Palantir's guide raise tilts bull, but share-count creep keeps the spread thin

Palantir Technologies Inc. is a data-software platform that runs two businesses under one roof: a U.S. government franchise growing on the back of programs like Maven, and a U.S. commercial franchise growing on the back of Artificial Intelligence Platform adoption by enterprises that want to deploy large language models against their own operational data. The bear framing writes the headline multiple and stops there, with trailing earnings elevated against larger software peers; on a stand-alone basis the multiple looks indefensible.

The sourced record supports a different claim: trailing-twelve-month revenue reached $6.2 billion with growth of 56.2%, and the Q2 2026 results showed year-over-year acceleration plus a full-year guide raised by approximately $500 million — the largest single raise in the company's history on a larger base. U.S. commercial grew 149% in the quarter. The setup the consensus multiple debate misses is that the share of contracted revenue already on the balance sheet is short-duration relative to peers, the dilution math is moving the wrong way, and the customer-concentration footprint is thinner than the headline backlog implies.

Setup ignores share-count creep behind the headline beat

Price performance vs peers (Software - Infrastructure and cross-industry comps)

Palantir shares closed at $176.24 on September 17, sitting about 14.9% below the 52-week closing high of $207.18 after a 4.7% twelve-month move, and short interest sits at 2.7% of shares outstanding — modest crowding either way. The narrative the tape has absorbed is the one Karp has leaned into: sovereign AI re-rates the multiple, the U.S. commercial step-up is structural, and the rule-of-40 print of 155 in the latest results release justifies a premium that no peer in the file can match on the same denominator. The stock has already given back ground from the peak, and the consensus framing still leans on growth versus a multiple that reads 147.2× on trailing earnings: extreme even next to a software peer set that itself sits at premium levels.

The filings push back on that read in two places the multiple debate does not reach. Stock-based compensation was $265.2 million in Q2 2026, up 66% year over year, with sales-and-marketing SBC alone doubling; SBC fell only 1.1% in FY2025 while revenue rose 56.2%, but the Q2 step-up reversed that discipline. The diluted share count has barely moved quarter-over-quarter, so the more useful read on equity is the SBC growth covered below. Adjusted diluted EPS of $0.41 in Q2 looks clean against that denominator only because the buyback book is closed; the Q1 call attributed dilution entirely to vesting and exercise, not issuance, and no repurchase program offset it.

The customer-concentration footprint thins the headline backlog. RPO of $4.9 billion at June 30, 2026 is split 43% within twelve months, 36% in months thirteen to thirty-six, and Palantir disclosed that many customers can terminate for convenience on less than twelve months' notice. Total remaining deal value runs $13.1 billion, which is roughly 2.7× the RPO; Karp and Sankar used that gap on the Q2 2026 call to argue the contracted book is understated, but it also means most of the visible backlog is cancellable. The September 10 NVIDIA partnership and the renewed Fujitsu agreement that Karp referenced extend the runway on the commercial side, yet neither converts to revenue until customers commit, and the customer count grew to 1,049 in Q2 from 954 at year-end, a thin base of expansion dollars against a $414 billion market cap.

The peer comparison below sets the multiple debate on a single page:

Valuation vs peers (Software - Infrastructure and cross-industry comps)

Karp opened the Q2 2026 call with the same demand-led posture he carried on the Q1 call — both filings sit at a net tone score of 92, and hedging barely moved (12 to 8 on a 0-to-100 scale). The continuity is the point: the gap between the two calls is in what he is willing to promise. On the Q1 call Karp said Palantir is at the limit doing the U.S. work it has; on the Q2 call he said he is now driving the business to grow at or above the U.S. commercial rate for the next 18 months.

Glazer called the latest guide the company's largest-ever full-year revenue raise, layered on top of the largest-ever raise the prior quarter. U.S. commercial alone was lifted from a 120% growth floor to 134%, on a base that already printed 149% in the quarter per the results release. The fresh themes that emerged on the Q2 call. U.S. revenue growth and customer momentum, AI sovereignty versus lab dependency, bench-making over bench-maxing: are commercial in orientation, while the prior call's government-leaning themes (Civilian Government Contracts, U.S. Government / Defense) fell off. Themes that did recur, AIP capabilities and profitability, came in at sharply lower mention counts, with profitability mentions falling from 40 to 9.

Sankar used the call to convert conviction into contract shape, citing a Fortune 100 tech company that converted a pilot into a three-year, nearly $370 million deal and a global asset management firm that committed to a three-year, $35 million TCV. Glazer disclosed that net dollar retention reached 157%, up 700 basis points quarter-over-quarter, and that U.S. commercial TCV printed at $2.132 billion, more than doubling year-over-year. The customer-count trajectory backs the read: 954 at year-end, 1,007 in Q1, 1,049 in Q2.

Karp used one swap-out story to frame the moat, a frontier lab failed to deliver a ticketing automation bake-off, the customer kept the same models and timeline, and the only variable was Palantir's forward-deployed engineers, which converted into a mid-single-digit-million ACV contract. The same model-portability point shows up in his pricing language: clients can be steered to other models if the frontier shifts, and the company is not pitching lock-in. Karp also acknowledged the constraint on the other side, noting an unrealized $0.03 tailwind to GAAP EPS from the SpaceX mark, a non-operating variable the consensus framing does not price.

The peer split runs along the same U.S.-government-versus-AI-software axis. Microsoft reported Azure growth of 43% in constant currency for Q4 FY2026, ahead of the 39-40% guidance from the prior quarter, and reiterated a fiscal-year-2027 capex envelope of approximately $175 billion after a useful-life reclass, a sign that demand continues to exceed supply at the infrastructure layer.

Oracle's Q1 FY2027 print carried a similar supply-led tone, with the CEO stating that demand continues to support the investment and that more than $30 billion of additional AI contracts were closed in the quarter without requiring additional capital from the issuer, alongside RPO growing to $664 billion. Snowflake's Q2 FY2027 results reinforced the same direction, with product revenue accelerating for a third consecutive quarter and net revenue retention holding at 126%; ServiceNow reported current RPO of $13.20 billion, up 21% year-over-year, on a 24% revenue print. The AI software set broadly corroborates Karp's read: the AI buildout is real, capacity-constrained, and U.S.-anchored.

The government-services primes diverge. Booz Allen posted Q1 FY2027 revenue down 4% year-over-year, with civil revenue an explicit high-single-digit decline expected for the year; the executive order shifting federal work toward firm-fixed-price contracts was named as a structural swing, and fixed-price share moved 300 basis points higher in the quarter. Leidos reported Q2 FY2026 revenue up 7% but adjusted EBITDA margin compressed 140 basis points year-over-year, with the Health segment the chief drag and management guiding civil-style revenue to a low-single-digit full-year increase.

CACI's full-year revenue rose roughly 11%, the strongest of the primes, yet backlog grew in the mid-single digits. CrowdStrike ran on the other side of the same divide, with ending ARR up 25% and guidance for net new ARR lifted by a cumulative 1,150 basis points since the start of fiscal 2027 on Falcon Flex commitments that uplift existing subscriptions by more than 40%. Valuation vs peers (Software - Infrastructure and cross-industry comps)

Performance vs peers (Software - Infrastructure and cross-industry comps)

Palantir Technologies Inc. management guidance — from the 8-K filed 2026-08-03

Adjusted free cash flow has stepped from $140.8 million in Q3 2023 to $1,220.4 million in Q2 2026, a near 9× trajectory across the table. The quarterly print also lifted the same-quarter margin from 25% to 63%, so each dollar of revenue now carries roughly four times the cash conversion it did in late 2023. The chart sits beside this paragraph because the curve, not the level, is the story: four out of the last five quarters cleared the $500 million mark, and the Q2 2026 print is the largest single quarter in the file.

Adjusted operating income tells the same direction in dollars, $163.3 million in Q3 2023 to $1,194.5 million in Q2 2026, a 7× expansion. The Q4 2025 step from $373 million to $798.5 million in the following quarter shows the slope steepened, and the Q2 2026 reading is the highest in the table.

Adjusted diluted EPS ran from $0.05 in Q2 2023 to $0.41 in Q2 2026, an 8× move on a share base that moved only single digits over the same period. The cleanest read on the equity story is here: $0.41 per share on a denominator that has barely moved, against revenue that more than doubled in the same window.

The commercial-versus-government split is where the segment table earns its keep. In FY2021 the company ran about three-fifths government and two-fifths commercial; FY2025 closed closer to a 54/46 split, with two-year CAGRs in the low forties on both sides of the mix. The Q2 2026 contribution math in the source excerpt sharpens the directional read: commercial contribution margin reached 78% versus 64% a year ago, government 71% versus 63%.

The mix is rotating toward the higher-margin side at the same time the commercial side is compounding faster, and the government's share of the slow-moving dollar base is dropping. Among the named peers, Microsoft's (MSFT $493.78 -0.80%) FY2026 print is dominated by Productivity and Intelligent Cloud (42% and 42%), Oracle's (ORCL $147.61 -1.98%) FY2026 by Cloud and Software (87%), and Leidos's (LDOS $127.69 -2.50%) FY2025 by National Security and Digital (44%) — none of which is a like-for-like segment, only context for how concentrated the dependence on a single segment can be.

The peer-versus-peer read cuts against the headline-multiple framing. Microsoft prints an EV/revenue above 11× on revenue growth in the high single digits; Oracle prints roughly 7–8× on mid-teen revenue growth; Palo Alto Networks (PANW $363.58 -3.06%) runs on revenue growth near 29% with no published multiple; CrowdStrike's (CRWD $237.65 -3.28%) growth in the low twenties pairs with an unprintable multiple. In the services half, Leidos prints about 1.2× EV/revenue on roughly 3% growth and CACI (CACI $626.85 -1.28%) about 2× on roughly 2.5%.

Palantir's growth, above 56%, sits at the top of the table. The companies in the services half run at low single-digit revenue growth with single-digit-low EV/sales, while the software-AI half prints multiples an order of magnitude wider. The valuation-vs-performance tension is between these two regions of the table, not against either of them individually, and Palantir's growth figure clears both halves without sharing either's multiple structure.

Capital return has been nearly dormant. Buybacks ran $64.2 million in FY2024 and $75 million in FY2025; no dividend has been paid in any year on file.

The management guide points the trajectory in one direction. The latest results release sets the full-year 2026 revenue guide at $8.2 billion, adjusted operating income at $4.9 billion, and adjusted free cash flow at $4.5 billion to $4.7 billion against a $3.4 billion-plus U.S. commercial floor. Adjusted FCF margin already printed 63% in Q2 2026, the highest in the table; the guided band implies the company is sustaining close to that conversion against the full-year revenue target. RPO was $4.9 billion at June 30, 2026, of which 43% sits in the next twelve months, a near-term conversion share that puts more than $2 billion of next-twelve-month revenue on contract today and that the headline backlog debate does not factor in.

The read-through

The peer tape that closed after Palantir's Q2 2026 results confirms the U.S. AI buildout, but it also tightens two of the bear pressures the headline multiple skips. The first is supply-led concentration at the customer; the second is U.S. federal budget risk absorbing the primes while the software primes print above them.

Microsoft's Q4 FY2026 call said demand continues to exceed available supply, and the Q4 FY2026 results put calendar 2026 capex at roughly $190 billion. If Palantir's U.S. commercial revenue grows on the back of enterprises that are themselves absorbing that compute envelope, then Palantir's guided floor of more than $3.4 billion becomes a function of hyperscaler capacity landing. The risk is the same one Karp named on the Q2 2026 call: demand exceeds what forward-deployed engineers can deliver, and the conversion slips. Microsoft's capex figure is Microsoft's; the mechanism is that Palantir sells into the same buyers.

CrowdStrike's Q2 FY2027 call lifted net-new-ARR guidance by a cumulative 1,150 basis points since the start of the fiscal year and cited more than 935 Falcon Flex adds in a single quarter. The same dollar-weighted duration story that boosts RPO for CrowdStrike, a Q2 FY2027 10-Q RPO of $10.7 billion with 46% to be recognized in the next twelve months — also shows that single-platform budget concentration is moving against customers, not in Palantir's favor. If a buyer's full endpoint wallet is already locked into one platform, the AIP displacement window narrows.

Oracle's Q1 FY2027 results release carried RPO of $664 billion, with management guiding roughly half to convert over thirty-six months and describing multi-cloud database revenue growing 353% year-over-year. The implication for Palantir is read-through only: if Oracle's data layer absorbs the enterprise AI plumbing, Palantir's ontology layer competes for the application dollars layered above it.

The services primes run the opposite direction. Booz Allen's (BAH $77.07 -2.31%) Q1 FY2027 results release posted revenue down 4%, with civil guidance a high-single-digit decline and the April 2026 federal contracting executive order driving fixed-price share up 300 basis points in the quarter. Leidos's Q2 FY2026 results release raised full-year revenue to $18.20–$18.40 billion yet compressed adjusted EBITDA margin by 140 basis points. The federal side is squeezing the labor-heavy integrators while platform spend stays open, which the management guide already mapped into Karp's U.S. commercial framing.

Microsoft's Q4 FY2026 call said demand continues to exceed available supply, and the Q4 FY2026 results put calendar 2026 capex at roughly $190 billion. If Palantir's U.S. commercial revenue grows on the back of enterprises that are themselves absorbing that compute envelope, then Palantir's guided floor of more than $3.4 billion becomes a function of hyperscaler capacity landing. The risk is the same one Karp named on the Q2 2026 call: demand exceeds what forward-deployed engineers can deliver, and the conversion slips. Microsoft's capex figure is Microsoft's; the mechanism is that Palantir sells into the same buyers.

The competitive field that the filings surface splits cleanly. The direct operating rivals, Microsoft, Oracle, Snowflake, ServiceNow, CrowdStrike, compete for the same enterprise AI wallet, and Palantir's sourced edge is the forward-deployed-engineer motion that converted the frontier-lab bake-off cited on the Q2 2026 call. The stock-market rows: Palo Alto Networks, Booz Allen, Leidos, CACI — share the customer file but not the cost structure. Customer concentration cuts the other way: one customer represented 31% of accounts receivable at March 31, 2026, and no customer cleared 10% of revenue in the latest quarter, a thinner concentration footprint than the headline backlog implies.

Liquidity rests on customer prepayments and short-duration Treasury holdings

Palantir Technologies Inc. balance sheet — as of Jun 30, 2026 (10-Q filed Aug 4, 2026)

The Q2 2026 10-Q shows a balance sheet with no long-term debt line item, no current portion of long-term debt, and the only balance-sheet encumbrance of consequence is the $211.4 million noncurrent operating-lease liability. Total liabilities of $1.8 billion against $11.7 billion of total assets, and $9.8 billion of stockholders' equity, leave the structure essentially unlevered.

Cash and equivalents of $2.0 billion sit alongside $7.4 billion of short-term investments, the bulk of which the 10-Q discloses as U.S. Treasury securities at fair value ($7,195.0 million) with the remainder in publicly-traded equity securities ($184.1 million). Restricted cash is immaterial; the $2,063.4 million "cash, cash equivalents, and restricted cash – end of period" line on the cash-flow statement is overwhelmingly unrestricted. Palantir funded the first-half 2026 build in cash and Treasuries entirely from operating cash flow — the half-year cash-and-securities balance grew by roughly $2.2 billion versus year-end 2025, with no underwritten equity issuance or debt raise in the period. The SpaceX mark Glazer flagged on the Q2 2026 call sits inside that $184.1 million equity-securities line, so the $0.03 EPS tailwind he attributed to it does not require cash to be touched.

The customer side of the liability stack is the part the unlevered read glosses over. Deferred revenue rose to $579.4 million current plus $33.7 million noncurrent ($613.2 million total) at June 30, 2026, up from $455.2 million at year-end. Customer deposits, the cash side of those customer-funded obligations, moved from $357.1 million to $452.6 million over the same period. Together the customer-funded line runs roughly $1.07 billion, versus total liabilities of $1.8 billion: customer prepayments are the dominant claim on the asset side. Glazer described Palantir's book as having "no debt" on the call, but those customer deposits carry their own implicit service-delivery obligation that the unlevered framing ignores.

Buyback activity has been the only capital-return channel, and it is small. The equity rollforward in the 10-Q shows an additional $18.6 million repurchase in the first half of FY2025 and only $1.4 million in other financing during the first half of FY2026. The dilution discussion in the prior section runs against a buyback program that is functionally dormant: the entire share-count offset has come from treasury-stock retirements that are not even close to absorbing the vesting flow.

What has to land is supply, conversion and a share count that stops drifting

The bull case rests on three load-bearing assumptions, and each has a visible failure mode in the filings. The first is that U.S. commercial growth stays above the 134% floor Glazer guided on the Q2 2026 results release, with Karp on the Q2 2026 call saying he is targeting that rate for the next eighteen months.

The supporting evidence is concrete: U.S. commercial TCV of $2.132 billion more than doubled year-over-year, top-twenty customer revenue averaged $124 million (up 67% year-over-year per the source excerpt), The failure mode is the dependency Karp himself named on the Q1 call, that forward-deployed engineers are the binding constraint, and that demand already exceeds what the team can deliver. If the 18-month goal requires more FDE capacity than the company can hire and clear, conversion slips and the guide breaks before the multiple does.

The second assumption is that the contract shape carries the growth through the multiple. The dollar-weighted duration metric rose 129% year-over-year, three-year deal sizes cited on the call ran from $35 million to nearly $370 million, and remaining performance obligations of $4.9 billion carry 43% in the next twelve months per the Q2 2026 10-Q. The bear counter sits in the same filing: many customers can terminate for convenience on less than twelve months' notice, and total remaining deal value of $13.1 billion runs roughly 2.7× the RPO. Karp and Sankar pitch that gap as the under-stated book; the same gap means most of the visible backlog is cancellable, so the headline RPO is closer to a coverage-of-next-quarter read than a multi-year commitment.

The third assumption is that the share-count math holds. Diluted weighted-average shares moved from 2,450.8 million in FY2024 to 2,565.2 million in FY2025 to 2,568.7 million in Q2 2026, an effective pause, but outstanding SARs doubled to 14 million, antidilutive awards tripled to 3 million, and stock-based compensation of $265.2 million in Q2 2026 rose 66% year-over-year with sales-and-marketing SBC alone doubling per the same 10-Q. The buyback book is functionally closed (no dividend on file; FY2024 and FY2025 buybacks totaled $64.2 million and $75 million respectively), so the cleanest adjusted EPS print of $0.41 sits on a denominator that drifts up as RSU vesting compounds. A break in any of these three paths resets the multiple debate the filings so far have kept at arm's length.

Risk: the multiple debate ignores the $4.9B RPO and the dilution math

The Q3 2026 print against the $2.160–$2.164B revenue band Glazer set on the Q2 2026 call is the decisive test: a clean in-line print preserves the multiple-debate standoff the filings have kept at arm's length, while a miss reframes the 147.2× trailing P/E immediately. The 134% U.S. commercial growth floor is the second number — break it and the largest-ever guide raise on a larger base stops being a re-rating catalyst and becomes the high-water mark. The 63% adjusted FCF margin printed in Q2 2026 is the third: sustain it against the $4.5–$4.7B full-year guide and the dilution math (stock-based compensation of $265.2M up 66% year-over-year, outstanding SARs doubled to 14M) absorbs into the cash story; slip on margin and the share-count creep becomes the lead headline again.

Method note: Equibles compiled the figures from SEC filings and deterministic datasets, and the commentary from earnings-call transcripts and filings. This article is for information only and is not investment advice.