. And the biggest customers stay anyway. Palantir's top three have been with it fifteen years on average.

But everything locked in today adds up to about an eighth of one year's sales in 2030, at the goal. Remember, if you bought today, you'd pay for Karp's goal up front. Seven eighths of it isn't locked in yet. Last time, it was about twice that. The goal was smaller, and a year closer. He came within twenty-five million anyway. So signed contracts don't settle it.

9:18 People or software?

Burry, who's betting against it, has his own answer: 'a consultant riding a bubble of AI FOMO demand.' Remember the engineers Palantir sends on site? That's what Burry calls consulting. Connecting the systems is work people do, and Palantir bills for it. Of what the US government bought in 2025, about a quarter was paid as hours. So Burry has a point, for about a quarter of it.

A consultant sells hours, so more sales need more people. Software is written once and sold many times. So that's the question from the start: people or software. In August 2025, Karp put a headcount on that goal. He told CNBC: 'The goal is to get 10x revenue and have 3,600 people.'

In 2030, at the goal, that's ten point six million dollars of sales per person. Today, one point four. So ten times with fewer people only works if Palantir really is software. He called it 'a crazy, efficient revolution.' Five months later: four thousand four hundred and twenty-nine people. More sales still took more people, but far fewer per dollar. Last year, sales grew fifty-six percent. Staff, thirteen.

In June: four thousand four hundred and one, still about eight hundred above his goal. But sales kept climbing. In 2022, sales per person were half a million dollars. Today, one point four: nearly three times. So the direction is right. To reach ten point six million, it has to rise almost eight times again.

Booz Allen, a government consultant, sells almost twice as much. It keeps seven cents of profit per dollar. Palantir, today: thirty-nine. At seven cents, Palantir's P/E at the goal would be about a hundred and eighty. So the price only holds if Palantir keeps its software margins. Another way to more sales per person: bigger bills.

Where Palantir is almost alone, in the Pentagon's targeting software, it has won every head-to-head contest on record. Its twenty biggest customers now pay a hundred and twenty-four million dollars a year each. A year ago, seventy-five. So more sales per person can also come from bigger bills. Handy, when you're almost alone. Good for profit. It just doesn't prove the software does the work.

In court this year, the ruling quotes Palantir's own witness: a software company, and a consulting company. On profit per dollar, the numbers side with software. Burry's other point is about bills already sent: money customers still owe Palantir. From 2023 to 2025, in nine of twelve quarters, that money grew faster than sales. Why it matters: sales can be booked before the cash comes in.

In 2021, customers paid in about forty-five days. Now it's over seventy. Their terms say thirty to sixty. On that, Burry's numbers check out. Palantir says it now bills some customers once a year, some afterwards. But the profit turned into cash: two point one billion dollars in the first half, on reported profit of one point nine.

The cash beats the profit because staff are paid in shares. That cost lands on you as new shares, and I counted it in stop one. So today's profit is real money. So: his record is good, the margins say software, and the profit is cash.

Against it: seven eighths of the goal isn't locked in, sales per person have to rise almost eight times, and sales take longer to turn into cash. Can he deliver? Maybe. For you, that's the second thing that has to be true: the software does the work. That's what keeps the thirty-nine cents per dollar. Ten times still has to come from somewhere.

13:01 Stop three: where the growth has to come from

Stop three: where the growth has to come from. In stop one, I counted thirty-eight billion dollars in sales for 2030. Thirty-three from America. Five from the rest of the world. First, the five. Does it hold? Last year, Palantir's sales outside the US were one point two billion dollars. About a quarter of all its sales. To reach five by 2030, that has to grow about four times.

In 2025, governments abroad bought forty-seven percent more. Companies abroad, two percent. This year, companies joined. In the first half, up twenty-six percent. Governments, forty-six. In total, abroad grew twenty percent last year. So far this year, thirty-five. Stop one counted with about a third a year. So if this pace lasts, the five billion holds.

More abroad is one way to beat the thirty-eight billion. Could it stay a quarter of sales, like today? That takes fifty-seven percent a year. So far this year, thirty-five. Governments carry abroad. For them, two things decide if it speeds up. The money they spend, and whether they trust American software. First, the money.

Europe and Canada's defense spending, after inflation: four hundred and eighty billion dollars in 2024. This year, six hundred and thirty-four. And NATO's summit this year committed to 'adopting powerful AI models.' It already bought Palantir's targeting system, Maven. In NATO's words, it took 'only six months.'

One country stands out. Over a third of Palantir's sales abroad are in Britain, and they grew every year. From a hundred and seventy-three million dollars in 2021 to four hundred and twenty-seven million in 2025. Last December, Britain's defense ministry signed a new deal: three years, two hundred and forty million pounds. About eighty million a year. There was no tender, so no one else got to bid.

So the money is there. Britain and NATO buy, and fast. But money is half of it. The other half is trust. In March, about the Strait of Hormuz, Trump told countries like Britain: 'the U.S.A. won't be there to help you anymore.' In May, German press reported that Germany's domestic intelligence agency picked a French rival of Palantir, ChapsVision.

In June, Washington ordered Anthropic, an American AI company, to cut off every foreign national from its two most advanced models. Four days later, France's prime minister said its domestic intelligence agency would replace Palantir. With the same French company.

One contract. But listen to his reason: 'We cannot depend on the goodwill of certain partners capable, as we have seen in recent days, of cutting off access to Anthropic's models.' Palantir says that contract is still in force. France's decision is to leave.

With the EU's new defense loans, at most thirty-five percent of a weapon's parts can come from outside the EU, counted by cost. Software isn't named, so the rule doesn't clearly cover Palantir. But the head of the EU Commission, Ursula von der Leyen, says: 'We must buy more European.' And Palantir has warned about this in every report since 2020: its 'perceived relationship with the U.S. government.'

My read: the direction is set. For intelligence software, the EU wants its own, and it's putting the money at home. For the rest of defense, like Maven, its own takes years to build. Until then, Europe buys American. The US supplied fifty-eight percent of European NATO members' arms imports. Karp, in February: 'we really don't have the bandwidth to do anything that's difficult outside of America.'

On Europe, he added: 'the issue ends up being their issue more than our issue.' And Britain keeps buying. So the five billion can hold. More than five, I wouldn't count on. For you, that's the third thing that has to be true: abroad keeps growing about a third a year. Nine tenths of the growth is America's. Can America get there? That's next.

17:03 Can America get there?

So, can America get there? Karp's ten times counts from last year's three point three billion. Palantir's forecast, minus abroad, gives about six point six this year. The goal is thirty-three. From here, that takes about fifty percent a year, for four years. Right now, US sales are up a hundred and ten percent in the first half. So today, America grows twice as fast as the goal needs.

Karp, in May: 'We are at our limit doing a hundred percent this year.' And: 'maybe we can do a hundred percent next year in the U.S.' So the speed is there. Next year is his 'maybe'. America has two halves: the government, and companies. About half each.

Growth has slowed before. US sales went from thirty-two percent growth in 2022 to nineteen in 2023. The government half, eleven and a half. Can each half keep up? The government first. Last quarter, the government half grew ninety percent.

If the split holds, at the goal the government buys about seventeen billion dollars a year from Palantir. Leidos is the government's biggest IT contractor. It sells about that much, to everyone.

Oh well. It's only taxpayers' money.

Of every dollar Washington spends on IT, Palantir gets about two cents today. At the goal, ten. For the goal: every president has committed more to Palantir per year than the one before. Both parties. Committed means money set aside in contracts, as the public database shows it. About two thirds of that money is defense. So the government half rides on the defense budget.

This year, the Pentagon asked for two point three billion dollars. For Palantir's Maven, the targeting system NATO bought, and a second program. Congress hasn't passed it. Until December 11th, the Pentagon runs on last year's budget. So that new money isn't there yet.

And committed is less than the headlines. Those are ceilings. The Army's deal says 'up to ten billion dollars'. So far, the Army has committed one point two. Remember 'almost alone' from stop two? Across all federal contracts, it has a number. Eighty-two percent of Palantir's federal contract dollars had one bid, or no competition at all. For the goal, few rivals. Good thing there's competition.

But Palantir's own annual report warns that government spending 'may not be sustained at the levels' of last year. Palantir reports on November 2nd. The next day, America votes for the Congress that writes the budget. Civilian orders follow who's in power. Health and Human Services ordered much less under Trump so far than under Biden. ICE, immigration enforcement, ordered more.

So today, the government half grows fast enough. But what it buys depends on who's in power, and on Congress, every year. Now companies. If the split holds, at the goal US companies buy about sixteen billion a year from Palantir. Snowflake from stop one and Databricks also sell data software. Combined, worldwide, they sell twelve point four billion. Less than Palantir's company half, in America alone.

Of every dollar American companies are forecast to spend on AI software, the goal needs between five and twenty cents. Washington's side needed ten. The forecasts only reach 2028, and they differ a lot. For the goal: US companies grew a hundred and forty-nine percent last quarter. Deal value still to come from them is up a hundred and twenty-four.

Against the goal: the tech research firm Gartner predicts that by 2028, seventy percent of companies will abandon AI agents built by engineers the seller sends in. They're called FDEs. Gartner doesn't name Palantir. Palantir's CTO did, in August: 'only Palantir has FDEs.' So Gartner's warning is about how Palantir sells.

And in McKinsey's survey, only thirty-seven percent of companies see AI in their operating profit. Companies like these are the buyers of the sixteen billion. Most don't see the payoff yet. Palantir's customer count grew twenty-four percent, down from forty-three. So more of the growth has to come from bigger bills.

For you, that's the fourth thing that has to be true: both halves keep about fifty percent a year. Four years in a row. The government on budgets and votes. Companies on AI paying off. Today both grow faster. In 2023, the government grew eleven and a half. That's the business. Now the stock.

21:36 Before the stock: the people who write about it

Before the stock, the people who write about it. If you're only here for the analysis, skip to the chapter 'What moves the stock?' Everyone else: I'm going to enjoy this. First, the name. Palantir is named after the seeing stones in Lord of the Rings.

Peter Thiel named it. The headquarters was called the Shire. His Palantir shares sit in companies called Rivendell. The New York office was Gotham. The ethics hotline? The Batphone. They're Lord of the Rings and Batman fans. Look into one, and you see what the others see. Let's hook it up to my database. More than eight hundred thousand finfluencer headlines, on every stock. Good god.

Yeah, I know.

22:18 The headline factory

Here's what the finfluencers sent out. Since January 2025: more than four thousand four hundred Palantir headlines.

Palantir reports four times a year. Motley Fool writes about it twice a day. That's about a hundred and ninety articles for every report. What do you write about one report a hundred and ninety times? Three hundred of them say buy. Nothing new. Finfluencers believe in rising prices. Even when it falls, it's really rising.

So what do they say? Here's the stock since January 2025. Nineteen times, Motley Fool has asked if it's too late to buy it. At seventy-nine dollars, the answer was 'stay away.' At two hundred and nine: 'just getting started.' Uuuuh?

But what did they say when it fell? On the highest close of 2025, two hundred and seven dollars, Seeking Alpha ran this. 'I regret moving to the sidelines, and I'm diving back in.'

Motley Fool, at a hundred and seventy-four: 'dips like this are golden opportunities.' At a hundred and sixty-eight: 'It won't be down for long.' Barchart, at a hundred and thirty-seven: 'Don't miss this chance to buy the dip.' At a hundred and fourteen: 'Why it's time to buy the dip.' Those two Barchart articles have the same writer, and mostly the same text. Only the price moved.

The bottom came the next day. A hundred and seven dollars. Not one of these said wait. It came back. The one who bought on Seeking Alpha's headline is still down ten percent. When they put numbers on it, the numbers get big. Motley Fool, January: 'Palantir is poised to mint ten thousand millionaires.' Their example: 'Microsoft minted ten thousand millionaires.'

By their own number, Microsoft went up more than a hundred times. A hundred times from that day makes Palantir worth about forty-three trillion dollars. Ha ha ha. Today it's four hundred and forty-seven billion. Benzinga, in March: 'Palantir may control the fifty trillion dollar physical AI system.' Same article: 'no single company owns physical AI.'

Even a sell headline can be a buy. Like this one. 'President Trump Sells Palantir Stock and Buys an AI Stock.' It ends: 'Investors should consider buying a small position in Western Digital.' Bold call. Two articles a day also means mistakes. Motley Fool, February 2025: 'Palantir's co-founder and CEO Peter Thiel.' The CEO is Alex Karp.

And if you missed Palantir, they have the next one. Twenty-six headlines used the words 'the next Palantir'. Ten were from Motley Fool. The typical one since its headline: down fifty-eight percent. Palantir itself: up forty-one.

Finbold goes further. It asks AI where Palantir will trade on a given date. Nineteen of those dates have passed. Thirteen times, just guessing today's price was closer. Finbold has published more than thirty of these predictions. Congratulations, Finbold.

25:20 Who checks any of this?

So who checks any of this? 24/7 Wall St., in June: 'Palantir will join the three trillion dollar club on this date.' For a broker, the regulator has a rule. Everything must be 'fair and balanced.' And no predictions of performance, unless it shows the basis and the risks. A website needs no license for that. It counts as a publisher.

The Supreme Court said publishers are exempt for 'disinterested commentary and analysis.' Disinterested means no stake in it. Under Motley Fool's Palantir articles: 'The Motley Fool has positions in and recommends Palantir Technologies.' Disinterested. And Google? Its manual for rating search results says stock advice with no proof of expertise gets a 'Low'.

This month, Motley Fool ran this as a Google ad. 'Where to put a thousand dollars right now. This could be like Netflix in October 2004.' So to say buy, a broker needs a license and has to stay fair and balanced. A website needs a headline. And Google takes the ad money.

And my own footnote? My Hype Fund buys the stocks I roast. The ones they shout about most. I'll get to that at the end. For you: none of these headlines checks the four things on your list. They come after the price moves. Motley Fool says its free articles reach 'millions of prospective members each month.' Seeking Alpha says seventeen million visitors. Benzinga, twelve million.

That's a lot of people to tell 'buy' three hundred times. I could make a whole series out of just this.

26:52 What moves the stock?

So what does move the stock? The day after the Army announced its up-to-ten-billion-dollar deal, the stock fell two point six percent. The whole market fell one point six that day. The Monday after the S&P 500 said it was adding Palantir, the stock rose fourteen percent. Index funds had to buy it.

The biggest moves come on report days. Eight reports in two years, every one moved it at least six percent. So reports and index money move it. Even an up-to-ten-billion-dollar deal barely did. An index fund buys every stock in the index, by its weight. Of every hundred dollars into an S&P 500 fund, sixty-nine cents buys Palantir.

So if you own an S&P 500 fund, you already own some Palantir. When people sell the fund, it sells Palantir too. Either way, it trades at today's price. Nobody in the fund picks it. So index money never asks if the price is fair. What about insiders selling? Every sale by the people running a company is filed with the S-E-C, on a Form 4.

Right after Palantir joined the S&P 500, Rivendell 7 sold a billion dollars of Palantir stock. Remember Rivendell? That's Peter Thiel, Palantir's chairman. The sale was on a plan set in May, before the index news. So he didn't pick the day. Karp has sold three and a half billion dollars of stock since the listing. At about thirty-five dollars a share, about a fifth of today's price.

His shares came as pay, mostly as options. He sold on a plan set in advance. And he still owns about eleven billion dollars' worth. In the last year, nine tenths of his sales were automatic, to pay the tax on new shares. My read: insider selling looks scary until you read the Form 4.

Nobody now running Palantir or on its board has bought a share on the market since the listing. They get paid in shares. None of them buys more. Remember the fall from last November's high? Almost half, to a hundred and seven. Then it rose seventy-three percent, to a hundred and eighty-six. Sales grew every quarter, the whole way.

Remember, the price is the profit times the P/E. The profit grew. So the whole fall was the P/E: money and opinions. Some people bet on those swings. There are funds that do twice what it does, every day. And funds that do twice the opposite.

Remember the one who bought on that November headline? Ten thousand dollars in is now about nine thousand. In the biggest fund that doubles it, about five thousand three hundred. In the one that doubles against it, about three thousand seven hundred. Doubling every day loses in a stock that swings this much. Both directions lost.

Betting on the swings didn't pay. Holding lost less. But at the bottom, your ten thousand was about five thousand two hundred. Anyone who sold there missed the way back. This time, it came back. If the first four hold, a fall on the way doesn't change 2030. It only costs you if you have to sell.

So the fifth thing that has to be true for you: you can watch your money halve, and not sell.

29:57 What has to be true for you

So here's what has to be true for you to make money. One: in 2030, the profit per share times the P/E beats about two hundred and eighty dollars. At Karp's goal, that takes a P/E of forty-eight. For the goal, three more. Two: the software does the work. Sales per person almost eight times higher. Three: abroad keeps growing about a third a year.

Four: in America, government and companies both keep about fifty percent a year, four years in a row. And five: you can watch your money halve, and not sell. Right now, US sales grow more than twice as fast as the goal needs. Keep that up, and it goes beyond the goal. It has to last four years. The business has almost no holes. The price is the hole.

That's the risk: miss the growth, and the price falls. It already halved once, while sales were still growing. That's why five is on the list. So the number to watch is American sales. About fifty percent a year, until 2030.

Dates we check. November 2nd: Palantir's next report, and report days move it most. December 11th: the government's stopgap budget runs out. June 2027: Burry's bet against it runs out. And any report can bring the tax day from stop one. Drop your hype in the comments: where is Palantir on November 3rd, the day after its report, and the day America votes?

31:23 The Hype Fund, and the buy

The Hype Fund so far. Micron, minus fifteen percent since it joined. SpaceX, plus five. The same money in the S&P 500, same days: about flat. In every episode, we set a plan before we buy. Then we follow it. We sit calm in the boat. So. Would my math buy it at this price? No. It needs more than Karp's goal.

The Hype Fund buys what the headlines shout about. So it buys it anyway. That's the experiment. If Burry is right, Palantir's spot in the fund loses about three quarters.

To the moon! Haha hahaha!