Palantir (PLTR) — Ninety-Six and Six

Palantir's revenue grew 96% in the third quarter, the fastest of its public life, and net dollar retention reached 165%. The contract value it closed grew 6%, against a comparison that had grown 151%, on deals the company says ran shorter than usual. We value the stock at $145.70, from $149 in August, against $178.34 at the close before the release. August's method, fed this quarter's data, would say $224.90; we explain why we no longer rest the value on it. At our margins and rates, the price takes about $136 billion of revenue in 2036, a decade about as fast as AWS's and Meta's from the same size, the best we have found in software and internet.

TON618 Capital Research · As of 2 November 2026 · Download PDF · Download model

In August we drew a fade line: the rate at which Palantir's growth could decay before the price stopped making sense. We also wrote that every leading indicator still outran revenue. Three months later revenue has run past nearly all of them. US commercial sales grew 148% while US commercial contracts still to be delivered grew 79%, a gap already open in August and wider now. That may say less than it appears to: the comparison is hard, the deals are shorter, and customers are spending far more than they did. This note sorts out how much of the gap is real, then asks what a business growing this fast is worth.

Terms used throughout, defined once. TCV (total contract value): the lifetime value of contracts closed in the period, assuming every option is exercised and nothing is terminated; the company says most of its contracts can be terminated. RDV (remaining deal value): the same measure for what is still undelivered at quarter end. RPO (remaining performance obligations): non-cancelable contracted revenue not yet recognized, leaving out contracts of twelve months or less and most government work. Net dollar retention: twelve-month revenue from the customers of the prior twelve months, over that prior revenue. Adjusted: excluding stock compensation and its payroll taxes. Free cash flow to the firm: after-tax operating income less the capital growth requires, with stock compensation an expense. ROIC: after-tax operating income over capital, tested against the weighted average cost of capital (WACC). Fade factor: the share of its growth rate a business keeps each year, August's single input. Terminal growth: the rate the business grows at forever after the forecast ends.

Revenue was $2.313 billion, up 96% and 20% on the quarter. US commercial was $984 million (+148%, +29% on the quarter) and US government $953 million (+96%, +18%). International was $376 million (+26%): commercial $184 million, +21% and only +1% on the quarter, and government $192 million, +31%. Customers were 1,090, +20%, and US commercial customers 686, +29%. The growth is American and concentrated: the US is 84% of revenue, and the top twenty customers, at $146 million each, are about 40% of trailing revenue. No single customer exceeds 10% of revenue; one accounts for 27% of receivables.

The guide calls for $2.649–2.653 billion in the fourth quarter, about +88%, and $8.530–8.534 billion for the year, +91%, from $8.154 billion in August. Revenue has run 5.9%, 6.4%, 7.6% and 7.0% above the guide's midpoint in the last four quarters, so we carry $2.785 billion for the fourth quarter, about 5% above it, and $8.67 billion for the year.

GAAP operating income was $1.085 billion, a 47% margin (33% a year ago), and each added dollar of revenue over the year brought 61 cents of added operating income.

Why the two figures are both $0.44. Adjusted earnings add back $319 million of stock compensation and $23 million of related payroll tax, which should put them well above GAAP. They end only $14 million higher because the company then subtracts $328 million for tax at an estimated long-term rate of 23.0%, while the GAAP line paid 1.2%: Palantir still carries a full valuation allowance and federal net operating losses that stood at $9.0 billion at the end of 2025. The match is a coincidence of those two lines. At 23% on GAAP pre-tax income, earnings per share would be about $0.34.

The cash. Cash and Treasuries were $10.38 billion, with no debt and no active buyback; the old $1 billion program was terminated in January 2026. At $178.34 that costs holders little: spending $10 billion on stock at the close would lower our fair value by about $0.73 a share.

Palantir closed $2.918 billion of TCV, up 6%, and $1.463 billion in US commercial, up 11%. Total RDV was $13.4 billion, +55% and +2% on the quarter. US commercial RDV was $6.517 billion, +79% and +4% on the quarter, the smallest sequential rise in the five quarters we show. RPO was $4.88 billion, +88% and flat on the quarter. Each of those growth rates is now below the revenue it feeds; only billings, +100%, still outruns it. August's claim about leading indicators no longer holds.

Before reading anything into that, three boring checks. Each explains part of the gap.

The comparison. The year-ago quarter's TCV of $2.76 billion was itself +151%, and US commercial's $1.31 billion was +342%. Over twelve months TCV was $12.96 billion, +56% (the prior window partly derived from reported growth rates), and US commercial TCV $6.114 billion, +62%. That is much better than +6%, so the comparison explains most of the headline. It does not explain the full year: revenue grew 87% and US commercial revenue 143%, and TCV per dollar of revenue fell to 1.78 from 2.13.

Lumpiness. Quarterly TCV has run $2.76, $4.26, $2.41, $3.37 and $2.92 billion, and one large deal moves it: the second quarter included a three-year deal of nearly $370 million. The third quarter's 13% sequential fall is well inside that range. Deal counts rose far more slowly than revenue: 64 deals of at least $10 million, up from 53 a year ago and down from 73 last quarter.

Contract length. This is the company's own explanation and the hardest to size, because the figure that would size it is missing. In the second quarter the company said TCV on a dollar-weighted duration basis grew 129%, against 49% as reported. For the third quarter neither the release, the deck, the call nor the 10-Q gives that figure. With the second quarter's gap, +6% would be about +63%: strong, and below revenue's +96%. But the company says this quarter's deals ran shorter than usual, so that gap may understate it: duration-basis growth would match revenue if length shortened about 46% on the year, against about 35% in the second quarter. That is our illustration, not a company number. The filings fit the company's account without proving it: long-term RPO fell to $2.74 billion from $2.81 billion, and the share due beyond 36 months went from 25% in March to 21% in June and 19% in September, though older contracts rolling closer account for some of that. No analyst on the call asked about TCV, RDV or contract length.

Shorter contracts cut both ways. They lower TCV without lowering annual demand, as the company says. They also lower what is locked in. A year ago US commercial RDV covered 2.29 years of the quarter's annualized revenue; now it covers 1.66, from 2.04 three months ago. Most of the contracts can be terminated, so a customer buying a shorter first contract has more room to expand, as Karp describes, and more room to stop.

What retention shows. Net dollar retention of 165% is the strongest fact for the price. By the company's definition, customers of the prior twelve months produced about 75% of the last twelve months' growth. The 10-Q's narrower measure agrees: customers on the books at the end of 2025 produced $891 million of the quarter's $1.132 billion increase in revenue, 79%, against 72% a year ago, and commercial customers added since produced $212 million, against $107 million, though total revenue also about doubled. That is what "much quicker to deployment" and "an ability to expand that contract dramatically" would look like. But both measures report revenue already earned. Neither can tell us what customers have signed for next year, which is the question bookings raise.

What would settle it. US commercial RDV over the next two quarters. Its sequential rise over the past five has been 30%, 21%, 12%, 27% and now 4%. Short deals come back fast, so if contract length is the reason, renewals and expansions should show up soon. Two readings of 15% or more would settle the question for the company; two of 5% or less, with twelve-month bookings still growing more slowly than revenue, would settle it the other way. One quarter cannot, so the slowdown enters our model only through the near-term path (§4).

August modeled a flat 40% free cash flow margin, and against 47% GAAP, 62% adjusted and 53% free cash flow that looks low. It was not, once two costs are counted.

Stock pay is the first cost. It was $319 million in the quarter, 13.8% of revenue, up 85% on the year. We expense it in every year and hold the diluted count at 2,573.9 million. August's 40% was already close to free cash flow after stock pay (38% this quarter), and it also grew the share count 2% a year, which charged the same cost twice. Removing that is the largest method change in the company's favor in our bridge.

Tax is the second. The company pays about 1% and tells investors to think about 23%. The losses that shelter it are real and finite. We carry 4% in 2027, 9%, 15% and 20%, reaching 23% from 2031; 23% from 2027 would cost only about $1.65 a share.

The path. GAAP operating margin rises to 52% in the base case, because costs are growing more slowly than revenue: adjusted expense rose 53% on the year against revenue's 96%. It holds through 2030, then falls to 46% in 2036 and 42% by 2041, landing with growth (bull 48% in 2041; bear 33% by 2036). Microsoft earned 46.8% in its latest fiscal year, Meta 41.4%, Adobe 36.6%, Oracle 30.6%, Salesforce 20.1% and ServiceNow 13.7%. A business as light as Palantir earns a place in the upper part of that range. The fall from 52% reflects the selling and service that a company ten times larger is likely to need, and competition: Karp says the company rarely competes with the AI labs for customers today, and adds, "I don't know at what point we become competitive with each other." After tax the base case ends at 32% of revenue, against 36% today at full tax.

Margin is not the main question. August said this was a growth-priced stock, and the test holds. On August's engine at today's price, margins of 35%, 45% and 55% move the demanded ten-year growth from 33.6% to 29.8% and 26.8%. In the rebuilt model five points of margin in every year moves fair value about $16, and the 2036 revenue the price requires from $122 billion to $154 billion at five points either way; two more years of faster US commercial growth moves it about $48. The stock's value turns on how long growth lasts.

August ran one number for the whole company: start from the guide's growth and multiply by a fade factor each year. That no longer works as the model. With this quarter's data the seed moves from 82% to 91% and trailing revenue rises 18%, which lifts August's fair value by $75.83, or 51%, because an acceleration in one quarter compounds through all ten years. The base fade of x0.78 now reaches $153 billion by year ten, against the $102 billion it reached in August, and the bear fade of x0.70 reaches $78 billion, more than Netflix earned in 2025. So we ask instead what drives each business, and check the answer against what has happened at this size.

What the record says. Palantir arrives at about $9 billion of revenue growing above 90%. Alphabet passed that size in 2006 growing 73% and compounded at 24% for ten years (28% from 2005, when it was growing 92%). AWS reached it in 2015 growing 70% and compounded at 32%. Meta reached it in 2013 growing 55% and compounded at 33%. Salesforce, Netflix and Adobe arrived growing about 25% to 30% and compounded at 19%, 20% and 16% over the eight or nine years they have reported since. We leave out Nvidia's 47% for eight years, earned as a chip maker on data-center spending; even our bull sits below it. Persistence like this is rare: across US companies from 1951 to 1998, sales growth stayed above the median ten years running for 0.9% of firms (Chan, Karceski and Lakonishok, 2003). And none of these had landed. In the latest year shown they were still growing: Alphabet 20%, AWS 20%, Meta 16%, Netflix 16%, Adobe 11% and Salesforce 10%. A compound rate means little without the growth still running at the end of it.

How long each part of the business grows faster than normal. Each stage's length comes from what drives it.

The 2,000 is our estimate, with no published count behind it, so we check the base's US commercial path against spend as well. About 720 customers at the end of 2026 reach about 1,750 by 2032 and 1,950 by 2036, leveling off just under that limit. Revenue per customer runs from about $5.6 million in 2026 to $23 million in 2032 and $37 million in 2036, and its annual growth falls from 65% in 2027 to 15% and 11%. Today's top twenty customers average $146 million across the whole company.

The three cases. The base grows +76% in 2027 and compounds at 28% a year to $100 billion in 2036, still growing 11%, and ends at $141 billion in 2041. Alphabet from 2005 is the closest match on entry growth, at about 70% of the size: it compounded at 28% and was growing 14% in its tenth year. AWS and Meta did more. The 2036 figure is close to August's published $102 billion, reached from a higher 2027 and a slower middle. The bull (+90% in 2027, 36% a year, $182 billion, still growing 16%) is the case where the company's claims about Sovereign AI are right and bookings recover: it runs ahead of AWS and Meta and matches Meta's tenth-year growth. The bear (+50% in 2027, 19% a year, $48 billion) has the contract slowdown as real and AI budgets digesting what they bought, at a pace near Salesforce's and Netflix's, and it still leaves a business more than five times its present size. It is the one case we land by 2036, with growth at 3.5%, below where those two were, because the precedents are all survivors and this case also stands in for the outcomes they cannot show. We apply that reasoning to it alone. On the bull and base schedule it would be worth about $56, and fair value about $147.55; with growth stopped at 3% after 2029, about $33 and $141.95.

The first year. This is where the slower order book enters the model. The base holds dollars added at about $412 million a quarter through 2027, between the $378 million added in the third quarter and the $472 million in our fourth-quarter estimate. They ran $226, $226, $303 and $378 million over the last four quarters; with US commercial deal value up only 4% on the quarter we do not assume they keep rising. Sequential growth therefore falls from about 15% to about 10% over the year. The bull runs about 16% a quarter, near the guide's 15%, and the bear about 6%. A steady 15% a quarter through 2027 would lift the base case to about $134.

The landing. The bull and base reach long-run growth in 2041, and margins and the cost of equity land with them. The date is a judgment from the record above: the slowest precedents were near 10% after eight or nine years, the base reaches that rate in 2036–37, and it takes about five more years to reach GDP. Landing the bull and base in 2038 gives $133.72, and in 2036 $118.75. The US commercial stage matters more. Two more years in the faster stage in every case, which also moves the landing to 2043, lift fair value to $193.82, above the price, and two fewer cut it to $115.34. The base case alone reaches the price if that stage runs three years longer, with its growth held through 2034; the forecast then ends in 2044 at $271 billion.

August's two-quarter rule. August said two more quarters at 80% or better would re-weight the bull upward. Growth ran 96% in the first, and the +88% guide implies the second. We retire the rule before it completes; that is our judgment, not its outcome. A quarter's growth comes from contracts signed earlier and is already in the base year and the 2027 path, so counting it again in the weights is the same evidence twice. The same test applies to the order book: its slowdown is in the 2027–28 path, so we do not count it again in the weights or the later years.

The value. We discount free cash flow to the firm in three scenarios, 2027 to 2041, by business, with stock compensation expensed, mid-year from 2 November, the fourth quarter of 2026 as a stub, $10.44 billion of cash and securities net of minority interests added once, and 2,573.9 million diluted shares.

What the price requires. At $178.34 the enterprise is worth about $449 billion on diluted shares, 61.5 times trailing revenue and 29.4 times our 2027 base revenue. Holding the other base inputs, one at a time, the price takes one of three things. Growth about 1.14 times the base case in every year, a 31.7% compound rate to $136 billion in 2036, still growing 13% that year, just below AWS's 32.2% and Meta's 32.9%. Or costs of equity about 1.4 points below ours, 10.0% and 8.6%. Or about 16 points more operating margin in every year, which has no precedent. Combinations get there sooner: August's weights with two more years of faster US commercial growth give about $221. Of the three cases, only the bull is above the price.

Terminal value is 50% of the base case's operating value (56% in the bull, 38% in the bear), or 16.6 times 2041 after-tax operating income at the end of that year. As a rough check, the median trailing price-to-earnings multiple of Microsoft, Meta, Salesforce, Oracle, Netflix, Alphabet, Adobe and ServiceNow is 19.7 today, an equity multiple on reported earnings rather than an enterprise multiple on operating income taxed at 23%, and those businesses are growing faster than 4.5%. On our measure a 20-times exit in every case would give $156.83, and 25 times $178.03.

From $149 to $145.70. The small net move hides large ones. We first re-ran August's configuration, which reproduces $148.94, then applied one change at a time, so the lines sum to the move.

The quarter's data on August's method take the value to $224.90. Mid-year discounting adds $10.82, and expensing stock pay without the 2% share growth adds $51.62 (§3). A GAAP margin path, a tax path and capital tied to revenue take $33.12: the near years are above 40%, the later ones below, and the later ones carry more.

Stages by business in place of one fade take $77.79, the largest line and the one to challenge first. The change of form adds value: stages set to the fade's 2036 revenues and running on to 2041 would be worth about $308, against $254 before this line. The fall is a lower revenue path. About $111 comes from returning the cases to the 2036 revenues August published ($215, $102 and $55 billion). The other $21, about $19 of it the bull, is a judgment about the range, not a reading of this quarter: August's bull took 38% a year from here, and ours, at 36%, still exceeds AWS and Meta, so at a 25% weight we go no further. August's bear took 20%, about level with Netflix and Salesforce; at 19% ours does not beat either.

Terminal growth from the Federal Reserve's longer-run projections of 16 September 2026, 2.0% real and 2.0% inflation, gives 4.0% in the base and bull and 3.0% in the bear, a share-loss case, against August's 3.5%: +$7.32. Charging reinvestment in the terminal value takes $3.20.

The cost of equity: 11.4% to 2030 (−$39.01), 10.0% by 2041 (+$22.05). August used 10% flat. The 10-year Treasury was 4.94% on 30 October, against 4.63% when August's note went out. Palantir's own 60-month beta is 1.56, with a standard error of 0.53 and a correlation of only 0.36, and it reaches back to a different business; alone it gives 12.0%. Blended toward the 1.20 median of ServiceNow, Snowflake, Salesforce, Datadog and Microsoft it gives 1.37 and 11.1%, with 53% of the weight on slower, shorter-duration software. We use the 36-month beta, 1.44, which covers the AIP business since 2023, for 11.4% through 2030. The rate then falls in a straight line to 10.0% by 2041 as growth lands. In 2036 it is 10.6%, close to Salesforce's 10.4% today at about the same growth. The end point is a beta of about 1.12, between the mature peers (Microsoft, ServiceNow and Salesforce average 1.07, or 9.8%) and the all-peer median (10.3%). Those two ends are worth $149.18 and $140.88; the blended read $143.81; August's flat 10% $160.64; and 11.4% with no convergence $126.48.

Return on capital. Palantir has no debt: market equity is about $429 billion on basic shares and operating leases under $0.3 billion (our estimate: $225 million noncurrent on the balance sheet, plus a current portion the 10-Q does not break out, $46 million at December), so the WACC of 11.4% and the cost of equity are nearly the same. The capital definition is the call. The conventional measure, equity plus debt less cash, averaged about $0.3 billion over the last year and has been negative in some periods, so its ratio says nothing. We leave the cash in, use average total equity, and tax operating income at 23% rather than the reported 1%. On that basis the return is 28.4%, up from 11.6% a year ago, and the regime is strength: profit is growing faster than a capital base that is mostly the cash pile.

What the ratio cannot show. The return on the next dollar of operating investment, because that investment runs through the income statement as engineers and stock pay. Capital added since December 2023 earned 41.7%, but that mostly divides profit growth by retained cash, so we do not set the terminal value from it. The forecast years charge $0.09 of capital per added dollar of revenue, about 1% of after-tax profit by 2041. A 40% return would reinvest a tenth of profit forever on top of the expenses already in the margin. We assume 100% in the base (bear 50%, bull 150%), which reinvests 4% of after-tax profit, more than the forecast years' charge, since the margin is held forever and a business of this size may need more capital than this one does. No terminal reinvestment gives $148.33; 40% in the base (bear 25%, bull 60%), $141.88; 25% everywhere, $136.30. A return equal to the 10.0% terminal cost of equity in every case, no excess return on new investment forever, gives $118.27; we do not use it because the margin already falls to the mature range and carries the investment that a return that low would charge a second time.

Our August notes left five tests open, plus one we closed in September, and made claims this print can check. The print cleared every line. The two price tests were not tripped, and there is no buyback program to test. We correct our own record on retention and customer count. The one August claim the print reverses is the one about leading indicators.

Each tripwire has its reading rule set now, so the next update scores it rather than reinterprets it. Effects are on fair value at today's other inputs.

1. US commercial RDV (about 1 February and early May 2027). At +15% or more on the quarter in both, with US commercial revenue on or above the base path, the duration account holds, and US commercial's 2027–28 path moves to the bull's pace in the base case (about $153.34, from $132.56). At +5% or less in both, with twelve-month US commercial TCV still growing more slowly than revenue, the bear's 2027–28 pace becomes the base (about $107.63). Anything between: no change, and the reading is recorded. A company figure for TCV on a duration basis, if it returns, is scored: above revenue growth in both readings, the duration account holds whatever RDV reads; below it in both, it does not.
2. The 2027 guide (Q4 release). The initial guide runs well below the outcome: 2025 was first guided at +31% and ended at +56%, and 2026 was first guided at +61% and is now guided at +91%. Allowing for the smaller gap, 25 points, the halfway marks between our cases correspond to an initial guide of about +38% and +58%. Between them, the base's +76% for 2027 holds. Below +38%, the bear's pace. Above +58%, the bull's.
3. Quarterly revenue. The base path is about $2.79 billion in Q4-26, $3.20 billion in Q1-27, $3.61 billion in Q2 and $4.02 billion in Q3 (+98%, +96%, +86% and +74% on a year earlier). Two straight quarters more than 5% below move the base toward the bear; two more than 5% above move it toward the bull. This replaces August's +70% line, which our own base case crosses in the fourth quarter of 2027.
4. Margin (each release). GAAP operating margin of 52% or better in two consecutive quarters of 2027 adds five points to every case in every year, about $161. Below 44% in two takes five points out, about $130.
5. Retention (each call). Below 150% by the second quarter of 2027, with customer growth below 20%, confirms expansion has slowed, and the base's 2028 US commercial growth moves toward the bear's. At or above 165% on the same date, with customer growth at 20% or more, it moves toward the bull's.
6. The cash. A buyback program, or any acquisition above $1 billion, is scored on the price paid against fair value.
7. Price. A close at or below $145.70 resolves the gap. $109 is a 25% margin of safety to fair value. Replaces $112 and $149.

Palantir has reported the best quarter of its public life, and at $178.34 the stock already prices a decade about as fast as the best we have found in software and internet. We value it at $145.70, from $149. August's method, fed the new data, says $224.90; we no longer rest on it because it lets one quarter compound for a decade. A base case close to Alphabet's decade from 2005 is worth $132.56. At our margins and rates the price takes about $136 billion of revenue in 2036, or two more years of faster US commercial growth than we assume, and our bull case, $269.43, clears it. Which is closer should show first in the order book: US commercial remaining deal value over the next two quarters. The Fund holds no position in PLTR.