Palantir Stock After Q3 2026 Earnings: The Contracts Got Shorter

I think the market has the wrong worry about Palantir's Q3 2026 bookings: the contracts really did get shorter, as CFO David Glazer said on the call, and that leaves the moat intact while letting customers test it sooner. The stock fell about 3% in extended trading after the report. US commercial revenue grew 148%, but the contracts signed there grew 11%, and that gap needs an explanation.

This print also hit the first half of a test I published in September, and that test turned out to be badly built. I'll own that further down.

Palantir reports two forward numbers for US commercial, its fastest-growing business. Total contract value (TCV) is what it signed in the quarter. Remaining deal value (RDV) is what's left on all contracts at quarter end. Revenue drains RDV and signings refill it, and in Q3 revenue took most of what was signed.

Divide RDV by a quarter's revenue and you get a rough count of how many months the book would last at today's run rate. Palantir doesn't publish this. It's my arithmetic and it's crude. It sat flat through most of 2025, slipped as revenue sped up, then dropped.

Here's why I don't read that chart as moat damage. Palantir's own definitions say TCV and RDV presume every option gets exercised, and that the majority of its contracts can be terminated, including for convenience. So a customer on a three-year contract was never really held there by the paper. It stays because its workflows and decisions are written into Palantir's ontology, the data model that its people and AI agents work from, and rebuilding that elsewhere is the real cost of leaving.

What shorter deals take away is time. Palantir hasn't said how short they are. If they run about a year, the first of them come up in late 2027, and the alternatives are moving quickly. Anthropic's S-1 says its revenue is primarily consumption-based and much of it isn't under long-term commitments, so a lab's customers aren't locked in either. When neither side holds the customer on paper, it comes down to whether the work is embedded.

I think business logic is the pillar a shorter contract tests, and our ratings have Palantir ahead of both rivals there. CTO Shyam Sankar's pitch on the call was that labs treat the knowledge worker as the unit of work when it should be the workflow, and I've written up the one bake-off Palantir has described. Microsoft is the harder rival, I think. Our ratings have it ahead on bundling, the system of record and transaction embedding, so a customer weighing a renewal has an alternative that's already paid for in its estate. Microsoft has its own gap between revenue and backlog, but the bundle doesn't depend on a contract term.

Glazer said Palantir signed "multiple 8-figure deals in the quarter that were shorter in duration than usual" as customers adopted its Sovereign AI products, its pitch for models a company owns and runs inside its own walls. He expects duration and size to grow over time. In the Q&A, CEO Alex Karp described "shorter duration of the contract, but then an ability to expand that contract dramatically". Palantir gave no number for how much TCV this cost.

The deal counts give that some support. From Q2 to Q3, deals of at least $1 million went from 220 to 212, at least $5 million from 98 to 107, and at least $10 million from 73 to 64. So the $5 million to $10 million band grew from 25 to 43 while the band below it shrank. My reading is that some eight-figure deals came in smaller, though Glazer only said they were shorter, and nine fewer big deals can explain at most half of that gain. The counts are also company-wide and the TCV fall was all in US commercial. On the call, remaining performance obligations were flat at $4.9 billion, which fits if these deals run under a year, since that figure leaves out shorter terms. Flat is also what softer signing looks like.

In September I said two straight quarters of US commercial RDV adding less than $1.3 billion would show Q2's record bookings were a pull-forward. Q3 added $279 million, so that's one down. I expect Q4 to miss as well: the full-year guide implies at least about $1.2 billion of US commercial revenue in Q4, so adding $1.3 billion of RDV would take about $2.5 billion of signings, above the record. If it misses, I'll mark that claim tripped on the old article and correct it there. That concedes Q2's pace didn't repeat. It doesn't concede a customer lost to a lab, which hasn't happened in anything Palantir has disclosed.

I still think the threshold was wrong, and you can discount that since I'm saying it after the miss. It was a fixed dollar figure taken from the single best quarter in the table, and three of the last four quarters missed it. So this article's test scales with the business: contract value signed against that quarter's revenue.

The strongest case against me is that I'm accepting management's explanation for the one bad number in a great quarter. Glazer gave no figure, and he didn't say shorter deals explain all of a 31% fall in US commercial TCV from Q2. The reading my September test was built to catch is still alive: customers rushed to sign in Q2 and Q3 was the hangover. Part of that fall may also be one deal, since the three-year contract of about $370 million Palantir described in Q2 may sit in the record, though it hasn't said which line it was booked on.

Customer growth points the bears' way, and it's the part of the print I like least. Palantir's counts are of customers it recognised revenue from over the trailing 12 months, so they aren't a clean read on new customers, but the direction is clear.

Net dollar retention doesn't rescue me either. It compares what last year's customers spent over the trailing 12 months with what they spent the 12 months before, so 165% means they spent 65% more. It mostly reports on deals signed earlier, on the older and longer terms, and it could keep rising for several quarters after signings soften. That's why it's only my late alarm. The book has a leak I can't explain too: RDV added fell short of signings minus revenue by about $200 million in Q3, against about $50 million in each of the three quarters before. Lapsed options, terminations, or Palantir's own exclusions for customers' financial condition could each explain it, and Palantir hasn't said.

Holders will object from the other side. Months of cover is falling because revenue is growing 29% a quarter, and holding 24.5 months in Q3 would have taken about $8 billion of RDV. On that view I've drawn a chart that punishes success. I give that some weight, and it partly survives in my test, since signings against revenue has the same denominator. I use that ratio because it's a flow of reported numbers with no option assumptions inside it. But Q2 revenue grew 28% from Q1 and the book barely moved. Karp also called the relationship with Anthropic "symbiotic" because it's growing the market, then said: "I don't know at what point we become competitive with each other, but at this point in the US, the market for us is insatiable." Today's demand backs him. I'm more interested in the first half of that sentence.

Shorter deals push this ratio down on their own, so I don't expect it back at 3. But Karp says these deals expand once customers see results, and expansions get booked as new contract value, so two of the next three reports should clear 1.5 times. On the full-year guide that's about $1.8 billion of signings in Q4. I picked 1.5 because Q3 is the series low and I'm betting it stays the low, so it's a line I drew and you can argue with it. Even at 1.5 the book keeps thinning, to about 18 months on that guide. Below it, the shorter contracts were softer demand with a better name.

That covers the explanation. The moat test comes later, because on my one-year assumption nothing short renews before late 2027. I said retention can't settle things today, but it's the right late alarm: below 130% by the Q4 2027 report would mean customers are leaving or shrinking as those deals end. The scorecard below shows where our ratings put Palantir and its two rivals at today's prices.