PLTR

We rate Palantir Avoid: our fair value is $69 a share, 66% below the $200.51 price. The business is excellent, but the price needs 65x our 2032 profit. Even the bull case is worth only $113.

How to read the numbers. Each report values a company three ways. The central case is our best estimate, the bear case a severe but plausible bad outcome and the bull case the matching good one. The central case sets the view. The exit price at the end of 2031 is the median of three estimates. The exit multiple is a judged multiple of 2032 after-tax operating profit (NOPAT). The long fade lets growth and the return on new capital drift toward mature levels over 15 years. The outside view is what similar companies trade at today. "Stock pay" is stock-based compensation, which we charge as a cost.

How the number moved. We built the model blind, at commit d961539 (central $83, Avoid), before reading the price. Two outside reviewers then reviewed it and moved it to $69 (section 11). The largest change holds the judged exit multiple to the model's own long fade at Palantir's own cost of capital (-$17); our method requires this and the blind output already printed it. Smaller fixes to the share count, award costs and tax deductions (+$1.9) and a faster 2027 for US commercial, based on the company's deal-value record (+$1.4), offset part of it. The view is unchanged.

The call in one paragraph

Palantir is the fastest-growing large software company we cover: Q2 2026 revenue grew 93% (US commercial +149%, US government +90%) at a 62% adjusted operating margin, and the company beat its own quarterly revenue guide in each of the last eight quarters, by a median 6.8%. Our central case is not cautious about the next two years: it is above guidance and the Street for 2026 ($8.50B against the $8.15B guide and the $8.21B Street) and 2027 ($13.3B against $12.3B), and within 3% of the Street for 2028. It then slows the way large software businesses do, to 15% growth in 2031 with 12.6% left in 2032, at a 50% operating margin after stock pay. At Palantir's own 11.18% cost of capital that business is worth 18x its 2032 profit at the end of 2031 and $69 a share today. The price of $200.51 needs 65x our 2032 profit. Even the bull case (2031 revenue $39.5B, 55% margin after stock pay) is worth $113 and returns -1.2% a year. Avoid.

1. Business economics: very fast growth, cancellable contracts, no debt

Palantir sells software platforms: Gotham (government), Foundry and AIP (commercial and government) and Apollo (deployment). Customers sign multi-year contracts, most cancellable for convenience; only $4.9B of the book is non-cancellable (remaining performance obligations, RPO, at 2026-06-30, 10-Q note 3), 43% of it due within a year. The wider, cancellable book is larger: total remaining deal value (RDV) $11.2B at 2025-12-31 (+105%), plus $12.3B of unfunded IDIQ awards (indefinite-delivery, indefinite-quantity government contracts; 10-K MD&A); US commercial RDV alone reached $6.24B at 2026-06-30, +124% (Q2 release). These are management metrics that assume every option is exercised.

The four parts grow at different speeds (H1 2026 against H1 2025, our arithmetic from the releases and 10-Q note 12):

The money is simple: no debt, $9.4B of cash and Treasuries, almost no capital spending ($22M in H1 2026), working capital near zero, and a 63% adjusted free cash flow margin in Q2. Stock pay is the main claim on owners: $684M in 2025 and $467M in H1 2026 (13.1% of revenue), plus payroll tax on it. It is falling as a share of revenue only because revenue grows faster than headcount (4,429 employees at end-2025).

What limits it: a US government that is 42% of revenue and sets budgets yearly; a commercial market where model vendors and hyperscalers are building the same agent-and-data layer; and scale (we know of no software business at $8.5B of revenue that then compounded above 50% a year for five years; our inference, no base-rate study cited). Total bookings are cooling from their peak: total contract value (TCV, the value of new deals signed) grew 61% and 49% in Q1 and Q2 2026 against 151% and 138% in Q3 and Q4 2025. US commercial is not yet: its TCV grew 153% in Q2 2026 and its remaining deal value 27% sequentially, after 12% in Q1 (releases 2025-11-03 to 2026-08-03).

2. Results against the bar: eight straight beats of guidance

(Releases; all eight quarters in output.txt.) Q2 came with the largest full-year raise in the company's history: midpoint $7.66B to $8.15B (+82%). The stock rose 40% that week (weekly close $123.06 on 2026-07-31 to $172.01 on 2026-08-07, Robinhood) against 5.1% for QQQ (a Nasdaq-100 fund), and 17% more since, to $200.51. We did not check other news that week.

Q3 2026 reports on 2026-11-02 (company release of 2026-10-12). Our forecast: revenue $2,290M (guide $2,160-2,164M; Street $2,176M, TradingView), US commercial $975M, adjusted operating income $1,390M (guide $1,292-1,296M).

3. Our forecast gives $69; the fade and the multiple agree, the outside view is higher

The model is model.py (output in output.txt); the blind inputs and their sources are in assumptions.md, and the changes since are in section 11.

• Stock pay and dilution, counted once. The share count is 2,585.9M: 2,402.9M shares plus every option (149.7M, proceeds of $1,506M at the $10.06 weighted strike added to cash), restricted stock unit (RSU) and performance RSU (P-RSU) outstanding at 2026-06-30. The 13.7M stock appreciation rights (SARs; weighted strike $131.61, gain capped, settled in shares only above a price threshold at expiry) are out of the money at our values and are left out; their cost stays in the stock-pay expense. Future stock pay is charged as a cash cost, net of the $916M of RSU and $361M of option expense not yet recognised for awards already in the count. The tax saved when options are exercised and RSUs vest above their cost (23% of the spread at our value) is added: $2.3B, about $0.9 a share.
• Cash at its own yield. Palantir pays no dividend and states no payout, so free cash is kept. Each year's free cash is valued when it is earned and today's $9.4B at face; neither is held at a deposit rate and then discounted at the cost of equity.
• Cost of capital. 5.10% Treasury (2026-10-15) + 1.45 beta (median Blume beta, a beta adjusted toward 1, of eight high-growth software peers) × 4.20% equity risk premium (ERP; Damodaran, 2026-10-01) = 11.18%, for the holding period and for the exit. At Palantir's own Blume beta of 1.72 (12.34%) fair value is $59 and the purchase price $60. At the mature-peer rate of 10.27% for the exit (the blind choice, dropped in review) fair value would be $76.

The outside view is 43% above the median. The input behind the gap is the discount rate: the peers' 31.6x prices mature software at the market's rate, well below the 11.18% we charge Palantir, whose own beta is higher still. We believe the fade, which prices this case's growth at this company's cost of capital; the outside view is the market's price for the same profit and is the case for the other side. In the bear case the outside view ($76) is 73% above the median ($44) for the same reason plus the bear's lower growth (6.7% at exit against the peers' 10-17%).

Cases. Bear $28: US commercial growth decays to 10% by 2031 and US government to 5% (model vendors commoditise the platform layer; a defence budget turn after 2028; our assumption), margin after stock pay 38%, 13.5x exit (its fade). Bull $113: US commercial +90% in 2027 decaying to 22%, revenue $39.5B in 2031, margin after stock pay 55%, 21.0x exit (its fade). Weighted 25/50/25 (our weights): $70.

Sensitivity (central fair value): NOPAT ×0.8 $56, ×1.2 $82; exit multiple 20x $75, 24x $86, 28x $98 (each a fixed multiple in place of the fade; the median then moves with it); mature return on new capital 30% $68, 80% $69. Each line recomputes the fade at its own inputs.

4. The price assumes $823B of enterprise value; the Street stops at 2028

unblinded.py (output in unblinded-output.txt) reverse-underwrites $200.51 with the model's cash flows at 11.18%:

• an end-2031 enterprise value of $823B, against our $229B;
• on our path, 64.6x 2032 NOPAT; on the bull path, 41.4x;
• at our 18x and our 50% margin after stock pay, 2031 revenue of about $105B, 65% a year from 2026 (ours 28%).

Against the Street: FMP consensus revenue (fetched after the blind commit) is 2026 $8.21B (23 analysts), 2027 $12.33B (23) and 2028 $18.05B (17). Our central is +4%, +8% and -3% against those. FMP also lists 2029 ($33.4B) and 2030 ($70.1B), but those rows are not estimates: their low-to-average and high-to-average ratios (0.9480 and 1.0891) repeat 2028's exactly, and their EPS ratios repeat each other (found by the model reviewer; we confirmed it). There is no real Street figure past 2028 to compare with. On the Street's revenue to 2028, then 33%, 25%, 20% and 16% (the model reviewer's path), the model gives $94 at the fade (20.9x), $106 at 24x and $117 at 28x.

Capfolio's ranking model (Alpha) says (data/alpha-rank.json, 2026-10-15): rank 32 of 713, score 95.7, tier Strong; growth A, quality A, momentum B, expectations B, valuation E (684th of 713). Alpha ranks for a 126-day horizon on quality, growth and momentum; it agrees the business is excellent and says the price is among the most expensive in its universe. We differ in the question, not the facts: this report asks what the shares are worth.

5. We agree with the market on the next two years and disagree on the five after

The market's thesis: Palantir is the operating system for AI in governments and large companies, revenue will compound near 50% for most of a decade, and today's 125x current-year Street EPS ($1.60, TradingView) is a waystation. Ours: the next two years are as strong as the market thinks (we are above the Street), but the price needs about 65% a year for five years at our exit multiple, or about 41x at exit on our bull path. What settles it is the decay of US commercial growth through 2027: our central has 2027 revenue at $13.3B (+56%); the FY2027 guide in early 2027 and the quarters after it show which path is real.

6. Why the other side could be right

• The deal-value record. In four quarters we can check, the year-on-year growth of US commercial deal value has matched the next year's US commercial revenue growth within 0.9-1.4x (bull-side reviewer, from the releases). The latest reading, +124%, points to US commercial growth of 110% or more through mid-2027; our central has about 79% for 2027. At +100% for 2027 the central is $73.
• The guide record and margins. Profit beat guidance by a median 14.9% over eight quarters, against 6.8% for revenue; H1 2026 incremental margin was 79% (bull-side reviewer). A flat 63% adjusted margin is worth $73 (+$3.8).
• Government budgets. Department of War revenue is "less than 25 basis points" of the Pentagon budget (management, call 2026-08-03); doubling that share would add up to about $2.5B a year, about 9% of our 2031 revenue (our inference from the quote).
• The multiple. The market pays 30x and more for mature software today (the outside view, $162 at exit). If the market's lower discount rate is right and ours is too high, the exit is worth more: at 28x the central is $98, and on the Street's path to 2028 at 28x it is $117. Nothing we can find gets to the price without both the Street's path and a 40x-plus exit.

The single assumption that would flip the call is growth left at year five: a Palantir still growing 30% or more in 2032 would earn 40x or more and a value near the price. Nothing dated supports that today.

7. Catalysts and what would prove us wrong

• 2026-11-02 (company-announced), Q3 2026: revenue against our $2,290M (range $2,240-2,340M), US commercial against $975M, adjusted operating income against $1,390M.
• Early February 2027 (not yet announced; Q4 releases were filed 2026-02-02 and 2025-02-03): the FY2027 guide. An initial guide at or above $12.7B (+50% on our 2026) confirms our central; it does not reach the price.
• What would prove us wrong: FY2027 revenue above $15B with US commercial RDV still growing 100%+: re-underwrite toward the bull. US commercial sequential growth below 10% in two quarters, or a major program cancelled for convenience: toward the bear ($28). Stock pay held at 15% of revenue through 2031 cuts NOPAT about 10%, about $7 a share (the NOPAT ×0.9 line).

8. Concerns the review checklist raised, and how they came out

A checklist review of the work raised or cleared these points.

• The move from the blind $83 to $69 (-$14, -17%), a medium concern. The largest change (-$17) holds the judged multiple to the model's own fade at the company's own cost of capital, on the method's rule and the blind output's own printed fade, not on a management claim or Street figure. It moves away from the price.
• The value is 66% below the price, a medium concern.
• The outside view is 43% above the median in the central case and 73% in the bear, a low concern; the discount rate is the input behind it (section 3).
• The central case against the Street, a medium concern, restated: within 8% of the real Street through 2028; FMP's 2029-30 rows are synthetic and are not compared. Above guidance for 2026, sourced by the eight-quarter guide record.
• The peer beta (1.45) over Palantir's own (1.72), a medium concern: fair value $69 against $59. The mature-peer exit rate, the other lenient input, was dropped in review.
• Not raised: the view is unchanged; the purchase price is $67 against $60 at own beta, 11% (raised as low; disclosed here); all in-the-money awards are in the count, and the 10-Q shows no convertible, offering or at-the-market program; the 10-K, 10-Q and nine releases were read; the cost of equity is below the hurdle; the reviewers moved the central 17%, under 20%; the bridge items are balance-sheet and note lines.

9. Sources (all read 2026-10-15)

• SEC EDGAR: 10-K FY2025 (0001321655-26-000011), 10-Q Q2 2026 (0001321655-26-000041), earnings releases (8-K exhibit 99.1) Q2 2024 to Q2 2026, saved in data/ with price-revealing figures masked; the 10-Q stock-pay note unmasked after the blind commit, data/10q-2026-06-30-note9-filing.txt.
• Q2 2026 call passages (Quartr via a Bigdata.com search, call 2026-08-03), data/bigdata-call-excerpts-q2-26.txt; the Q3 date from the company's 2026-10-12 release (wire reprint), data/q3-26-earnings-date-source.txt.
• FMP: statements, segment revenue, peer ratios and prices for betas, and consensus estimates (after the blind commit).
• US Treasury par yield curve (2026-10-15); Damodaran implied ERP (2026-10-01).
• Robinhood: last trade, official 2026-10-14 close ($200.97), weekly bars, analyst ratings. TradingView: forecasts. Alpha: alpha rank PLTR.

10. What went wrong in preparing this report

• Our masking of the SEC filings missed table cells where $ and the number sit in separate cells, the stock-pay note's award prices and option intrinsic values, and the equity statement's repurchase rows (shares and dollars give the price). We read them before fixing the mask, which made this report partly unblinded. The fix drops the whole stock-pay note and the repurchase rows; it is recorded in our tooling notes (tools.md).
• Dropping the stock-pay note also hid the options' weighted strike, and we substituted the exercised options' $4.72 for the outstanding $10.06; a reviewer caught it. Next time keep the note's share counts, strikes of out-of-the-money awards and unrecognised expense in a separate file read only after the blind commit, or mask only the price columns.
• TradingView and Robinhood sign-ins lapsed mid-task and came back. The 10-year yield came from home.treasury.gov's CSV instead, first time.
• A helper agent made two Bigdata.com searches: the Q3 date and six call passages, nothing on stock pay, headcount or the Army agreement.
• Before the blind commit we moved the exit to the mature peers' discount rate when the first run's 28x lacked support from the fade. That was the wrong fix: the method values the exit at the company's own cost of capital, and the lesson from an earlier report on another software company (NOW) was to cut the multiple, not to find a rate that supports it. The review reversed it (-$17).
• FMP's 2029-30 consensus rows are scaled copies of 2028; recorded in tools.md.

11. How the two reviewers moved the value from $83 to $69

One round, two reviewers: a model reviewer who checks the model and a bull-side reviewer asked to argue the other side. Effects are measured in order on the model, central fair value per share (a scratch script run against model.py; each step on top of the last).

Net: $82.62 blind to $68.91 (-$13.71). Purchase $79.89 to $66.70; exit $137.10 to $112.70; bear $29.44 to $28.27; bull $137.81 to $113.17. The view stays Avoid on every line of the review: the bull-side reviewer's two changes at the levels proposed give $73 each, and the highest value any reviewer path reaches, the model reviewer's Street path at 28x, is $117, 42% below $200.51.