X-Ray Analyses›Technology›Palantir Technologies Inc.
X-RayNº 11222 November 2026
◆ Follow-upSecond X-Ray of Palantir Technologies Inc.. Our previous, dated 2 Oct 2026 at $190.04, concluded “Pass at $190 — Begin Near $110”. This report checks that call against what has happened since. Read it →
X-Ray Analysis◆The Buffett Lens
P

Palantir Technologies Inc.

NASDAQ: PLTR·Software — Infrastructure·United States·Explore PLTR live ↗

Palantir — a quarter-cheaper, and still not cheap

◆ The Buffett LensIn October we called Palantir one of the best businesses we have studied, at a price that needed a miracle, and said we would begin near $110. Seven weeks later the shares are 24.5% lower, although the company beat its own forecasts again and raised its guidance. Two things happened: the one weakness we named, how little new business is being signed, showed up in the third-quarter numbers, and the market fell hard for reasons that had nothing to do with Palantir. At $143.49 the price still asks for about 28% growth a year for a decade. Closer, still a pass. Begin near $115.
◆ Educational analysis & opinion — not investment advice. Figures as of 22 November 2026. See full disclaimer below.
8
Moat
4
Management & Capital
10
Financial Strength
9
Growth
2
Valuation
6.4
◆ The Scorecard · one-second read
"The company beat its forecast. The price fell a quarter. Both can be right."
−24.5% since October · revenue +96%, guide raised to $8.53bn · bookings +6%, backlog flat · Q4 guided +88% · ~47× sales, ~85× free cash flow · the price implies ~28% growth for ten years
◆ Type · Hypergrowth at a hyper-price, a quarter cheaperDividend · None; $10.6bn of cash, no debt★ Follow-up · bookings cracked, revenue did not; level $110 → $115
Price & total return · live
The Tape, since we wrote
The gold pin marks the price we used. Daily closes, live.
—
Drag across the chart to measure
Live price history is momentarily unavailable · price at this X-Ray: $143.49.
Go deeper — the Trading Desk, 15 years of financials, DCF & peers for PLTROpen PLTR →
◆
Part
I

What we said in October — and what happened

The verdict held; the weak spot we named is the one that cracked

On 2 October we published Palantir at $190.04 with a verdict we rarely give a business we admire: pass. We said revenue was growing 93% and accelerating, that the margins and the customer economics were among the best in software, and that none of it mattered at about 74 times sales, a price that assumed about 32% growth a year for a decade. We said we would begin near $110, where 25% a year would earn a buyer about 10%, and that "the price can halve again on a change of mood". The shares closed at a record $209.05 a week later and at $143.49 on 20 November. Here is the account.

What we said on 2 OctWhat happened by 20 NovGrade
'Pass at $190; begin near $110'−24.5%; the low was $142.28 on 20 November, still 29% above $110.✓ Right so far
'The price can halve on a change of mood'−32.6% from the 12 October intraday high in six weeks, with no bad news except bookings.✓ Right, in progress
Growth accelerating (Q2 +93%)Q3 +96%, a ninth acceleration; but the Q4 guide implies +88%, the first slowdown in ten quarters.✓ Right; Q4 too early
Backlog '$4.9bn, small for the valuation'; contracts terminableBacklog (RPO) flat at $4.9bn; contract value booked +6%; US commercial +11%. The market's stated reason on results day.✓ Right — it cracked
Net dollar retention 157%; 1,049 customersQ3: 165%; 1,090 customers, +20% (from +24%), from transcripts.✓ Right
Gross margin 85%Q3 83.4%, as cloud hosting costs rose $135m in a year.◆ Slipping
Insiders sold ~$1.2bn in 2025, ~$0.7bn in 2026Re-summed from 134 Form 4s: correct. 2026 now $772.5m, with ~$176m more filed as proposed sales.✓ Right
Government risk: stopgap to 11 Dec, midtermsThe midterm result produced the period's worst day, −9.7% on 5 November.✓ Right

Our errors. None changed the verdict, but six details were wrong or loose when we wrote them. We showed company buybacks in 2025 as nil; they were about $75m, still a token. We called Peter Thiel a director; Thiel is Chairman. Our analyst split ("13 buy, 10 hold, 3 sell") came from a block in our feed that contradicts the feed's own history, which shows far more buy ratings. We gave the cash pile as $9.4bn in one place and $9.2bn in another; both figures are in the filings, but they measure different things and we should have used one. A chart caption said the 2026 guide was $8.2bn while the arithmetic used $8.15bn. And price to sales was 74 times, not 73. Every insider-sale total and every second-quarter number we re-checked was correct.

Our October price → today
$190.04 → $143.49
−24.5%; −31.4% from the 9 Oct record
Revenue guide for 2026
$8.15bn → $8.53bn
raised on 2 November
Our level
$110 → $115
the 25% path on the higher base
◆
Part
II

★ Up 95%, then down 31%

A record in October, a rout in November

Up 95%, then down 31%PLTR daily close, $, 1 Jun – 20 Nov 2026.100125150175200225our level, $115 (was $110)1234567JunJulAugSepOctNov125 Jun · the low, $107.2724 Aug · Q2 results, +29.5%31 Oct · our X-Ray, $190.0449 Oct · record close, $209.0553 Nov · Q3: bookings +6%65 Nov · midterms, −9.7%720 Nov · AI sell-off, −6.6%, $143.49★ −24.5% since our X-Ray, and still 25% above where we would begin.Sources: our price history; Palantir releases; press reports for the 5 and 20 Nov moves.
DateWhat happenedPLTR that day
9 OctRecord close after broker upgrades (Barclays initiated at Overweight, as reported).+5.2%, $209.05
16 OctNo company news; doubts about AI funding hit software shares (as reported).−7.2%
30 OctA week of disappointments from Alphabet, Amazon and Intel; no Palantir news.−6.3%
3 NovQ3 results: beat and raise, but contract value booked +6% and backlog flat; shares −8% before the open, −1.7% at the close.−1.7%
5 NovDemocrats win both chambers in the midterms; defence and AI shares sold (TIKR).−9.7%
19 NovAnthropic's listing; Palantir up 8.7% during the day, then down.−1.9%
20 NovAI sell-off: Nasdaq −5.3%, Nvidia −14%; no Palantir news (Yahoo, Bloomingbit).−6.6%, $143.49

Mostly the market, partly the business. Of the seven big days, one was about Palantir's own numbers, and it was the smallest. The midterms cost more in a day than the results did in three, and the last week was the AI trade unwinding across the board: over the month, ServiceNow fell 41% and Datadog 39%, against Palantir's 25%, according to Yahoo. Consensus earnings estimates for 2026 to 2028 rose 5–10% while the price fell. Everything that went wrong was the price of the shares, not the profits of the company.

That is the classic setup for a value investor, with one difference: the price was so high to begin with that a 31% fall from the record leaves it where it was in early August. In June the shares were $107. We said in October that a change of mood could halve the price, and the mood has changed only part of the way.

What would finish the job is unknowable in advance: a weak fourth quarter, a government shutdown in December, a larger supply of insider shares, or simply more weeks like the last one. What would reverse it is equally unknowable: a strong quarter of bookings, or a return of appetite for AI shares. We do not try to predict either. We set a price at which the business, on generous assumptions, earns a buyer 10% a year, and wait for it. The price has moved a long way towards that level in seven weeks; it can move the rest of the way as quickly, or not at all.

◆
Part
III

★★ Revenue accelerated; bookings did not

The weak spot we named in October is the one that cracked

Revenue accelerated; bookings did notGrowth, %: Q2 2026 (pale) and Q3 2026 (dark).Revenue (year on year)+93%+96%Total contract value booked (yoy)+49%+6%US commercial contract value (yoy)+153%+11%US commercial remaining deal value (qoq)+27%+4%Q2 2026Q3 2026★ Remaining performance obligations: $4.9bn, flat on the quarter. Q4 guide: +88%.Sources: Palantir Q2 and Q3 2026 releases and 10-Qs.

On 2 November Palantir reported third-quarter revenue of $2.31bn, up 96%, well ahead of its own forecast of about $2.16bn. US commercial revenue grew 148% to $984m and, for the first time, exceeded US government revenue. The operating margin was 47% on the reported measure and 62% on the adjusted one. The company raised its 2026 revenue guide to $8.53bn and its adjusted free-cash-flow guide to $4.8–5.0bn. By any normal standard it was an exceptional quarter.

But a company's revenue tells you what it delivered; its bookings tell you what it will deliver. The total value of contracts signed in the quarter grew only 6%, after 49% in the second quarter. In the US commercial business, the engine of the story, contract value grew 11%, after 153%. Remaining performance obligations, the backlog the company must report, were flat at $4.9bn. The finance chief said, according to the call transcript, that several large "Sovereign AI" deals had been signed for shorter periods than usual, and that size and duration should grow. That may be true. But in October we pointed to this exact number as the thin part of the case, and it is the number that moved the wrong way.

Q3 2026Valuevs a year earlier
Revenue$2,313m+96% (+20% on Q2)
US commercial / US government$984m / $953m+148% / +96%
Operating margin, reported / adjusted46.9% / 61.7%Rule of 40: 158%
Free cash flow$1,197m52% of revenue
Total contract value / US commercial contract value$2.92bn / $1.46bn+6% / +11%
Remaining performance obligations$4.9bnFlat on Q2
Q4 guide$2.65bn revenue+88% (first slowdown in ten quarters)

Why it matters at this price. A company that is growing at 96% does not need bookings to grow at 96% in every quarter; deals are lumpy. But the market is paying for many years of very high growth, and the consensus already expects 2027 revenue of about $13.2bn, 55% above this year. Revenue a year from now comes largely from what is signed today. If the fourth quarter shows contract value and backlog growing strongly again, the third quarter was a pause. If it does not, the 2027 consensus is too high, and so is the price. Palantir's own guide for the fourth quarter, +88%, is still extraordinary; it is also the first step down after nine steps up.

The other backlog measure. Palantir also reports "remaining deal value", which counts the full value of contracts including options customers have not yet exercised and periods they may cancel. It is a generous number, and we said in October that it assumes every option is taken up. Even on that generous measure, the third quarter was quiet: total remaining deal value of about $13.4bn grew 2% on the quarter, and the US commercial part 4%, after 27% in the second quarter. Two measures, one strict and one generous, telling the same story is worth more than either alone. A further small item points the same way: contracts with customers in which Palantir itself invested, about $326m in total, have no remaining options and will stop contributing revenue after this year.

◆
Part
IV

The customers: deeper, slower to arrive

Existing customers spend more; new ones come more slowly

The customer numbers tell the same two-sided story. Existing customers are spending more than ever: net dollar retention rose to 165% from 157%, and the twenty largest customers spent an average of $146m each over twelve months, up 76%. New customers are arriving more slowly: 1,090 in total, up 20% after 24%, and the US commercial count grew 5% in the quarter. The business is getting deeper faster than it is getting wider. That is the profile of a product customers love and of a market that is not yet as large as the price assumes.

Costs are rising underneath. Gross margin fell to 83.4% from 84.7% in the second quarter, because cloud hosting costs rose $135m in a year; Palantir has committed to at least $5.6bn of cloud spending over ten years. Stock-based pay rose 85% to $319m in the quarter. And two small details deserve a footnote: receivables grew 27% in the quarter against revenue growth of 20%, and Palantir received about two million shares of Surf Air Mobility in September "as payment for certain outstanding receivables", taking its stake in that customer to 8.5%. Neither is large. Both are the kind of thing an owner should watch in a company growing this fast.

Concentration, and what it means. The three largest customers provided 16% of revenue in the first nine months, and have been with Palantir for fifteen years on average; no single customer exceeds 10%. Depth of this kind is a strength while those customers keep expanding, and it is why net dollar retention can rise while the customer count slows. But it also means that a large part of the growth comes from a few hundred organisations spending more each year, and an organisation that has already doubled its spending twice will not double it forever. At some point the growth has to come from new customers, and the third quarter is the first in which that point looked further away rather than closer.

Customer measureQ2 2026Q3 2026
Net dollar retention157%165%
Top-20 customers, average revenue (12 months)$124m (+67%)$146m (+76%)
Customers / growth1,049 / +24%1,090 / +20%
Gross margin84.7%83.4%
Rest of world revenue growth+34%+26%
◆
Part
V

The government, after the midterms

41% of revenue, a new Congress and a funding deadline

The US government provided 41% of Palantir's third-quarter revenue, and that revenue grew 96%, from existing programmes and new awards, according to the finance chief. The Army's TITAN programme and the Maven system, which now has more than 100,000 users across the services according to the technology chief, are the flagships. We found no new defence award dated October or November.

On 5 November the shares fell 9.7% after the Democrats won both chambers of Congress. The fear, as reported by TIKR, was the defence budget: the White House asked for $1.5tn for 2027 against $901bn approved, and Jefferies expects about $1.15tn. Even the lower figure would be a large increase. The more specific risk for Palantir is scrutiny. In October we noted a $243.9m no-bid Pentagon award; a new Congress with oversight committees in opposition hands is more likely to ask about such awards. And the government's stopgap funding was due to expire on 11 December; Palantir's 10-Q says continuing resolutions and lapses in funding "have and could in the future" delay its government revenue.

Patronage cuts both ways. In October we wrote that Palantir's closeness to the current administration was an asset that could become a liability. The midterms did not change who runs the Pentagon, but they changed who funds and oversees it. A company whose government revenue is growing at 96% is visible, and visibility attracts questions: about no-bid awards, about the use of its software in military operations (Part XII), about its work for immigration enforcement. None of these has produced a finding against Palantir. Each is a reason why government revenue, which looks like the steadiest part of the business, may carry more political risk than its contracts suggest. For an owner the sensible response is not to guess at politics but to demand a price that does not depend on them.

Government itemStatusRead
US government revenue, Q3$953m, +96%✓ Strong
Defence budget for 2027$1.5tn requested; ~$1.15tn expected (Jefferies, as reported)◆ Still rising
Stopgap fundingDue to expire 11 December (as reported in September)◆ Watch
UK NHS data platformReported kept to March 2029; successor tender from 2027◆ Not guaranteed
Germany (Bundeswehr)Award reported due by year-end◆ Open
◆
Part
VI

The moat, rechecked

Deeper with existing customers; the width is the question

The claim (October)Evidence nowVerdict
Switching costs: the software becomes the operating systemNet dollar retention 165%; the top three customers have been with Palantir 15 years on average.✓ Stronger
AIP turns AI models into productsUS commercial revenue +148%; but contract value +11%.◆ Proven, slower to sign
Government trust and clearancesUS government +96%; Maven open to all services.✓ Intact
Threat from AI labs and cloudsAnthropic now listed; the CEO calls the labs 'symbiotic' (transcript). No evidence of share loss.◆ Too early

We keep the moat at 8. Customers who adopt Palantir spend more each year, and the old ones do not leave; that is the signature of a real moat. The question raised by the third quarter is not depth but width: how many organisations will adopt it, and how fast. The new-customer count and the bookings say the answer is "many, but not as fast as the share price needs".

How knowable is the next ten years?
2/5 — the product is clear; its market size is not. We can see why customers stay and spend more. We cannot see whether Palantir will be selling to 3,000 large organisations in 2036 or to 30,000, and the price depends on that.
What you must believe to buy at $143.49
  • That revenue grows about 28% a year for ten years, to roughly $97bn in 2036, from $8.5bn this year.
  • That the third quarter's weak bookings were a pause in signing, not the first sign of a smaller market, and the fourth quarter shows it.
  • That operating margins stay near Microsoft's, despite rising cloud costs and stock-based pay growing faster than revenue.
◆
Part
VII

The competition

A listed AI laboratory, a re-accelerating data platform, and Europe's own champions

RivalWhere it meets PalantirThe latestRead
AI laboratories (Anthropic, OpenAI)Selling AI directly to companies.Anthropic listed on 19 November; Palantir says it works with Anthropic. A sell-side analyst on the call claimed about half of Anthropic's customers buy directly (unverified).◆ Partner and rival
Snowflake, DatabricksData platforms for AI.Snowflake's product revenue accelerated to +37% and its backlog is $9.0bn, nearly twice Palantir's.◆ Growing
Anduril and defence primesMilitary software.Shares Army work with Palantir (TITAN production split, as reported).◆ Partner and rival
European providersSovereign data platforms.Germany's domestic intelligence service chose ChapsVision; the Bundeswehr is testing alternatives (as reported).◆ Narrowing abroad

Across our board. Two comparisons matter. Snowflake, a data platform that sells to many of the same companies, has a contracted backlog of $9.0bn, almost twice Palantir's, on revenue that is smaller; Palantir's contracts are shorter and more often terminable, which is one reason its backlog understates its revenue and one reason it is a weaker guarantee of it. And Anthropic's listing gives investors a pure AI laboratory to own for the first time. Some of the money that bought Palantir as the way to own AI can now buy the laboratory directly. That is a flow, not a fundamental, but at 47 times sales flows matter. Anthropic is expected to join the Nasdaq-100 in December, according to reports, which would require index funds to sell other members, Palantir among them.

Sovereign AI. Palantir framed its third quarter around "Sovereign AI": governments and national companies that want their own AI systems, run on their own terms, often on open-weight models rather than a laboratory's. That is a good market for a company that already holds security clearances in many countries, and it explains some of the shorter contracts. It is also the market in which European buyers are most inclined to choose a European supplier. International commercial revenue grew 21% in the quarter but barely moved from the second quarter, and international revenue is now only 16% of the total. Palantir's growth is, more than ever, an American story.

◆
Part
VIII

Management, insiders & capital allocation

Large awards, steady selling, and one plan to buy

A
Alex Karp · Co-founder and Chief Executive
No open-market sales from October to 19 November. A Form 144 filed on 20 November gives notice of 675,000 shares (~$100m), most of it to cover taxes on shares vesting that day. A trading plan adopted in August covers up to 7.08m shares to June 2027.
S
Shyam Sankar · Chief Technology Officer
Sold 384,680 shares for $64.9m on 10–11 November, under a plan adopted in August, and filed notice of about 385,000 more on 19–20 November. Received 564,830 new shares in an award on 28 October.
P
Peter Thiel · Co-founder and Chairman
No sales since October. An entity Thiel is deemed to own adopted a plan in August to buy up to 4m shares, subject to price conditions, until December 2027; no purchase had been filed by 20 November.

Capital allocation. We lower the score from 5 to 4. Palantir has no debt, $10.6bn of cash and securities, and very little need for capital; the question is only what it does with its shares. On 28 October the board granted large new awards: 564,830 shares to the technology chief and 211,208 each to the finance chief and the chief revenue officer, worth roughly $111m and $42m each at that day's price; the terms were not disclosed in the filings. Stock-based pay grew 85% in the quarter, faster than revenue. Buybacks were $1.5m in nine months. Options, rights and restricted shares outstanding amount to about 8% of the share count. A company this profitable, issuing this many shares to its executives while the executives sell, is not treating its share capital as scarce.

Insiders. Sales so far in 2026 total $772.5m, with about $176m more filed as proposed sales on 19–20 November. Some of that is selling to pay tax on vesting shares, which is mechanical. The one counter-signal is the Thiel entity's plan to buy up to 4m shares. A purchase would be the first insider buying of any size we have recorded at Palantir. It has not yet happened.

The plans ahead. The 10-Q filed on 3 November lists the trading plans now in force. The chief executive's covers up to 7.08 million shares until June 2027, about $1.0bn at today's price; the technology chief's covers 1.45 million until May 2027; smaller plans cover the finance chief, the chief revenue officer and a director. Shares also vest every quarter, and the last two vesting dates, 20 August and 20 November, each brought a large sale to cover taxes. An owner should expect a steady supply of insider shares on the market for at least the next six months, whatever the price does. That supply is known and planned; it is not a signal about the business. But it is a reason why a falling price can keep falling for longer than the news would justify.

Who owns it
★ The founders
49.999999% of the votes
Through Class B and the Class F shares, on about 9.5% of the economic interest; officers and directors 50.8% of votes (2026 proxy).
Insiders
$772.5m sold in 2026
Plus ~$176m filed on 19–20 Nov; one plan to buy (Thiel entity, up to 4m shares), no purchase yet.
Analysts
Mean target ~$224
Our feed: median $246.5; Yahoo/TIKR ~$223. No downgrade found in October–November.
The company
$1.5m of buybacks
In nine months of 2026; ~$75m in 2025. Options, rights and RSUs ≈ 8% of shares.
◆
Part
IX

The numbers

A cash machine with no debt, and six faults in our feed

The quality is not in doubt. Over twelve months Palantir earned free cash flow of about $4.05bn on revenue of $7.29bn, a cash margin few software companies of any size have reached. It has no debt and $10.6bn of cash and securities. Two adjustments are needed before using its earnings. Stock-based pay, about $982m over twelve months, is a real cost; subtracting it leaves owner earnings of about $1.19 a share. And Palantir pays almost no tax (about 1% of pre-tax profit in the third quarter) because it still carries a full valuation allowance against its US and UK deferred tax assets; at a normal 23% rate, earnings would be about a fifth lower. When the allowance is released, reported earnings will jump once, and the jump should be ignored.

The multiple in history. Measured at each quarter-end since 2021, Palantir's price has ranged from 7 times trailing sales at the end of 2022 to 111 times in September 2025; the median is about 23 times. Today's 47 times is twice the median and close to where the shares stood in June 2026, when they were $107 and revenue was smaller. On earnings the picture is similar: about 100 times trailing earnings today, against 154 times in March and 93 times in June. The fall since October has taken the multiple from extreme to merely very high. It has not taken it anywhere near what investors in the best software companies have historically paid.

★ Six faults in our feed — read these before trusting a screen
What the screen showsWhat is wrong, and the number to use
Market value $331bn; P/S 45.4×Counts Class A shares only. All classes (2,409.5m): <b>$345.7bn</b> at $143.49, P/S <b>47.4×</b>. The same fault was in our October numbers' source.
Enterprise value snapshotDated 31 December 2025, with $229m of 'debt' that is lease liabilities. Palantir has <b>no debt</b>; EV about $335bn.
Net debt −$3.4bn (Q3)Ignores $6.94bn of marketable securities. Net cash is about <b>$10.56bn</b>.
Non-operating income, Q2 −$169m, Q3 −$63mWrong sign and size: Q2 was <b>+$91.8m</b>, Q3 <b>−$25.8m</b>. The same kind of fault appears in our feed for Salesforce, Broadcom and Meta.
Government revenue, FY2021: $8,974mTen times too large: <b>$897m</b>, out of revenue of $1,542m.
Analysts: 17 buy, 9 hold, 3 sellContradicts the same payload's history (24 buy or strong buy, 9 hold, 3 sell or strong sell, in November). This produced the split we quoted in October.
◆
Part
X

No dividend — and where the cash goes

A company that returns nothing, and pays its staff in shares instead

Palantir has never paid a dividend and has given no sign that it will. That is not a fault in a company growing at 90%: a business that can reinvest at very high returns should. But Palantir does not need to reinvest much. Its capital spending was $43m over twelve months, against $4.05bn of free cash flow, and the cash pile grew from $7.2bn at the end of 2025 to $10.6bn in September. The cash is not being reinvested, returned or used to offset the shares being issued to staff. It is accumulating and earning interest, $89m in the third quarter.

That is a reasonable choice for a company that values flexibility and serves governments, but an owner should see it clearly. Of every dollar of free cash flow, about 24 cents is matched by new shares issued to employees, and none is returned. An income investor has nothing here, and a value investor is being asked to accept dilution in return for growth. At 47 times sales, the growth has to be very large to make that a good trade.

Use of cash, last twelve monthsAmountRead
Free cash flow~$4.05bn✓ Exceptional
Capital spending~$43mAlmost nothing to fund
Stock-based pay (non-cash, paid in shares)~$982m◆ +85% in Q3
Buybacks / dividends$1.5m (2026) / none◆ Token
Cash and securities$7.2bn → $10.6bn (Dec → Sep)Accumulating
◆
Part
XI

★ Valuation

A quarter cheaper than in October; the 25% path is still a long way down

Price to sales, 12 months
47.4×
Was 74× in October; the median since 2021 is 23×.
Free cash flow per share
~$1.58
~91× on diluted shares; a 1.1% yield.
Owner earnings per share ◆
~$1.19
Free cash flow less stock pay: ~120×.
2027 consensus
$2.52 EPS · $13.2bn sales
57× earnings; 26× sales.
Cheaper than in October, still not cheapPrice to trailing sales at quarter-ends; dashed line: the median since 2021, 23×.23.8Dec 217Dec 2216.9Dec 2360.8Dec 2493.7Jun 25110.7Sep 2594.8Dec 2567Mar 2645.4Jun 2661.7Sep 2647.4NowWhat ten years of growth is worth today$ a share; revenue from the 2026 guide, Microsoft-like margins, 10% a year.20% a year$5725% a year (our level)$11727.6% a year (the price)$143.4932% a year$202★ $143 needs ~28% a year for ten years; our level is the 25% path.Sources: our price history and feed; Palantir filings; our arithmetic (method of our October X-Ray).

Our method, unchanged. Palantir's price can only be judged by asking how fast it must grow. As in October, we start from the 2026 revenue guide, now $8.53bn, grow it for ten years, apply net margins as high as Microsoft's (40%) and an exit multiple of 30 times earnings, allow for 2% a year of dilution, and ask what a buyer earning 10% a year could pay today.

PathRevenue in 2036Value todayvs October
20% a year~$53bn~$57$54
25% a year ◆ (our level)~$79bn~$117$112
27.6% a year (the price)~$97bn$143.49—
32% a year~$136bn~$202$193

What is priced in? At $143.49 the shares assume about 27.6% a year for ten years, revenue of roughly $97bn in 2036, about eleven times this year's, at Microsoft's margins. In October the price assumed about 32%. The fall has taken out about a third of the excess, not all of it. For comparison, the consensus expects 55% growth in 2027; even if that arrives, the following nine years would need to average about 25%.

Our level. Our rule for a business of this quality is that the level is where the price becomes fair, not where it becomes a bargain: the 25% path. Because the company raised its revenue guide, that path is now worth about $117, against $112 in October. We round it down to $115 for the bookings crack. At $115 the shares would trade at about 38 times trailing sales and 21 times next year's expected sales: still a premium price, for what is still one of the best businesses we have studied. Below the June low of $107 it would be a bargain.

How sensitive is the level? The 25% path depends on two generous assumptions: net margins of 40% in 2036 and an exit multiple of 30 times earnings. At a 35% margin the path is worth about $102; at 25 times earnings, about $98; at both, about $85. In the other direction, a business that kept growing at 25% for a decade would deserve more than 30 times at the end, and Palantir's margins are already above 40% on the adjusted measure. We keep the generous assumptions because the business has earned them. But it is worth seeing that the price of $143.49 is above the 25% path even on our generous terms, and well above it on ordinary ones.

The 20% path ($57), our level at the 25% path ($115), the price ($143), the 32% path ($202) — closer, still a pass
$57
$115
$143
$202
$40$220
★ Closer, still a pass — begin near $115. The price implies ~27.6% a year for a decade; our level is the 25% path. Analysts' mean target: ~$224. → Interactive valuation on the company page
◆
Part
XII

Risks, lawsuits & controversies

Verified 22 November 2026 against the Q3 2026 10-Q and attributed reports

Price implies ~28% a year for ten yearsBookings +6%, backlog flatFirst guided slowdown (Q4 +88%)Government 41% of revenue; new CongressStock pay +85%; large executive awardsInsider selling (~$950m in 2026 incl. notices)Index flows to AnthropicMaven reporting (allegation)Securities appeal pending (Cupat)

Legal. A securities class action (Cupat v. Palantir) was dismissed with prejudice in April 2025; the appeal was argued in March 2026 and the 10-Q of 3 November says it is still pending, and that no pending matter is material. We found no new securities suit or government investigation in 2026; notices of law-firm "investigations" are solicitations, not suits.

Controversy. Bloomberg reporting, as described by Foreign Policy Journal on 12 October, linked reliance on the Maven targeting system to a February 2026 strike on a school in Minab, Iran. Palantir's technology chief called the reporting wrong and pointed to likely human error, pending the official investigation. This is an unproven allegation; we record it because government reputation is part of Palantir's moat.

★ How we will know — six dated tests
TestTodayWhat would change our mind
Q4 results, early FebruaryContract value +6%; backlog flatBookings and backlog growing strongly again would make Q3 a pause; another flat quarter would make the 2027 consensus too high.
The first 2027 guideConsensus +55%A guide well below the consensus would test the 25% path itself.
Government funding, 11 DecemberStopgapA shutdown would delay government revenue for a quarter; not a reason to change the level.
Insider Form 4s, late November~$176m filed as proposedA purchase by the Thiel entity would be the first real insider buying signal.
Gross margin83.4%Back to 85% would say the cloud costs were a step; lower would say they are a trend.
The price$143.49Near $115 with bookings recovering: begin.
◆ PART XIII · To our shareholders
The Letter ⓘ

In October I told you Palantir was one of the finest businesses I had studied and that I would not buy it at a hundred and ninety dollars. Seven weeks later the shares are at a hundred and forty-three. It would be pleasant to say I saw it coming. I saw that the price could fall a long way on a change of mood; I did not know when, and I still do not know where it stops.

The business did nothing wrong. Revenue grew ninety-six per cent, the company raised its forecast again, and customers who already use it are spending sixty-five per cent more than a year ago. Most of the fall came from outside: an election, a stumble at Nvidia, the listing of an AI laboratory that gives investors a new way to own the same idea, and a week in which the Nasdaq lost a tenth of its value.

But one thing did go wrong, and it was the thing I worried about. The value of new contracts signed in the third quarter grew only six per cent, and the backlog did not grow at all. Management says the deals were shorter, not smaller. Perhaps. A company priced for a decade of growth cannot afford many quarters in which it signs less than it delivers, and the fourth quarter will tell us whether this was one.

At a hundred and forty-three dollars the shares still ask for growth of about twenty-eight per cent a year for ten years. I would be glad to own Palantir if it delivered twenty-five, which is still an achievement almost no company of its size has managed, and that is worth about a hundred and fifteen dollars today. The company raised its guidance, so my level rises a little with it; the bookings crack, so I round it down. I would rather pay a fair price for the best than a hopeful price for the best, and the difference here is still a quarter of the share price.

Meanwhile the executives are selling, the board is granting them large new awards, and the company returns nothing to its owners. None of that is unusual in a young technology company. All of it is a reason to insist on the price.

— The Buffett Lens. A wonderful business at a falling price is still not a bargain until the price is.

▲The Bull Case
★★ The business keeps beating — revenue +96% (nine accelerations), US commercial +148%, Rule of 40 158%, 2026 guide raised to $8.53bn; free cash flow 52% of revenue.
★ Existing customers spend more — net dollar retention 165%, top-20 customers $146m each (+76%); estimates for 2026–28 rose 5–10% while the price fell.
★ A fortress balance sheet — $10.6bn of cash and securities, no debt; and a Thiel entity's plan to buy up to 4m shares.
▼The Bear Case
★★ Bookings cracked — contract value +6%, US commercial +11%, backlog flat at $4.9bn; the Q4 guide (+88%) is the first slowdown in ten quarters.
★ Still priced for a miracle — ~47× sales, ~85× free cash flow, ~100× earnings; the price implies ~28% growth a year for ten years.
★ Owners come last — stock pay +85%, large new executive awards, $772m of insider sales in 2026 and more filed; no dividend and token buybacks; 41% of revenue from a government with a new Congress.
Closer, Still a Pass —
Begin Near $115
One of the best businesses we have studied, a quarter cheaper than in October and still priced for about 28% growth a year for a decade. ★ At $143.49 the price is above our 25% path of ~$117, within a range of ~$57 (20%) to ~$202 (32%). Begin near $115, the 25% path rounded down for the bookings crack. Q4 results in early February.
⚡The 25% path, where a buyer of one of the best businesses we know earns about 10% a year. Add the $115 price trigger to your Watchlist.
◆ The Buffett Lens · Dividend Line Research · As of 22 Nov 2026 · Price $143.49 (20 Nov close) · Follow-up to our 2 Oct 2026 X-Ray
Found this useful? Share the X-Ray.
◆ More on Palantir Technologies Inc.

Articles that take this apart

◆ Keep reading

More X-Rays from the same lens

Disclaimer: This is an editorial analysis for information and education, not investment advice, and not a recommendation to buy or sell any security. ⚠️ Price and market data are from our live data pull of 21–22 November 2026; the price used is the 20 November close. Results, guidance, contract value, backlog, customer concentration, stock-based pay, buybacks, trading plans and legal matters are from Palantir's releases, Forms 10-Q, its 2026 proxy, Forms 4 and 144 and a Schedule 13G/A; customer counts and net dollar retention for Q3, management remarks, analyst actions, the causes of market moves, government budget figures and the Maven reporting are from transcripts and press reports, as marked. The growth paths are our own arithmetic, using the method of our October X-Ray. This report corrects six details in that X-Ray. Past performance does not guarantee future results. Do your own research or consult a licensed adviser.
Dividend Line · X-Ray Analyses — written in the house methodology