Palantir connects separate systems into one model of factories, orders or soldiers, so staff and AI agents can act on it. Armies, spy agencies and big companies buy it.
The company · 2 · What exactly does it sell?
US government and US companies lead
Each $1 of sales, second quarter of 2026
42%40%9%9%
US governmentUS companiesForeign governmentsForeign companies
Sales to US companies grow fastest, up 148% last quarter.
The company · 3 · How does it make money?
Large contracts that grow each year
$7.29Bsales in the past year
46%operating profit per $1
$9.41Bof cash and securities, and no debt
Customers start with one project and expand, spending 65% more than a year ago.
Is it a good business? · 4 · What gives it an edge?
Leaving means rebuilding the model
65%
more spent by customers than a year ago
$146M
a year on average from the top twenty
75
Army contracts folded into one
The top three customers have stayed about ten years.
In defence, Palantir is only on Anduril's team for the Army's next system.
Is it a good business? · 6 · What does its record show?
Growth sped up and profit soared
Sales, $B
1.5
1.9
2.2
2.9
4.5
7.3
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025last 12 mo
Sales$1.54B → $7.29Bsales, 2021 to the past year
Profit−27% → 46%operating profit per $1
Share pay50% → 13%of revenue paid in shares
Every overhead line fell per sales dollar while still growing in dollars.
Is it a good business? · 7 · How much room is there to grow?
The market is not the limit
$452.5BGartner's 2026 AI software spending
$283.1Bin 2025
About 1.4%Palantir's share of it
US federal contract payments to Palantir roughly doubled in the year to September 2026.
Is it a good business? · 8 · Who is in charge?
The founders still run it
2005Alex Karp began running Palantir
47%of each sales dollar kept as operating profit last quarter
Karp, Thiel, Cohenthe founders, who control about half the vote
Under Karp it went from heavy losses to high profit.
The story right now · 9 · Why is Palantir an AI winner?
AI needs a map of the business
96%sales growth last quarter
148%growth in sales to US companies
6%growth in contract value signed
To let AI act inside a company or army, it first needs a live map of how things run. Palantir builds that map, so AI spending flows to it.
The story right now · 10 · Why is the price so high?
The price assumes the best case
Sales grow 96%A record quarter.
Easing to about 88%Next quarter's forecast.
The price needs 40%+For many years.
Below that hurtsA long way down.
Even on the analysts' path, the price is about a quarter above our value.
The story right now · 11 · What changed lately?
Sales up 96% and the forecast raised again
$2.313Brevenue in the third quarter
148%growth in sales to US companies
6%growth in contract value signed
64.5%of analysts now say buy, up from 44.4% in February
Strong quarter
Management said large deals signed were shorter than usual and expects them to grow.
The story right now · 12 · What is everyone arguing about?
The clearest AI winner, at any price?
Many investors think
The clearest winner in AI software
We think
Forecasts may be hit, but the price is still high
Against us: three forecast raises this year and customers spending more each year.
The price and our view · 13 · What is a share worth?
The price sits between our value and our best case
$109what we think a share is worth · the last close is of that
$109our value
Attractive below $87.05
Neutral
Unattractive above $131
$29.12gloomy: growth halves, less kept per $1
$109our estimate: experts' path, then slowing
$205sunny: growth beats the forecast by a quarter
Value per share in three futures. None is a prediction of the share price.
We estimate what a share is worth from the cash the business should earn over many years. Our rule: buy only when the price is at least a fifth below our estimate, as a cushion for being wrong.
The price and our view · 14 · How could we be wrong?
Growth could slow faster than priced
Growth slowsBiggest
Below 40% hurts.
Contracts can be endedShowing now
Signed value slowed.
AI labs sell directWatch
OpenAI's platform.
Army ordersUnknown
Under a $10B ceiling.
Defence budgets, rivals and founder control add to the risk.
The price and our view · 15 · How will we know?
February tests whether growth is holding
Signed contract value grows fasterGood sign
Growth stays above 40%Good sign
OpenAI wins big customersBad sign
FEB 20271
Fourth-quarter 2026 results, 1 February. Companies report every three months; each report is a test of the story.
Our view also changes on price alone: if the shares fall to within a fifth of our estimate, we move to Neutral.
The price and our view · 16 · So, what do we think?
A strong business priced for the best case
$109our value
Attractive below $87.05
Neutral
Unattractive above $131
UnattractiveMedium conviction
Palantir's software maps how a business or army runs and lets people and AI act on it, growing faster with high profit and no debt. But even on the analysts' path, the price is about a quarter above our value.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
Large organisations keep their information in hundreds of separate systems that do not talk to each other. Palantir's software connects them into one live model of the organisation: its factories, orders, soldiers or patients, and the rules that link them. Staff, and now AI agents, can then make and carry out decisions on top of it.
Its buyers are armies, spy agencies and big companies. The alternative is building the same thing in-house with consultants, which is slow and rarely finished. Palantir wins them with free pilots and week-long trials.
How it makes money
Customers sign contracts of one to five years and pay for the software over that time. Government and commercial customers each bring in about half the money.
Once installed, the software costs little to run: in Q2 2026 Palantir kept 85 cents of each dollar after its direct costs, and each half of the business kept more than 70 cents after its own costs and sales spending. So each new dollar of sales is mostly profit.
Growth and profit
What growth costs
Growth keeps making Palantir more profitable. Sales have grown far faster than spending on selling, research and running the company, so each new dollar of sales costs less to win. Paying staff in shares has shrunk as a share of sales, though it is still large. The business needs almost no buildings or equipment to grow.
Its record
The numbers
Is it growing?Fast
Sales grew 87% over the twelve months to Sep 2026, faster than the 56% in the year to Dec 2025.
Is it profitable?Strong
Operating margin is 46%, up from 5.4% in the year to Dec 2023.
Does growth pay off?Pays off
Running costs take 39¢ of each sales dollar, down from 105¢ in the year to Dec 2021.
Does it earn good returns?Strong
Each dollar put into the business earns 37% a year after tax, up from 3.6% in the year to Dec 2023.
Is it making cash?Strong
It turned 56% of sales into free cash, $4.05B over the last twelve months.
How steady is it?Uneven
Yearly sales growth has ranged from +17% to +56% over the last 5 years.
Are shares being diluted?Diluting
The share count rose 34% in 4.8 years, and pay in shares costs 13% of sales.
Why it wins
The evidence
Once an organisation runs its decisions on Palantir's model of itself, leaving means rebuilding that model somewhere else. Customers rarely do.
The evidence is how much more they spend. Customers from a year ago spent 65% more in Q3 2026. The twenty largest pay an average of $146M a year, and the top three have stayed about ten years. The US Army folded 75 separate contracts into one Palantir agreement rather than put them out to tender again.
Who could beat it
The rivals in full
The AI model makers now sell directly to the same companies: OpenAI launched a platform for running AI agents across a business in February 2026. In defence, the Army gave Anduril the prototype for its next command system, with Palantir only on Anduril's team. Palantir says it beats the labs in trials, but nobody outside has confirmed one.
Each rival
Customers' own builds on data platforms such as Snowflake and DatabricksSNOW
The whole job: the 10-K's first-named competitor is the customer's in-house team, assembling data, models and applications itself.
OpenAI Frontierprivate
AIP's agent layer: a model lab selling agents straight into a company with its own engineers and consulting partners.
Anduril Industries (Lattice and the Army's next command system)private
Part of the battlefield software layer Gotham and Maven occupy, from a defence prime with its own hardware.
C3.ai, IncAI
The same product: enterprise AI applications sold to companies and governments.
Booz Allen Hamilton and other federal integratorsBAH
Custom government systems built by billed engineers rather than licensed software.
How big it could get
The market and who has signed
The market
Gartner, a research firm, expects the world to spend $452.5B on AI software in 2026, up from $283.1B in 2025. Palantir's sales over the past year were about 1.4% of that, so the market is not the limit.
Government money is growing too: US federal contract payments to Palantir roughly doubled in the year to September 2026.
Who has signed
The Army's enterprise agreement is worth up to $10B over ten years, and the Maven targeting software contract about $1.3B to 2029. Both are ceilings rather than promises. Most Palantir contracts can be cancelled at will, so only $4.9B of the $13.4B customers have agreed to is firm.
Who runs it
Leaders, pay and the cash
Who runs it
Alex Karp has run Palantir since 2005, with co-founders Stephen Cohen as president and Peter Thiel as chairman. Under Karp it went from heavy losses to keeping almost half of each sales dollar as profit. Insiders own about a tenth of the shares, but a special share class gives the founders half the votes, so other shareholders cannot outvote them. Karp's pay is a large grant of options from the year it listed that simply vests over time: no bonus, nothing tied to sales or profit, only the share price.
The last report
The latest report
Palantir's sales rose 96% to $2.313B and it raised its forecast again
What it did to the thesis Strong quarter
It proved the demand is real and speeding up, so the business case is stronger. Contract value signed grew far slower than sales, which the report left unexplained.
The report
In Q3 2026 revenue was $2.313B, up 96%, against a forecast of about $2.16B. Sales to US companies rose 148% and passed sales to the US government for the first time. Existing customers spent 65% more than a year earlier, up from 57% in June, and US company customers grew to 686.
The one weak figure was commitments. Contract value signed rose only 6% on the year, and the $4.9B that customers cannot cancel was flat on the quarter. Management said the large deals signed were shorter than usual and expects them to grow. Management raised its 2026 forecast to about $8.53B, 91% growth.
What's happening now
The background
The shares fell almost half from November 2025 to June 2026 in a sell-off across software stocks, then jumped 29.5% the day after Q2's results. Analysts have turned more positive: about two thirds now rate it a buy, against less than half in February.
The Q3 test came on 2 November and passed: revenue beat the company's own forecast by about $150M and the 2026 forecast rose again. The next test is the fourth quarter, due around 1 February, and whether contract value signed grows faster than the 6% of Q3.
Where we disagree
Both sides in full
The market
Most analysts call it the clearest winner in AI software; a loud minority, including Michael Burry, calls the price a bubble.
We believe
The business may well hit 2028's forecasts, but even then the price is high, so the return depends on growth beyond 2028 nobody can yet see.
The argument
Bulls are projecting Q3's 96% forward, while the company's own forecast has fourth-quarter growth easing to about 88%.
The case against us is strong: Palantir raised its 2026 growth forecast from 61% in February to 82% in July and 91% in November, and customers keep spending more each year.
What it's worth
How we value it
Valued on the analysts' revenue path for three years, each later year closing 30% of the gap to 4%, on official operating profit with stock pay a cost and tax paid in full; the thesis's own evidence is the check on that path.
The three cases
Bear case$29.12
Growth runs at half the analysts' path, 30%, 24% and 30%, the margin settles a fifth lower at 29.2%, and new capital earns only the 8% rate: the downturn test.
Our value$109
Revenue grows 60%, 48% and 60% over three years, the analysts' path (C51), held to at most 60% a year, then fades toward 4%, closing 30% of the gap each year; the margin moves to 36.5%, the analysts' year-three figure after stock pay and on the filed basis (7.3% less); tax is 25%, the federal 21% plus state tax, the full rate once past losses are used up.
Bull case$205
Growth runs a quarter above the analysts' path, 75%, 61% and 75%, the margin reaches 40.2%, and new capital earns 15%.
What could go wrong
Every risk
The risks
Growth slows faster than priced
At 30% growth a year, 2028 revenue lands 23% below what the price expects.
Contracts can be cancelled
Most contracts can be ended at will; only $4.9B of $13.4B agreed is firm, and that $4.9B did not grow in Q3.
AI labs sell direct
OpenAI and others now offer companies their own platforms for AI agents, aimed at the same buyers.
Defence budgets and rivals
US government is 42% of revenue, and Anduril won the prototype for the Army's next command system.
Founders control the vote
The three founders hold close to half the votes through a special share class, so outsiders cannot change pay or strategy.
Share pay is a real cost
Stock given to staff costs about 14% of revenue and is left out of the adjusted profit figures.
What we still do not know
How much the Army actually orders each year under its $10B ceiling.
How much of US company growth comes from a few very large deals.
Whether the AI model makers are winning trials against Palantir; only Palantir has described any.
What changes our view
Every route to a new view
Every route to a new view
→ Neutral
The share price falls to about $131, back within 20% of what we think the business is worth.
→ Attractive
The share price falls to about $87.05, 80% of what we think the business is worth.
The next test
It supports the case if
Revenue above $2.653B, US commercial up at least 130%, net dollar retention at or above 150%, total contract value closed growing faster than 6% on the year, and remaining performance obligations above $4.9B.
It pressures the case if
Revenue below $2.649B, US commercial growth under 110%, net dollar retention below 150%, or remaining performance obligations below $4.9B.
We would also be wrong if
Growth in the number of US company customers falls below 20% a year.
One of the three largest customers cuts or cancels its contract.
The verdict
Scores and how we weighed it
The scores
Growth
4 of 4
Business quality
3 of 4
Financial position
4 of 4
Execution record
4 of 4
Per share capture
3 of 4
Potential
3 of 4
Risk profile
3 of 4
Valuation
2 of 4
Evidence quality
3 of 4
How we weighed it
ValueNegative
The price sits above our value even on a very steep path; the bear case is far below.
BusinessPositive
Strong execution, high profit on each sale and no debt.
GrowthPositive
Growth is speeding up and demand is real.
RisksNegative
Growth may slow faster than priced, most contracts can be cancelled and AI labs offer their own platforms.
RivalsMixed
No filed rival compares directly; against its own normal it is far dearer.
The groupMixed
It sits in none of our ranked groups; AI and defense news move it.
Latest newsPositive
The last report proved demand is speeding up, though signed contract value grew more slowly.
In plain words
Our estimate of value
We forecast how much cash the business can earn over the next ten years and beyond, then ask what that stream is worth today to an investor who wants an 8% yearly return. We do it three times: a gloomy future, an expected one and a sunny one. We call a share Attractive only when its price is at least 20% below the expected estimate, and Unattractive when it is 20% above.