Palantir's Real Bet Is Control of AI, Not the AI Itself

DA Davidson raised its price target on Palantir to $250 on Thursday, up from $200, and kept its Buy rating. The reason, per the note: AIPCon11, the company's eleventh customer conference, focused on "AI sovereignty," a theme the analyst says is "gaining traction with customers." That $250 is roughly 50% above where the stock actually trades, near $167.

Before treating a new target as a signal, consider what a price target is: one analyst's estimate of what a share should be worth, revised whenever the stock moves or a conference changes the story. It is an opinion, not a value. DA Davidson's own path tells you how much weight to give it. In February it cut Palantir to Neutral and a $180 target. In July it flipped back to Buy at $175. Now $250. The target chased the stock, not the other way around, and the stock did not move on the news — it fell on the conference day as the broader market retreated. So the interesting question is not whether the analyst is right. It is what, exactly, he is betting on.

Abundant intelligence, scarce control

Here is the mechanism underneath the AIPCon talk. The past two years made raw AI abundant. Base models are everywhere, inference gets cheaper by the quarter, and every software vendor wraps an assistant around its product. When something becomes abundant, its price collapses — and what becomes scarce is whatever sits on top of it.

For enterprises and governments, the scarce asset is not the model. It is control: where the model runs, which data touches it, who can see the answers, and whether all of that respects a country's or a company's borders. That reservation of control is what "sovereign AI" means — models deployed inside the organization's own walls under its own rules. Palantir is selling the harness that manages it, positioning itself as the orchestration layer rather than a model maker. Its new partnership with NVIDIA applies exactly this: a supply-management system for NVIDIA, built to run AI inside a critical supply chain rather than hand a foreign model the data.

This is the same squeeze the AI trade has been missing. The market's default story frames Palantir as a hypervalued fad. Set that against the actual numbers, quarterly, and the picture inverts. In the second quarter revenue grew 93% year over year to $1.94 billion. U.S. commercial revenue grew 149%. U.S. government revenue grew 90%. Gross margin sits near 85%. And it is profitable — not pro-forma profitable, GAAP profitable. It raised full-year guidance again, to roughly $8.15 billion, an 82% jump. There is nothing speculative in that run of digits. The growth is measurable, and it is accelerating in exactly the sovereign, control-demanding segments the conference was about.

A compression, or a ceiling

Now the counterargument, because it could change the trade. All of that growth still leaves the stock enormously expensive. Market cap is around $400 billion against roughly $8 billion of forward revenue — a price-to-sales multiple in the high 40s. On earnings it trades near 74 times forward profit. Whatever Palantir does, a multiple like that assumes years of near-perfect execution, no big competitive threat, and no stumble in contract timing.

And there is a real tension inside the very guidance the bulls celebrate. Q2 grew 93%. Full-year guidance implies 82% growth. That is not a rounding error; it is a step down, and the gap widens in later quarters as the base grows. An 82% full-year average means the exit rate is below the quarter you just saw. For a business priced for acceleration, deceleration is the thing to watch, even when the absolute growth is still stunning.

The deeper issue is that the market already believes the sovereign-AI story. The stock fell about 29% year to date even as earnings exploded, which means the forward multiple compressed dramatically rather than expanded. A stock that goes down while revenue grows 93% is not getting more expensive, it is getting cheaper — and the market was paying that price precisely because it had already loaded the expectation into the multiple. That is not a mispricing staring you in the face. It is the correct market doing its job: growth priced in years ahead.

What the bet really is

Strip the conference theatrics away and the honest read is this. DA Davidson's $250 is a bet not that AI grows — everyone believes that, and believing it is worth nothing on its own — but that Palantir owns the scarce layer within it, the control plane, and that governments and companies will keep paying premium rates to reserve their own AI. The revenue data supports the mechanism so far. The valuation says the market has already paid for it.

The judgment that follows is a contradiction you have to hold at once. Palantir is compounding faster than almost any large company on Earth, and it is doing it in the part of the economy where control is becoming scarce. That combination is real. It is also fully priced, and the multiple leaves almost no room for the guidance to disappoint even slightly. A price target says where one analyst thinks the stock goes. The numbers say the duration of the conviction is now the whole game.

What would change that? Watch the exit-rate deceleration, and whether growth holds into next year as the base doubles. If revenue keeps compounding at the rate the guidance promises, the multiple does the work of catching up to earnings and the stock finds a floor. If the quarter after next comes in short of the trajectory, an 80-times-earnings stock has a long way to fall before the price resembles anything like value. That is the risk the $250 target leaves out of the headline.