Its customers renew 98% of their contracts every year. The market took two-thirds off the shares.
The argument was clean and it is worth stating properly, because it is not silly: this company charges by the seat, seats exist because people do the work, and if software agents do the work then the seats go. Price per person, in a world with fewer persons.
Our report's conclusion was four words long: the customers did not leave.
Retention is the only number that can answer this
Revenue growth cannot settle it — revenue can be bought, and this company spent $11.7 billion on acquisitions. New customer counts cannot settle it either; they say nothing about whether the existing ones are being displaced.
Renewal and retention can. If agents were replacing the seats, it would appear first as customers renewing for fewer of them, and the series would bend before anything else did.
| What it measures | What we found | |
|---|---|---|
| Gross renewal | Do they stay at all? | 98% — implying an average relationship of around fifty years at that rate |
| Net retention | Do they spend more or less than last year? | 157% at Palantir — the opposite of displacement |
| Contracted order book | What has just been signed, before revenue sees it | Adobe: decelerated from 13% to 8% — this is a real signal |
| Subscription margin | What it costs to serve them | Slipped — AI work adds computing cost |
A company losing its customers to a new technology does not renew 98% of its contracts. Whatever is happening, that is not what is happening.
It is a cost story, not a demand story
Here is what actually moved: the subscription margin slipped, because AI work added computing costs.
That is a different and much more boring problem than disruption, and it is also a real one. Software's whole appeal to investors is that serving the hundredth customer costs almost nothing. When each interaction consumes computing time that must be bought, a little of that property goes away.
So the honest summary is not "the fear was wrong". It is: the fear was about the wrong line. The market priced an exodus of customers and what arrived was a margin that is a few points harder to defend.
Where the warning did show up
Our September report found revenue accelerating to +12.9% while the contracted order book decelerated from 13% to 8%, the gross margin broke an eight-quarter band, and net new recurring revenue fell by about a third.
Those two series measure different moments. Revenue is what was signed before, arriving now. The order book is what is being signed now. When the leading series slows while the lagging one rises, the headline is describing a year that has already finished.
That is what a genuine demand problem looks like in the numbers — and it looks nothing like a 98% renewal rate.
What two-thirds off does and does not buy you
A fall of two-thirds needs a threefold rise to undo. That is why these repricings matter even when the fear turns out to be overdone — and why it is worth being precise about what you are getting at the end of one.
Counting stock-based pay as the real cost it is, our report put the shares at 44 to 60 times owner earnings for a business growing near 20%. That is a fair price, not a bargain — and the gap between those two words is the whole reason our conclusion was to start small and add below $115 rather than to buy the dip.
Note the adjustment, because it is where most analysis of this sector goes wrong: the adjusted margin is 29.5% and the GAAP margin is 4%. Almost all of that gap is pay settled in shares, and it is a real cost that a shareholder pays in ownership rather than in cash.
How to check this for anything you own
Most software companies disclose it. A number above 100% means existing customers are spending more than they did a year ago — which cannot be true of customers being replaced. The trend matters more than the level.
If the order book is growing slower than revenue, revenue is going to slow, and you have roughly a year's notice. This single comparison is worth more than any amount of argument about the technology.
The disruption that is actually arriving shows up as cost of serving customers, not as customers leaving. A software gross margin drifting down two or three points is the most underreported development in this sector.
Renewal below 95%. Net retention below 100%. Two consecutive quarters of order book growth under revenue growth. Write the numbers down now, while nothing is happening, and you will not have to decide in the middle of a fall.
One whose customers stayed, one whose order book slowed, and one whose customers spend half as much again each year.
Including the case for and against AI agents under three different futures.
Frequently asked
Will AI agents destroy per-seat software pricing?
It has not happened yet in the numbers we can see. ServiceNow's customers renewed 98% of their contracts while the market took two-thirds off the shares on exactly that fear. What did change is the cost side — the subscription margin slipped because AI work adds computing costs. So far the threat is visible in the multiple and in the cost line, not in customer behaviour.
What is the best number for checking software disruption?
Net revenue retention, and the gross renewal rate underneath it. Retention above 100% means existing customers are spending more than a year ago, which is incompatible with being replaced — Palantir's was 157%. A renewal rate of 98% implies an average relationship measured in decades. Revenue growth can be bought; retention cannot.
Why did Adobe's order book slow while revenue accelerated?
Because they measure different moments. Revenue is what was contracted in the past being recognised now; the contracted order book is what has just been signed. Adobe's revenue accelerated to 12.9% while the order book decelerated from 13% to 8% — the leading series slowing while the lagging one rises is the warning, and it arrives a year before it reaches the headline.
Is ServiceNow cheap after falling two-thirds?
Our October 2026 report said fair, not cheap. Counting stock-based pay as the real cost it is, the shares were 44 to 60 times owner earnings for a business growing near 20% — which is a reasonable price for quality rather than a bargain. Our conclusion was to start small and add below $115.