At roughly 74 times sales and 100 times forward earnings, Palantir has become less a conventional software valuation exercise than a referendum on how valuable the operating layer of enterprise AI could eventually become.
Wall Street rarely agrees on anything, but Palantir Technologies has achieved something special: it has given analysts an unusually large number of ways to disagree.
At $189.67, $Palantir Technologies Inc.(PLTR)$ carries a market capitalisation of approximately $456 billion. Analyst targets stretch from $80 to $255, while the average sits at $195.57. That is not ordinary forecasting noise. Investors are effectively valuing two different companies wearing the same ticker.
One is an extraordinarily profitable AI platform entering a potentially enormous commercial and defence opportunity. The other is an excellent software business whose share price has already borrowed several years of success from the future.
Curiously, I think both descriptions contain some truth.
Wall Street is weighing two Palantirs on one very expensive scale
The Moat Isn't the Model
The most interesting part of Palantir is not artificial intelligence itself.
Large language models are becoming cheaper, more capable and increasingly interchangeable. Palantir does not need to build the world's best foundational model because its real proposition sits above and around those models.
Its Ontology connects an organisation's data, rules, assets, workflows and permissions to the decisions people — and increasingly AI agents — actually make. A factory is not merely rows in a database; it becomes machines, inventory, orders, employees and actions that software can understand and manipulate within defined controls.
That distinction matters.
An enterprise can swap one LLM for another relatively easily. Reconstructing the operational representation of the enterprise underneath it is considerably less entertaining. Think root-canal treatment, except the dentist is also migrating your ERP system.
This is where Palantir's stickiness may ultimately reside.
There is also a counter-intuitive possibility: commoditisation of LLMs might strengthen Palantir rather than destroy it. As models become cheaper and interchangeable, economic value could migrate towards the orchestration, governance and operational layer deciding what those models are permitted to see and do.
That turns one of the standard bear arguments partially on its head.
Commercial AI Has Left the Laboratory
The financial evidence is becoming difficult to dismiss.
Trailing twelve-month revenue reached $6.16 billion through June 2026, up 78.9%. U.S. commercial revenue reached approximately $2.26 billion, versus $702 million in FY2024. Total U.S. commercial customers increased from 382 in 2024 to 653, while net dollar retention climbed from 120% to 157%.
That last number deserves attention.
A 157% net dollar retention rate indicates that existing customers are expanding spending dramatically before Palantir even adds another one. This supports the argument that its AIP bootcamps are not simply impressive demonstrations designed to make executives nod thoughtfully over expensive coffee. They are translating into production deployments and larger contracts.
Total remaining performance obligations have risen from $1.73 billion in 2024 to $4.90 billion on a trailing basis.
Palantir is therefore doing something particularly valuable: widening its customer funnel while extracting considerably more revenue from customers already through the door.
The valuation exploded, but the underlying financial machine changed too
< Insert CHART 1 Here > Caption: The valuation exploded, but the underlying financial machine changed too
The Cash Machine Has an Asterisk
Revenue increased from $2.23 billion in 2023 to $4.48 billion in 2025 and $6.16 billion on a trailing basis. Profitability has expanded even faster.
Gross margin now stands at 84.8%, operating margin at 42.8% and net margin at 49%. Palantir generated $3.36 billion of trailing free cash flow, equivalent to a remarkable 54.6% FCF margin. Cash and short-term investments total $9.41 billion against only $211 million of debt.
But that $3.36 billion deserves an asterisk.
Trailing stock-based compensation is approximately $835 million — nearly 25% of reported free cash flow. SBC is non-cash for accounting purposes, but it is emphatically not free for shareholders: employees receiving shares create an economic cost through dilution. Treating reported FCF as though SBC simply evaporates therefore flatters the underlying economics.
Palantir has moderated dilution as profitability has scaled, but investors should still distinguish between accounting cash generation and owner economics.
The cash machine is formidable. Shareholders simply need to remember that some of its fuel arrives in shares.
The $15 Billion Question
CEO Alex Karp's projection that Palantir could generate $15–18 billion of free cash flow within two years is central to the valuation debate.
Against trailing FCF of $3.36 billion, even the bottom of that range requires extraordinary expansion.
Yet the mathematics explains the excitement. At $15 billion in FCF, today's $456 billion market capitalisation would represent roughly 30 times that future cash flow. At $18 billion, it falls towards 25 times.
Suddenly today's apparently absurd valuation becomes considerably less absurd.
There is, however, an obvious footnote: Karp is not an independent analyst. His substantial personal share sales are well documented, and investors should consider both his incentives and those transactions when placing weight on exceptionally ambitious management forecasts. Insider selling does not invalidate the forecast — executives sell shares for many reasons — but neither should management guidance be mistaken for a discounted cash-flow model delivered from Mount Sinai.
That is the real Wall Street schism. Bulls are valuing the company Palantir might become surprisingly quickly; bears are valuing the cash flows investors actually own today.
Everyone Wants the Middle Layer
Competition also requires more nuance than declaring Palantir untouchable.
$Microsoft(MSFT)$, Databricks and Snowflake increasingly overlap with enterprise data, analytics and AI infrastructure. Frontier-model companies such as OpenAI and Anthropic are pushing agents deeper into enterprise workflows. In defence technology, Anduril and established contractors compete for portions of expanding budgets.
Palantir's advantage is that few rivals combine data integration, operational ontology, AI orchestration, security, deployment and years of mission-critical government experience within one architecture.
But integration is a moat only while customers value integration more than modularity.
Open standards, cheaper inference and increasingly capable autonomous agents could eventually allow enterprises to assemble competing stacks from multiple vendors. Palantir therefore needs Ontology to become infrastructure customers cannot sensibly remove, rather than middleware they merely prefer not to remove.
That distinction could determine hundreds of billions of dollars in future market value.
Defence Is Both Moat and Exposure
Government revenue now represents approximately $3.21 billion, or roughly 52% of trailing revenue.
Palantir's Western alignment gives it credibility in sensitive defence and intelligence environments where security clearances, trust and deployment history matter enormously. Rising Western defence expenditure and demand for sovereign AI infrastructure potentially extend that runway.
Yet concentration cuts both ways.
With more than half of revenue still coming from government customers, changes in procurement priorities, defence programmes or the loss or delay of major contracts could materially affect growth. Government relationships may be unusually sticky, but government budgets ultimately remain political documents.
Palantir's Western orientation creates another constraint. It cannot position itself as strategic infrastructure for allied governments while behaving like a politically neutral global SaaS vendor selling indiscriminately everywhere.
Its philosophy strengthens the moat inside its chosen ecosystem while limiting parts of the theoretical market outside it.
Perfection Has a Price
My difficulty with Palantir is not finding flaws in the business. It is deciding how much perfection investors should pay for in advance.
The company has accelerating revenue, exceptional margins, formidable cash generation, a fortress balance sheet and exposure to two structural growth markets: enterprise AI and defence modernisation.
The counterweight is equally concrete. SBC equates to nearly 25% of reported trailing free cash flow. Government customers still account for roughly 52% of revenue. Competition will intensify. And a valuation of around 74 times sales assumes extraordinary execution for years, not quarters.
Price tells the story; volume reveals where investors actually committed
This is why the $80-to-$255 analyst range is less ridiculous than it looks. Small differences in assumptions about growth, margins and terminal valuation create enormous differences in present value because so much of Palantir's implied worth resides in cash flows that have not yet arrived.
The finish line is priced in. Fundamentals still have to arrive
Palantir may ultimately prove that conventional software valuation models were looking through the wrong end of the telescope. Equally, an exceptional company can remain an unforgiving investment when expectations become exceptional too.
For me, the debate is no longer whether Palantir's business is real. The financial statements have largely settled that argument.
The question is whether Palantir can grow quickly enough — and cleanly enough on a per-share basis — for reality to catch a valuation already sprinting several years ahead.
At $456 billion, reality had better bring running shoes.
@TigerStars @Daily_Discussion @Tiger_comments @Tiger_SG @Tiger_Earnings @TigerClub @TigerWire
Comments