Palantir’s gross margin went up while its revenue nearly doubled. From 80.6 percent in 2023 to 84.7 percent in the second quarter of 2026, with revenue at $1.94 billion, up 93 percent year over year.

Accenture runs around 32 percent.

That single comparison is the strongest argument against the case Michael Burry made this month, and almost nobody covering the story mentioned it.

What he actually said, and what he actually measured

In early September he published a long piece on X called “Palantir: An Accounting,” arguing Palantir is a consultant riding a bubble of AI FOMO demand and warning the market cap could fall below $100 billion from roughly $430 billion.

Most of it is about accounts receivable, and that part is careful work. One customer has sat at roughly a quarter of Palantir’s receivables for three straight years without ever crossing 10 percent of revenue, and average days sales outstanding nearly doubled between 2020 and 2025. Those are real questions. This post does not answer them.

The part everyone quoted is the other one. He charted five years of quarterly deferred revenue for nine companies out of the fiscal 2025 10-K and read the ratio off each panel: SaaS peers between 80 and 207 percent, Accenture at 31 percent, Palantir at 32 percent. Deferred revenue on the balance sheet, divided by that quarter’s revenue.

Worth being precise about the base, because most of the coverage was not. Those figures come from the 10-K Palantir filed in February. Burry has said he will run the 10-Qs in a follow-up, and at the time of writing he has not published it.

So I ran his test on the most recent quarter myself.

Palantir’s 10-Q shows $613.2 million of total deferred revenue against $1,935.5 million of quarterly revenue as of June 30, 2026. That is 31.7 percent, within a rounding error of his annual figure. The test travels. For contrast on the same basis, Salesforce at its fiscal 2025 close carried $20,743 million against $9,993 million of quarterly revenue, or 207.6 percent. The 207 at the top of Burry’s own range is CrowdStrike, not Salesforce.

The reasoning behind the ratio is sound too. Subscription businesses bill a year ahead and carry a mountain of unearned obligation. Consultancies bill for work already delivered and carry almost nothing.

Two things the number leaves out

Both come from the same document he cites.

Palantir books customer deposits on a separate line from deferred revenue. At June 30 those were $452.1 million current plus half a million noncurrent. The 10-Q adds them together itself, calls the total contract liabilities, and reports $1.1 billion. Run the ratio on that and you get 55.1 percent.

Still nowhere near a subscription business. But 55 and 32 do not tell the same story, and the filing hands you 55 without any interpretive work.

Then the margin. A consultancy’s cost of revenue is salaried humans, and no amount of scale turns that into 85 points of gross margin. Palantir is posting 47 percent GAAP operating margins and 63 percent free cash flow margins alongside it.

One smaller thing. His Accenture figure of 31 percent comes off that same annual panel. Run the quarter instead, the way I ran Palantir’s, and Accenture comes in at 40.4 percent. Higher than Palantir, not lower, which is the opposite of the direction the comparison is usually deployed in.

Where he is completely right

Palantir does not disclose how much of its revenue comes from professional services. Anywhere. The 10-Q’s revenue disaggregation splits government from commercial and stops there.

So anyone quoting you a services percentage for Palantir made it up.

Alex Karp himself called the company a hybrid of infrastructure software, forward deployed engineers, orchestration and business know-how on the Q2 call. That word is his. Shyam Sankar went further and said only Palantir has FDEs while everyone else has sparkling sales engineers.

We wrote about half the industry hiring forward deployed engineers and renaming the job in the decade’s most expensive rebrand.

The test worth running on your own vendors

Here is what I take from this, and it has nothing to do with where the stock goes.

The ratio works. It is misapplied here, but it works, and you can run it on any publicly traded vendor of yours in an afternoon using documents that are free.

Add every contract liability in the latest filing, deferred revenue plus customer deposits, not just the first line. Divide by that quarter’s revenue. Then read gross margin next to it, which is the half nearly everyone skips.

Low ratio and low margin, you are buying hours on an invoice that says software. Low ratio and high margin, you are buying a product that still needs people to install it, which is a different business and a legitimate one, but it changes your implementation budget. High ratio, you paid in advance and the execution risk sits on their side of the contract.

Which of the three is your main AI vendor? Do you know, or are you assuming? Did anyone check before the last renewal went out?

One more thing before you cite this in a meeting. Burry states his position plainly and repeatedly in the piece: he is short Palantir stock and owns put options on it. What is not verifiable is the size or the timing, because Scion deregistered as an investment adviser on November 10, 2025, and its last 13F covers the third quarter of that year. None of that touches the arithmetic. It does mean the usual disclosure trail stops in 2025, so weigh the conclusion accordingly and check the filings yourself.

Run the numbers on your three most expensive vendors this week. Half an hour, public filings, and you come out knowing what you are actually buying.

Let’s evaluate your vendors before the next renewal