Palantir Shows Why AI Won’t Do Your Homework

Palantir Shows Why AI Won’t Do Your Homework

Guest Post by Philip van den Berge, Founder and CEO, Intrinsiqq

The pitch behind most AI investing products is that the hard part is already solved. Point the model at the filings, let it read faster than you can, and the answer comes out the other end.

The hard part is not reading the filings. The hard part is that the answer changes, and nothing warns you when it does. Palantir Technologies (PLTR) is the cleanest example I know.

For years the standard objection to Palantir was that its free cash flow wasn’t real. Subtract stock-based compensation, the argument went, and the cash generation vanishes. That objection was correct. Then it quietly stopped being correct, and a great deal of commentary never caught up.

Here is what Palantir filed. Figures are in millions of dollars, taken from the company’s annual reports. Free cash flow is operating cash flow less purchases of property and equipment.

In 2021 and 2022, Palantir’s free cash flow genuinely did not cover its stock compensation. Not narrowly. The company came up $457 million and $381 million short. Anyone calling the cash flow an accounting artifact was reading the filing correctly.

By 2025, free cash flow is $2.10 billion against $684 million of stock compensation, a gap of $1.42 billion the other way. The same test that failed twice now passes comfortably, because free cash flow grew 6.5 times while stock compensation went sideways. Palantir’s 2025 compensation expense was lower in absolute dollars than 2021’s, against nearly three times the revenue. That inversion happened in 2023. The 2022 answer is still being repeated in 2026.

Meanwhile the number that actually costs a shareholder kept climbing. Diluted shares went from 1.92 billion to 2.57 billion, up 33% in four years. Palantir did start buying stock back, $64 million in 2024 and $75 million in 2025, but against $684 million of issuance that offsets about 11%. The rest is permanent.

Run it the way an owner would. Free cash flow grew 6.5 times between 2021 and 2025. Free cash flow per diluted share grew 4.9 times, from roughly $0.17 to $0.82. Both are excellent. But a quarter of the improvement went to new shareholders rather than existing ones, and no expense line will ever show you that, because dilution is not an expense. It is a change in the denominator. So the popular objection expired and the durable one got dropped.

I should be straight about my own position, because I sell software that computes fundamentals from SEC filings, Palantir included. That software can build the table above in seconds. It cannot tell you that the 2022 answer expired in 2023, because that is not a computation, it is a judgment about which comparison still means something.

This is the gap I would want investors to hold onto when they see an advertisement promising that a model will handle the analysis. Retrieval is cheap now and getting cheaper. Knowing that a ratio you have been carrying around for four years quietly inverted is not retrieval. A tool that hands you a number without telling you the question has moved has made you faster at being wrong.

Two habits cover most of it, and neither requires a view on whether Palantir is cheap. Recompute rather than recall: if a thesis rests on a relationship between two line items, re-derive it from the newest filing every time you use it, because a ratio you remember is a memory of evidence and not evidence. And track the share count as its own line, because stock compensation and dilution are related but not the same measurement, and at Palantir they moved in opposite directions.

None of this settles what the company is worth. Reasonable people will disagree sharply about what to pay for 56% revenue growth at a 47% free cash flow margin. The disagreement is just more useful when both sides are arguing about the current filing.

Investors should be cautious with anything that offers to end that argument for them.

Disclosure: I have no position in Palantir.

About the Author:

Philip van den Berge is the founder and CEO of Intrinsiqq, which computes company fundamentals directly from SEC filings. Every figure in this article comes from Palantir’s annual reports and can be checked against the source documents.

Published by NCV Newswire
NCV Newswire
The NCV Newswire by New Cannabis Ventures aims to curate high quality content and information about leading cannabis companies to help our readers filter out the noise and to stay on top of the most important cannabis business news. The NCV Newswire is hand-curated by an editor and not automated in anyway. Have a confidential news tip? Get in touch.

Get Our Sunday Newsletter