ServiceNow's Moat Is Widening — and Getting More Expensive by the Quarter

On November 5, 2003, a 49-year-old engineer sat down at a desk in his own house with a single laptop and started writing code, because his fiftieth birthday was nineteen days away and he had decided he was running out of time.⁶ Fred Luddy had just watched his net worth — roughly $35 million in company stock — go to zero when Peregrine Systems, where he was CTO, collapsed into bankruptcy amid an accounting fraud.⁹ What he built next, first incorporated as Glidesoft, would become ServiceNow.⁵ Two decades later that home-desk experiment books $10.98 billion in annual revenue,¹ and the more interesting number is the one it hasn't earned yet.

The official story is that ServiceNow is an IT help-desk company — the place where you file a ticket when your laptop dies. That description was roughly true around 2011 and is now actively misleading. ServiceNow stopped being a tool a long time ago. It became the layer that work itself runs on, and it is now trying to become something larger and far more expensive.

The $22 billion the customers haven't paid yet

The cleanest way to read a lock-in is to look at what customers have already committed to but not yet spent. As of December 31, 2024, ServiceNow carried $22.3 billion in remaining performance obligations — contracted revenue, signed, sitting in the backlog waiting to be recognized — up 23% in a single year.² That figure is twice its entire annual revenue. It is the sound of an enterprise that has decided, for years in advance, that it is not going anywhere. And it concentrates upward: 2,109 customers now pay ServiceNow more than $1 million a year, and nearly 500 pay more than $5 million.² You do not write a $5 million annual check for a ticketing system. You write it for the nervous system of your operations.

Here is the mechanism, worked down. ServiceNow's stickiness has nothing to do with screens being nice. It comes from the platform becoming the system of record for how work happens — the approval chains, the escalation paths, the SLAs, the integrations across IT, HR, security, and customer service. A company can switch a CRM the way it switches a phone plan: painful, but bounded. It cannot switch the layer where every department's process logic is encoded without re-engineering the company itself. The cost of leaving is not the software license. It is the operational re-architecture, and that cost rises every quarter the platform absorbs another workflow. That is the whole engine: ServiceNow doesn't lock you in by being indispensable; it locks you in by becoming the place your indispensability is defined.

Buying the AI layer before it eats the platform

A moat this real should let a company relax. ServiceNow is doing the opposite. In March 2025 it agreed to acquire Moveworks for $2.85 billion in cash and stock.³ The logic is defensive in the most aggressive way: Moveworks had built an agentic, conversational AI front end that reached nearly 5 million employee users in about eighteen months, with roughly 90% of customers rolling it out to every employee — and the two companies already shared about 250 customers.⁸ Both companies frame Moveworks as a complementary front-end AI assistant rather than a captured rival — but the strategic logic is plain: ServiceNow moved to own the conversational layer sitting on top of its own platform, the interaction point that increasingly mediates between the user and the workflows underneath.¹¹

Read that line carefully, because it contains the bet. ServiceNow's thesis is that raw intelligence — the language models everyone is racing to train — is commoditizing, and that the durable advantage lies one level up, in the autonomous workflows that decide what an AI agent is actually allowed to do inside a regulated enterprise.¹¹ If that's right, ServiceNow doesn't need to win the model race. It needs to own the orchestration layer, the place where models get pointed at real processes with real permissions. The moat widens from 'where work is recorded' to 'where AI is allowed to act.' That is a much bigger pond. It is also a much more crowded one.

The honest objection: maybe the moat is buying itself a problem

The fair counter is that none of this is fragile — that ServiceNow is simply a category leader making sensible tuck-in acquisitions, that 22% revenue growth on an $11 billion base with $1.43 billion of net income is a company executing, not flailing.¹ That's true, and it's the strongest version of the bull case. But it sidesteps where the strategy actually strains. ServiceNow is doing two hard things at once. It is paying 20x revenue to graft on an AI layer whose data model fights its own,⁴ and it is pushing outward from IT workflows into the CRM market, taking direct aim at Salesforce and other incumbents entrenched in exactly that territory.¹⁰ A moat earns its keep when defending it is cheaper than the rent it collects. ServiceNow's rent is enormous — but the bill for defending it just moved from cash flow to stock dilution and integration debt, and that is a different kind of moat than the one the RPO number describes.

Fred Luddy started ServiceNow because he was about to turn fifty and couldn't stand the idea of not building one more thing.⁶ What he built turned out to be unusually hard to leave — a layer the enterprise records its own logic onto, then can't unwind without rewiring itself. That moat is genuine, and it is still widening. The open question is the one the backlog can't answer: ServiceNow is now spending real money and real equity to own the place where AI gets to act, on the bet that orchestration outlasts intelligence. It might. But for the first time since that desk in 2003, the moat is being widened on credit — and one of the most expensive ways to defend a fortress is to keep buying up the land outside its walls.