Palantir Technologies Inc. PLTR

Key data

The business

Palantir sells two software platforms and the engineers who wire them in. Gotham runs on the government side, pulling scattered data, intelligence, sensors, logistics, into one operating picture a commander or analyst acts on. Foundry does the same for companies, and AIP, the newest layer, lets an operator point a large language model at that data and get an action back, not just an answer. Revenue splits roughly between government and commercial, with the US commercial book growing fastest as AIP lands. The moat is depth of integration: once Palantir's people have mapped a customer's messiest data into an ontology and the daily workflows run on it, ripping it out means rebuilding the nervous system of the operation. The thing a customer actually touches is a screen where a supply-chain manager types a question and watches the system reroute a factory. No customer is more than 10% of revenue, though one is 27% of receivables, a reminder that a few large government deals still swing the timing.

The numbers

Every line is accelerating, and the change in the rate is the story.

Revenue nearly doubled year over year in the June quarter and net income more than tripled, and sequential growth reaccelerated in 2026 after a steadier 2025. Earnings have come in above the average analyst estimate every quarter shown, most recently $0.41 against $0.34 expected.

The company lost money through 2022, turned GAAP-profitable in 2023, and in the first half of 2026 alone earned more, $1.93B, than in all of 2025, $1.63B. Over four years revenue compounded about 31% a year, and the last twelve months ran far hotter than that.

Margins are the rare kind, and 55 cents of every revenue dollar becomes free cash while stock compensation stays a moderate 14% of sales for a company this young. Contract liabilities grew to $1.1B from $0.8B, so the backlog is filling ahead of revenue, not behind it. What the market does not fully price is deceleration in the growth rate itself: consensus puts revenue near $33B by 2029, roughly 65% compounding for four more years, and the print that settles the bet is not the next earnings line but the year-over-year growth rate, the first quarter it slips from the 90s toward the 50s.

Management

Insiders have sold and not bought, $207M across 68 sales in the last year against zero purchases. CEO Alex Karp accounts for the two largest, $53.6M in August and $30.5M in May, and President Shyam Sankar sold $19.5M in July. Plan status is not disclosed in the feed, so whether these ran on preset 10b5-1 schedules or were discretionary cannot be stated, only that the selling is steady and one-directional. Buybacks are token, $75M last year against a $400B company, so they neither support the stock nor offset dilution, and stock compensation, not repurchases, sets the share count. The balance sheet holds $2.0B in cash against almost no debt, returns on capital run above 25%, and management reinvests rather than returns. The guidance record has been to set a bar and clear it.

How it fails or surprises you

Growth rate decelerates. At 148x earnings the price assumes the near-doubling holds. If year-over-year revenue growth slips from the low 90s toward 50% over the next two or three quarters, still excellent, the multiple can compress faster than earnings rise. A move from 148x to 70x halves the stock even as profit grows. Watch the YoY revenue rate each quarter.

Government concentration bites. One customer is 27% of receivables and a handful of large public contracts drive timing. A budget delay, a continuing resolution, or a single deal slipping a quarter would surface first as a sequential revenue stall, the one thing the current price cannot absorb.

AIP inflects wider (right tail). The upside is AIP spreading from flagship logos to thousands of commercial seats while government adopts agentic AI faster than budgeted. If US commercial growth stays triple-digit and government AI spend accelerates, the $17.9B 2028 number arrives early and today's 51x on 2028 looks cheap in hindsight. The tell is US commercial customer count and revenue in the quarterly release.

Closing thoughts

What settles this is the year-over-year revenue growth rate, printed each quarter and watched closely. While it stays in the 80s and 90s the multiple is self-justifying and the stock can keep running. The first two or three prints that show a clean step down convert the story from acceleration to merely-great growth, and at 148x that re-rating is violent. An ambiguous print, growth easing to the 70s with a full backlog, resolves nothing and leaves the reader waiting for the next rate reading rather than acting on one quarter. The left tail is a growth stall meeting a triple-digit multiple, the right tail is AIP compounding into a number that makes today's price look modest, and the downside is fuller and faster than the upside is generous, because the multiple does the punishing.

The bet is still that governments and big companies keep handing Palantir more work and paying more every year for it, and that AIP spreads before anyone builds it cheaper. What breaks it is the growth rate, not the growth, and the pair to watch is year-over-year revenue and the sequential change in that rate. The read is wrong the first quarter revenue growth decelerates while the multiple stays above 100x.