ServiceNow fell 6.47% on July 22 and closed at $95.46 hours before it reported one of the strongest quarters in its history. Almost nothing in that selloff was about ServiceNow. A Pegasystems miss dragged the enterprise software complex, IBM had warned on July 14 that clients were rotating budget out of software and into AI hardware, OpenAI had just launched Presence, an enterprise agent product that reads uncomfortably close to the core workflow business, and CLSA had come in with an Underperform at $72. The stock was trading as a proxy for a narrative, not as a claim on a cash flow stream. Then the numbers landed and it recovered above $100 after hours. It closed at $115.76 on July 29, roughly 20% above the pre-print price.
The narrative being proxied deserves to be stated precisely, because it is falsifiable and it was falsified. The agentic disruption thesis holds that autonomous agents substitute for the human seats that per-seat enterprise software is priced against, so the customers who deploy agents most aggressively should be the customers whose contract values shrink first. That is a prediction, and it runs in a specific direction. ServiceNow’s second quarter runs in the opposite one on every available cut of the data.
AI annual contract value crossed $1 billion, arriving roughly a quarter and a half ahead of the original year-end target, with net new AI ACV accelerating more than 40% sequentially. First-time agentic buyers grew 45% year over year. Customers running agentic AI in production grew ninefold over nine months. The accounts furthest into agent deployment are the accounts spending more, not less. That is substitution running backwards, and it is the reason the AI ACV headline carries weight beyond its size. As a share of a book that will recognize roughly $15.77 billion in subscription revenue this year, $1 billion is six to seven percent. It matters for its slope and its direction, not its magnitude.
The harder number is the renewal rate, and it gets almost no coverage. ServiceNow renewed at 98% in the quarter. Disruption theses die at the renewal line before they die at the growth line, because an enterprise that has genuinely replaced a workflow platform with an agent stack does not renew quietly at a smaller number. It leaves. Nothing left. Nor is the remaining book fragmenting: 123 transactions cleared $1 million in net new ACV, up nearly 40% year over year, and the company ended the quarter with 658 customers above $5 million in ACV, up 23%. Spend is concentrating into larger and longer contracts at the top of the customer base. Erosion does not look like that.
The competitive question underneath all of this is whether the moat survives a change in what enterprise software is for, and the answer emerging from this quarter is that the moat was never the interface. It was the configuration management database and the permission layer sitting on top of it. ServiceNow does not replace a customer’s systems of record; it sits above them and knows what every asset, identity, approval path and dependency is. That inventory is exactly what an agent needs before it can be allowed to act, and exactly what an agent cannot bootstrap for itself. AI Control Tower is built on that position, it governs third-party agents rather than only ServiceNow’s own, and management pointed to a kill switch for rogue agents days after OpenAI disclosed that one of its own agents had escaped a test environment and reached another company’s infrastructure. Security and risk is now the fastest-growing business among the top ten enterprise cybersecurity companies, which is what the governance layer monetizing looks like on an income statement. The strategic consequence is that agent proliferation, including proliferation by ServiceNow’s competitors, is a revenue input rather than a substitution risk. Every agent an enterprise deploys is another object requiring inventory, permission and revocation, and the toll is collected where the workflow data already lives.
Profitability is the part of the print that speaks directly to the internal redeployment question. Non-GAAP operating margin came in at 29.5%, three points above guidance, while the company absorbed Armis, Veza and Moveworks. Those are the same acquisitions that forced two margin guidance cuts on the April call and crashed the stock despite a clean revenue beat. This time both lines held: 31.5% operating margin and 35% free cash flow margin for the full year. April was a beat and a crash on margin cuts. July was a beat and a rally on margin holds. The demand picture was similar in both quarters, which tells you the market has been pricing whether ServiceNow could deploy AI internally fast enough to defend its own cost structure, not whether customers still wanted the product. Management’s commitment to exit 2026 at the same headcount it carried before three acquisitions is that claim stated with a date attached.
The cost of riding this wave is visible and should not be waved off. Full-year subscription gross margin guidance came down to 81% from 81.5%, attributed to greater hyperscaler utilization and accelerating AI adoption. AI revenue is structurally more expensive to serve than seat revenue was, so the mix shift the bulls are celebrating is mildly gross-margin dilutive by construction. A regulatory filing after the print disclosed $4.8 billion in total cloud infrastructure commitments through 2030, including $1.2 billion with Google. Free cash flow margin printed 16% against 44% in the first quarter, which is seasonal and reverses, but will be quoted out of context.
The genuine soft plank is the backlog. Current remaining performance obligations reached $13.20 billion, up 21.5% in constant currency against a 19.5% guide, a two hundred basis point beat that did not decelerate from the first quarter’s pace. Morgan Stanley had published days before the print that stripping acquisition contribution put organic growth implied by guidance closer to 17%, and that informed positioning already expected 20.5% to 21.5% headline. The print cleared the official bar and roughly matched the real one. Roughly half the revenue beat came from federal on-premise business pulled forward out of the third quarter, which is timing rather than demand. The full-year raise was $15 million at the midpoint on a $15.77 billion base, which is management declining to spend the beat. Third quarter cRPO is guided to 20% in constant currency against a $35 million foreign exchange headwind.
At $115.76 the stock carries a market capitalization near $119 billion and trades around 25 times forward earnings, sitting 42% below its split-adjusted 52-week closing high of $201.15 and 42% above the $81.24 low set on April 10. The dispersion in published targets is the widest in large-cap software: Morgan Stanley at $180, DA Davidson at $170, Citi at $156, JPMorgan at $145, Cantor at $141, Oppenheimer at $140, against UBS at $115, KeyBanc at $85 and CLSA at $72. Consensus sits near $140. Twenty-five times forward earnings for a business compounding subscription revenue at 21% in constant currency with a 35% free cash flow margin is a multiple assigned to something the market expects to stop compounding.
Base case, the stock works from here through multiple repair rather than estimate revision, moving into the $135 to $150 band as organic cRPO holds near 17% and the AI ACV line clears $1.5 billion on schedule. Bull case takes it toward $170 to $180 and requires the governance layer to become a distinct disclosed revenue line, security and risk to keep outgrowing the core, and the headcount commitment to be met so 2027 margin guidance opens above 33%. Bear case is not a company-specific unwind. It is cohort derating: another IBM-style capex rotation datapoint, a hyperscaler bundling agent governance into an existing enterprise agreement, or an OpenAI enterprise product landing a marquee displacement, and the whole complex compresses to 18 to 20 times forward regardless of what ServiceNow prints. That path retests $90.
The disruption thesis and the deceleration thesis are not the same thesis, and the market has been pricing them as one. This quarter closed the first. The second is untouched and always was a question about compounding at $16 billion of scale rather than about artificial intelligence. Everything that matters between now and late October reduces to one figure: whether third quarter cRPO clears its own 20% guide the way this quarter cleared 19.5%, measured after the acquisitions are stripped out.