ServiceNow, Inc. NOW

Key data

The business

ServiceNow sells one thing, subscriptions to the Now Platform, the workflow system where an enterprise routes its IT, HR, customer-service and security work. About 97% of revenue is recurring subscription, billed a year up front, and the company runs as a single reportable segment. The moat is switching cost: once a bank or an airline builds its ticketing, approvals and security workflows on Now, ripping it out is a multi-year project nobody volunteers for, which is why renewals hold and prices rise. The current chapter is Now Assist, the AI-agent layer sold on top of the platform. It is landing on the top line and, so far, punishing the cost line. This is the same setup as every software vendor that bolted inference onto a subscription: the demand is real, the unit cost of serving it is the question.

The numbers

Subscription revenue compounds cleanly; profit does not. The last five quarters:

Revenue never stopped, up 24% year on year in Q2. Net income fell to $298M from $385M a year earlier, and operating income collapsed to $162M, a 4.1% margin, from $358M and 11.1%. The cause is one line: subscription cost of revenue jumped to $1B from $625M, up 65% against 24% revenue growth, as the GPUs behind Now Assist showed up in COGS. A watch-item flagged last time was whether gross margin would turn back up. It did not. It fell again, from 75.1% in Q1 to 70.7% in Q2, so that earliest read is bending against the bull, not for it.

The five fiscal years show the operating leverage the AI cost is now interrupting:

Operating margin climbed for four straight years to 13.7%, then fell back to 8.6% in the first half of 2026. The sector lens says watch subscription gross margin, and it tells the whole story:

Six hundred basis points gone in a year, entirely on cost, with the revenue growing the whole time. Free cash flow, the number that actually matters here, still compounds: roughly $1.8B in 2021 to $4.6B in 2025, about 26% a year, near $4.43 per share. But the reported 34% cash-flow margin flatters the owner. Stock-based compensation runs 14.8% of sales, so the cash that reaches a shareholder after dilution is closer to 20% of revenue. What this memo believes the market underweights: the margin drop is a deliberate purchase of AI attach, not a demand crack, and the single print that settles it is subscription gross margin turning up while growth holds.

Management

Bill McDermott (CEO not independently verified this run) runs a company that just made two large bets with borrowed money. Long-term debt went from $1.5B at year-end to $5.4B by June, with another $2.1B short-term, roughly $7.5B total debt outstanding to fund the Armis and Veza security acquisitions and to keep buying stock. Buybacks ramped hard: $0.5B in 2023, $0.7B in 2024, $1.8B in 2025, and $2.2B already in the first half of 2026, much of it at 8 times forward sales, which is defensible, and at 88 times earnings, which requires believing the earnings recover. The guidance record is a metronome of small EPS beats, four straight quarters. Insider selling is noise: no buys, about $2.1M sold across five officers over twelve months on a $146B company, the largest a $1.08M sale in August, plan status not disclosed. Routine comp-driven trimming, nothing to read into.

How it fails or surprises you

AI compute cost outruns AI revenue (downside). Subscription COGS grew 65% against 24% revenue, and subscription gross margin fell to 73.4% from 79.9%. This is the fact the case for owning it explains least well. If serving Now Assist keeps costing more per dollar than it earns, the 8.6% operating margin has further to fall. The print: Q3 subscription gross margin.

AI monetization re-rates the stock (right tail). Subscription still compounds 24% with Now Assist attach rising. If those agents convert to seat-plus-usage pricing and bookings reaccelerate, the market is currently paying a decade-low 8 times forward sales for accelerating growth. The print: remaining-performance-obligation growth pulling ahead of revenue growth next quarter.

Debt and integration bite at the wrong moment (downside). Net debt went from near zero to about 1.7 times EBITDA to fund Armis and Veza while margins were compressing. If integration stumbles or interest expense crowds out buybacks, the per-share compounding slows. The print: interest expense and whether the $2B-a-quarter repurchase pace holds.

Closing thoughts

Subscription gross margin next quarter tells you everything. Nothing about the demand side is broken: subscription revenue grew 24%, deferred revenue is healthy, renewals hold. The entire debate is whether AI helpers earn more than they cost, and it resolves on one line. That makes the setup unusually clean, and it means the person on the other side of you is selling because GAAP earnings look ugly and the chart is down 27% from its high. The fatter tail is up, but only modestly, because the downside is real and near-term: if margin keeps sliding, the 4% operating quarter is not the bottom. What is at risk if the downside linchpin breaks is a full year of depressed earnings and a stalled buyback; what the upside is worth is a re-rate from 8 times forward sales toward the 12-to-17 it normally commands, on a business still growing a quarter a year.

The bet is still that ServiceNow stays the system big companies run their work on, and that its AI helpers earn more than they cost to run. What breaks it is the second half of that sentence. The one pair that tells you first: next quarter's subscription gross margin against subscription revenue growth. If margin keeps falling while growth holds, the problem is cost, not demand, and the read is wrong.