ServiceNow

IThe business

ServiceNow began as a better way for an IT department to log and close tickets. An employee’s laptop breaks, a ticket opens, someone is assigned, the fix is recorded. That is IT service management, ITSM, and it still sits at the centre of the company. In the third quarter of 2026 it was in 11 of the top 20 deals.¹⁶ Fred Luddy, the founder, still owns 768,565 shares.¹³

What grew out of the ticket is a general engine for moving work between people and systems. A request comes in, the software knows who owns what, who must approve it, which systems it touches, and what happens next. The same engine now runs IT operations, IT assets, security response, HR cases, procurement, legal requests and customer service. The 10-K groups these into Technology, CRM and Industry, Core Business and Creator products.⁶ Each customer gets its own application layer and database, a design the company calls multi-instance.⁶ That costs more to run than one shared database, and it is part of why large regulated customers trust it.

Revenue comes from subscriptions, usually on multi-year contracts paid a year in advance. Subscriptions were 97% of revenue in 2025. Services are a small remainder, and they now lose money on purpose: the services gross margin was −32.5% in the third quarter, as the company pays people to get AI live inside customers.²⁶ At the end of 2025 there were about 8,700 customers, of which 603 paid more than $5M a year. That number is now 680.⁶⁴ About 63% of 2025 revenue came from North America, and the United States is 94% of that.⁶

The numbers that matter

Two measures of backlog run this stock. Remaining performance obligations, RPO, is all contracted revenue not yet recognised: $29.7B at 30 September, up 22%. Current RPO, or cRPO, is the part due in the next 12 months: $13.65B, up 20.5%, or 21% at constant currency.² cRPO is the closest thing a software company has to next year’s sales already in hand, which is why a small change in its growth can move the shares a lot.

Annual contract value, ACV, is the yearly value of a contract. The renewal rate is the share of expiring contract value that stays. It was 98% in the third quarter, up from 97% a year earlier.⁴ This is the single best fact about the business, and later I will explain why it is less comforting than it looks.

The bet the company has made

In 2026 ServiceNow changed what it says it is. The release now calls it “the Super Intelligence Control Tower for business reinvention”.² Strip the words away and the claim is simple. Companies will run thousands of AI agents from many vendors. Someone has to know which agents exist, what they may touch, what they did, and how to stop them. ServiceNow says that someone is ServiceNow, because it already holds the record of the company’s systems and processes.

To back the claim it spent about $11.2B on three companies that closed between December 2025 and April 2026. Moveworks, an employee assistant and search product, closed on 15 December 2025 for $2.4B. Veza, which maps who and what can reach which data, closed on 2 March 2026 for about $1.2B. Armis, which finds every device on a network, closed on 20 April 2026 for about $7.6B in cash.⁴ Armis was reported to have passed $340M of annual recurring revenue when the deal was announced, so the price was about 22 times that figure C (calculated).¹⁸ Before this, ServiceNow’s largest purchases were small. From 2016 to 2025 it spent $2.6B on acquisitions in total.¹⁷

The company also changed how it charges. Customers now buy seats plus a bundle of “assists”, units of AI work. Management said in July that 50% of net new business is already not based on seats, and that new AI product tiers lift price by 20% to 30%.¹⁹ It targets $1.7B of AI ACV by the end of 2026, up from a $1B target in January.² It counts only the extra contract value from AI, not the seats sold with it.⁸

What happened on 29 October

The third quarter beat the company’s own guidance on every top-line and margin measure.² It raised its AI target and its free cash flow margin target. The next day the shares fell 26.95%, the worst day in their history, from $150.13 to $109.67, and they fell another 4.1% on 30 October.¹⁹ A week earlier they had closed at $158.96.¹

The trigger was one line of guidance. cRPO growth was guided to 18.5% for the fourth quarter, about a point or more below what analysts had, depending on whose consensus you read, after 21% in the third.²²⁰⁹ Adjusted earnings of $1.00 also came in two cents light.⁹ The real reason a one-point gap took out $42B of value C (calculated) is what the gap might mean. Investors had spent a year asking whether AI would shrink seat-based software. Management had answered that AI would speed growth. Here was a guide that showed growth slowing just as AI net new contract value rose 65%.¹⁶²

Citi’s Tyler Radke called it “a little bit uncharacteristic for ServiceNow standards”, CNBC reported.⁹ I will take this apart in Section VII, because I think the headline number overstates the slowdown, and the truer number still isn’t comfortable.

IIThe moat

A durable business keeps its customers and its pricing without having to spend more each year to do it. On the first test ServiceNow is excellent. The second is now being tested for the first time.

Evidence it is wide

The renewal rate is the hard evidence. It was 98% in each year from 2022 to 2025, and 98% again in the first nine months of 2026.⁴⁶²¹ The figure is measured on contract value and excludes price and user changes, so it says customers stay. It doesn’t say they buy more.

Why they stay is plain enough. Over a decade a large company pours its own rules into this system: its approval chains, its asset records, its escalation paths, its links into SAP, Workday and a hundred others. The configuration management database, the CMDB, is a live map of what technology the company runs and how the pieces depend on each other. Rebuilding that elsewhere means a project with a real chance of failure and no direct gain. The person who approves that project gets no thanks if it works and gets fired if it doesn’t.

Contract length is going up. RPO grew faster than cRPO, 22% against 20.5%, and management says average duration is rising.²¹⁶ Customers don’t usually sign longer with a vendor they plan to leave. The number of customers paying over $5M rose about 23% in a year.² Nineteen of the top 20 deals in the quarter included six or more products.¹⁶

Evidence it is narrowing

The first crack is in gross margin. Non-GAAP subscription gross margin has fallen from 85% in the third quarter of 2024 to 80% in the third quarter of 2026.²²² Some of that is the cost of running acquired products. Some is customers choosing to run on the hyperscalers’ clouds. Some is AI compute. A protected business doesn’t usually give up five points in two years, even for good reasons.

The second is who owns the screen. An employee who used to open a ServiceNow portal may now ask a question in ChatGPT, Claude, Gemini or Copilot. The agent then goes and does the work through ServiceNow’s connections. If that becomes normal, ServiceNow keeps the plumbing and loses the front door. The company sees this. It launched EmployeeWorks as its own front door, reached $100M of ACV in under three quarters by its count, and it also built a ChatGPT plugin and a Gemini link so it can be called from inside theirs.¹⁶² Both moves are sensible. The second one admits the risk.

The third is the question no filing can settle. Software that charges by the seat wins when customers hire more people. A level-one service desk agent that resolves incidents 85% faster than a person, as management described on the call, is a good product.¹⁶ It also means a customer needs fewer people on the service desk, and those people are seats. The renewal rate tells you the customer stays. It can’t tell you how many seats renew.

My judgement: the switching cost is real and will last. Whether the pricing power lasts, I can’t yet say. A company that keeps 98% of contract value can still shrink per customer if each renewal comes with fewer seats and consumption doesn’t fill the gap. The evidence so far says it is filling the gap. There hasn’t been a recession or a real AI seat cut to test it.

IIIThe competitive advantage

The 10-K names Microsoft, Oracle, SAP, Salesforce and Workday as rivals, then adds new AI entrants, in-house builds, consultants and integrators.⁶ It also says plainly that its markets have low barriers to entry.⁶ I doubt many customers would put it that way about ITSM, but the sentence is there. The table compares the listed companies that sit closest to it, using the same simple owner measure for each.

The table says two things at once. ServiceNow is the fastest grower of the large names here, about twice the pace of Salesforce. It also costs far more per dollar of owner cash. Adobe’s market value is now smaller than ServiceNow’s, on 3.5 times the owner cash. Investors are paying for the belief that ServiceNow is on the right side of AI and Adobe is not. That belief may be correct. It is still a belief.

Where ServiceNow wins

Neutrality first. Salesforce wants the customer record. Workday wants the employee record. SAP wants the ledger. Microsoft wants everything. ServiceNow sits across them and connects them, and it has built its business on not being the system a customer must abandon. That is why OpenAI, Anthropic, Google and NVIDIA all appear as partners in its October release.² A partner list proves little by itself, but a vendor that threatens no one’s core system gets invited in.

The record of the estate second. An AI agent acting inside a company needs to know what systems exist, who owns them, what depends on what and who may approve a change. The CMDB holds much of that, along with 20 years of process history. Armis adds a live list of devices, Veza adds who can reach what.¹⁶⁴ Whether that bundle is worth what the company paid is a question for Section IV. The strategy itself holds together. In security the company says its products were in 16 of the top 20 deals.¹⁶

Distribution third. Management says quota-carrying salespeople are up by double digits even as total sales and marketing headcount falls.¹⁶ Sales and marketing cost 47% of revenue in 2016 and 32% in the last twelve months C (calculated).¹⁷ A new AI entrant can build a good agent. It can’t easily build that sales force or pass a large bank’s security review.

Where it loses

Microsoft is the serious one. It owns the employee’s desktop, the identity system, the email, Teams and Azure, and it bundles. A customer who already pays for Copilot has a reason to ask what ServiceNow’s assistant adds. In CRM, ServiceNow is the challenger: it has built about a $2B ACV business there and describes customers ripping out incumbents, but that claim comes from management and I can’t check it.¹⁹¹⁶ In security it is entering a crowded field with products it bought this year.

The AI labs are the unknown. Today they are partners. A partner today can be a competitor in three years, and the labs have every reason to move up from models into work itself. ServiceNow’s answer is that a model can reason but can’t act safely inside a company without its rails. That is a good argument. It is also the argument every incumbent makes.

In-house building is the last. The 10-K lists custom and in-house solutions as competitors.⁶ If AI makes software much cheaper to write, some customers will try to build their own simple workflows. Management’s answer in July was a question: which of the alternatives has a system of record, governance and global distribution?¹⁹ I think that is right for the largest customers and less right for the rest. ServiceNow launched Flow in October, a simple service desk that runs outside the main platform, which suggests it sees the threat at the lower end too.²

IVManagement

The question I care about here is what these people are paid to do, and how that compares with what an owner wants.

McDermott ran SAP as chief executive or co-chief executive from 2010 to 2019 before he came here.¹³ He has signed on to stay through at least the end of 2030, as chief executive, co-chief executive or chairman, at the board’s discretion.³⁰ So the next four years are his, and the board has already written the words “co-CEO” and “Executive Chairman” into the contract. Succession is a known subject.

How they are paid

McDermott’s 2025 pay was $51.6M, 251 times the median employee’s $205,259.¹³ About $43.5M of it was stock. Follow the measures. The annual bonus is 70% net new annual contract value and 30% non-GAAP operating margin, with a 10% adjustment either way for non-financial goals. Long-term pay is 60% performance stock and 40% stock that vests with time. The performance stock pays on one number, 2027 non-GAAP subscription revenue at fixed currency, with a plus or minus 20% adjustment for share return against the S&P 500. Nothing is earned below 96% of target.¹³

What does that pay them to do? Sell, and get revenue up. Those are sensible things to pay a software team to do. Look at what isn’t in there. Operating margin is measured before stock pay, so issuing more stock to employees costs the bonus pool nothing. Subscription revenue, as the proxy defines it, is GAAP revenue at fixed exchange rates. I found no language in the proxy that removes acquired revenue from the target, and the targets aren’t published. If they weren’t reset, buying Armis helps the 2027 number. If they were, the proxy doesn’t say so. Either way, no measure charges management for the $11.2B of capital. Return on that capital appears nowhere in the plan.

Shareholders noticed the pay. At the May 2026 meeting 110.9M shares voted against the say-on-pay proposal and 654.7M for it, about 14.5% against C (calculated).³¹

The 2021 award

In 2021 the top team received an unusual grant: stock options that vested only if subscription revenue targets and share price targets were both met by 30 September 2026. The top of the ladder needed a share price above $290.40 sustained over six months.¹³ It was a real test and I give the board credit for setting it. Six revenue targets were met, the sixth certified in April 2026. Four price targets were met, the last at $198.¹³³² The 10-Q confirms the outcome: the period ended on 30 September, and only the first four tranches vested.⁴

So McDermott holds 1,387,680 vested options with an exercise price of $139.55, which are $34.43 under water at $105.12. The other 1,387,705 did not vest. He also has 642,585 options at $53.26 from when he joined, worth about $33M C (calculated).¹³ On a pure incentive reading this is healthy. He now has money at risk in the same direction as a shareholder who bought in 2021.

A new grant, four weeks later

The proxy says the board committed not to grant one-time equity awards to any named executive holding a 2021 award while its performance period was running.¹³ That period ended on 30 September. On 26 October, two days before the third-quarter report, McDermott was granted 349,900 restricted stock units.¹⁴ They vest 25% at the end of 2027, 35% at the end of 2028 and 40% at the end of 2029, and the Form 4 attaches only one condition: that he is still employed.¹⁴ At the $158.18 close that day they were worth about $55M. At $105.12 they are worth about $37M C (calculated).¹

I don’t know the board’s reasons. The proxy that would explain them comes out next spring. The facts I can see are these. A grant that pays for performance lapsed in part. Within four weeks it was followed by a large grant that pays for staying. McDermott was not alone: the Form 4s filed the same day show five other officers received RSUs on the same schedule, from 79,078 shares for the chief legal officer and the chief people officer to 174,950 for Zavery, 157,455 for Fipps and 124,215 for Mastantuono.⁴⁰ So this looks like a team-wide retention grant, and his is the largest. His vesting dates match the contract that runs to 2030.³⁰ If the board’s view is that McDermott is worth keeping, and his record of growth is a fair case for that, the cheaper way to say so would have tied it to the share price or to cash per share. This one doesn’t.

How much of their own money is in

McDermott directly holds 184,199 shares and 24,405 through a trust, after the August vest, about $21.9M at $105.12.²⁹ That is a little under half of one year’s pay. It clears the guideline of six times salary, but it isn’t large for a man paid over $125M in three years C (calculated).¹³ He has bought, too. On 27 February 2026, after the January fall, he bought 28,682 shares on the open market at about $104.60, roughly $3.0M of his own money.²⁹ In 2025 he sold at least 90,440 shares, on a split-adjusted basis, at about $180 to $204, about $17.4M, in a series of small sales.²⁹ The sales look like routine selling of vested stock. The purchase was a choice, made with his own money, and I count it in his favour. I found no open-market sales in 2026. Luddy’s 768,565 shares are worth about $81M C (calculated).¹³ No one at the top owns enough for the company’s fate to be their own. BlackRock and Vanguard held 8.8% and 8.7% at the proxy date.¹³

What they did with capital

Here I look at the full record, not one good year. From 2016 to 2025 the company produced $18.14B of free cash flow. It gave $10.62B of that back to staff in stock pay C (calculated).¹⁷ That leaves $7.51B for owners over ten years.

Since 2023 it has spent $5.30B on buybacks and another $2.38B paying tax on employee stock, a total of about $7.7B C (calculated).¹⁷² Diluted shares rose from 1,028M in 2023 to 1,043M in the third quarter of 2026.¹⁷² So $7.7B did not shrink the share count. It stood still against stock pay and the shares issued in the deals. I think of that as a cost of doing business.

Buyback timing has been ordinary, neither clever nor foolish. The company paid about $185 a share in the first nine months of 2025 C (calculated). It bought $2.0B through an accelerated program in early 2026 at an average of $107.97, about what the stock fetches now.⁴ It then bought nothing in the third quarter, and it has $4.2B of authorization left.⁴ At the end of 2025 it had no short-term debt and $1.5B of notes. At 30 September it had $7.6B of debt, including $2.1B of commercial paper due in about two months, and cash and securities of $7.0B.²⁴ Net cash of about $8.6B became net debt of about $0.6B C (calculated). It is not a stretched balance sheet. It is a different one.

The acquisitions are the real test and the verdict isn’t in. In January, with Armis and Veza already announced, McDermott said no further large deal was on the roadmap.⁷ That is a promise worth tracking, because $11.2B in about four months is a large change for a company that calls itself organic. The company says the deals are running ahead of plan.¹⁶ I can’t check that, because it reports neither Armis’s revenue nor its margin. Pro forma results were left out of the 10-Q as not material.⁴ If Armis earns the 10% return I require at about $7.6B, it needs to produce about $760M a year after tax and stock pay. At a 25% owner margin, about what my base case reaches in ten years, that means about $3B of revenue, nearly nine times its size when the deal was announced E (estimate).

VFinancials

Here is the ten-year record. ServiceNow’s reported earnings are a poor guide for three separate reasons, so I’ll start with those.

• Tax releases. Net income of $627M in 2019 and $1,731M in 2023 each includes a large one-time tax benefit, $560M and $723M.¹⁷
• Gains on stakes. The first nine months of 2026 include $367M of unrealized gains on strategic investments, which is why GAAP net income beat GAAP operating income.²
• Non-GAAP profit. The company’s preferred measure leaves out stock pay, amortization of the deals, severance and deal costs. It was $3.96B in the last twelve months C (calculated).²³ All of those are real, and stock pay is the largest.

The table uses numbers as filed, in calendar years. Diluted shares are restated for the five-for-one split of December 2025.

Three things stand out. Revenue rose 11-fold in ten years, nearly all of it organic. Operating leverage is real: sales and marketing fell from 47% of revenue to 32%. And GAAP operating income has fallen this year, from $1.38B in the first nine months of 2025 to $1.05B, as amortization of the deals, severance and deal costs landed.² The GAAP operating margin in the third quarter was 9.5%. The non-GAAP one was 32.5%.² The whole of the gap is items the company asks you to leave out, and 66% of it is stock pay C (calculated).²

Cash, and what an owner gets

Free cash flow has grown every year and is the number management leads with: a 35% margin in 2025 and a 36% target for 2026, up a point in October.³² But the finance chief said on the call that the extra point comes from spending less on capital equipment.¹⁶ Capital spending has run at 6.5% to 9.3% of revenue for a decade, and was 3.8% in the last twelve months C (calculated).¹⁷ Meanwhile hosting cost is climbing inside cost of revenue, because more customers run on the hyperscalers’ clouds.⁴ The company has committed about $4.8B to cloud providers through 2030, $2.8B of it in 2030, and about $1.6B more to an equipment provider.⁴ So some of this rise in free cash flow is the move from owning machines to renting them. Gross margin pays for what capital spending saved. I don’t think the board is hiding anything. I do think “we cut capex” deserves a second look.

The free cash flow figure also adds back deal costs, and the company defines it that way. In the first nine months of 2026 it added back $371M, which is why the company reports $3.12B for that period and my simple calculation gives $2.75B.² Those were real dollars out of the door. The add-back is disclosed and I understand the logic, but I don’t add it back.

Stock pay

The direction is right. Stock pay was 22.8% of revenue in 2016 and 15.1% in the last twelve months C (calculated).¹⁷ The company has promised under 10% by 2029.⁵ That is a real commitment, and worth a lot: if it stays three points higher for good, my base value falls by about $12 a share E (estimate). But the share price matters here in a way people forget. At 30 September staff held 41.8M unvested stock units, and $4.2B of cost is still to be expensed.⁴ The units granted this year were valued at an average of $105.76.⁴ If the stock stays near $105.12, a company that wants to pay people the same dollars has to issue more shares. Staff who joined for stock at $200 will notice, and the best AI engineers have other offers.

The backlog

Reported cRPO growth held near 21% for two years. That steadiness is less clean than it looks, because deals added to it from late 2025. I estimate organic growth near 20% a year ago and about 18% now E (estimate). The company doesn’t publish that number, so I show my working in Section VII.

Margins and costs

Management says AI is doing real work inside the company. It claims more than $1B of annualized value from its own tools and says total headcount is slightly down while quota carriers and AI engineers rise.¹⁶ The 10-Q records $167M of severance in nine months under what it calls a workforce optimization strategy.⁴ The year-end 2025 headcount was 29,187.⁶ Non-GAAP operating margin guidance for 2026 is 31.5%, up from 31%, and that is after absorbing Armis.²³ Considering the deals diluted it, that is good cost discipline. It came partly from cuts, and cuts can’t be repeated forever.

One customer to watch

The 10-Q carries a line that most summaries miss. One customer, a U.S. federal channel partner and systems integrator, was 12% of total revenue in the first nine months of 2026 and 20% of receivables at 30 September, up from 11% at year end.⁴ This is a reseller. The money comes from many federal agencies. It still ties more than one dollar in nine to one buyer’s budget, and management says public sector demand is driving more on-premises deals.¹⁶

VIValuation

Before any numbers, I start from the far end. Say it is 2031 and this has gone badly. What happened? I can see three stories.

The three ways this goes badly wrong

The seats go and the assists don’t replace them. AI does what management says it does. Service desks shrink, and with them the seats those desks pay for. Consumption grows, but customers buy assists in bulk at falling prices, because the cost of a model call keeps dropping and rivals pass that on. Growth settles at 10% to 12% by 2028, not the 16% I use. Take five points off growth every year and three points off margin and my base value falls to $64, 39% below the price E (estimate).

Someone else owns the agent. Microsoft, or a lab, becomes the place where enterprise agents are controlled, and ServiceNow becomes a very good database they call. Customers keep it, because moving is hard, but they stop paying up for the newest tier. Gross margin keeps falling as ServiceNow pays for compute it can’t mark up. Five points less margin for good gives $98 E (estimate).

The capital goes the wrong way. Faced with slower growth, management buys it. Another Armis or two, paid for with debt and stock, while stock pay stays near 15% of revenue. Owners get a larger company and no more cash. If stock pay stays three points above my path and deals take 5% of revenue every year, my base is $91 E (estimate). Of the three, I find this one the most likely, because the company has just done it once and the pay plan doesn’t discourage it.

These can arrive together, since slower growth makes deals more tempting and makes each stock grant cost more shares. So I don’t add them, and I don’t put them in my cases. They are why my margin-of-safety price sits well under my base value.

What an owner can really take out

Now the cash. Reported figures give several answers, and they are far apart.

So after my tuck-in allowance an owner’s cash today is about $2.1B, a little less than half of free cash flow and 47% below non-GAAP earnings. The last twelve months are a low point, though. They carry deal costs, new interest and severance. The 2026 guide of a 36% free cash flow margin, on about $16.2B of revenue, implies about $5.8B of company-defined free cash flow and about $3.4B after stock pay C (calculated).² That is the starting point my model uses. On it the stock trades at about 32 times owner cash, against 28 times trailing non-GAAP earnings and 76 times trailing GAAP earnings C (calculated).²³

On capital needed to stand still: capital spending is small now, and tuck-in deals are a judgement call. The largest real cost of standing still is stock pay. I showed in Section IV that $7.7B of buybacks and share tax left the count higher. If you take stock pay out of my model the base goes from $118 to $155. That $37 a share is what you are asked to ignore when someone quotes free cash flow E (estimate).

How I built it

The model runs on calendar years from 2027 to 2041, with the fourth quarter of 2026 as a stub. Each year it takes subscription revenue, applies non-GAAP operating margin, takes off stock pay as if paid in cash, and charges cash tax on what is left, rising from 13% to 21% by 2031 as credits run down; cash tax was 11% of that profit in 2026, on my estimate. Then it adds depreciation, takes off capital spending and tuck-in deals, and adds 30% of each year’s new revenue for customers paying in advance. Deferred revenue has run at about 60% of revenue, less receivables.⁶ The terminal year takes that working capital gain down to the rate of long-run growth.

From the value I take off $7.60B of debt at face value, and add $6.96B of cash and securities and $2.05B of strategic investments at their carrying value.²⁴ I use 1.05B diluted shares, the company’s guide for the fourth quarter.² The cover of the 10-Q shows 1,038M outstanding, and 41.8M unvested stock units exist, but their future cost is already inside my stock pay line, so I don’t count them twice.⁴

The discount rate

I use 10.0%. The capital asset pricing model starts from the 4.94% ten-year Treasury yield and adds beta times Damodaran’s 4.20% implied equity risk premium for 1 October 2026.¹⁵³³ ServiceNow’s beta to the S&P 500 is 0.99 on 60 months and 1.30 on 36 months, which gives 9.1% and 10.4% C (calculated).¹³⁴ The betas disagree because the stock changed character this year. I take a number near the top, since the risk here isn’t well captured by past price moves.

Bear, base and bull

Base, $118. Growth of 17.5% in 2027 sits on the 18.5% cRPO guide and my organic estimate. Citi was reported to expect about 18%.²⁰ I then fade growth by a point or more a year. Subscription revenue reaches $27.8B in 2030, about 7% short of the company’s $30B+ target. Margin rises 2.5 points by 2030 and 8.5 by 2036. My owner margin of 25% in 2036 compares with 26% at Salesforce and 33% at Adobe today, so it isn’t a stretch for a mature company, though AI compute may make it harder.¹¹¹⁰ Stock pay misses the 2029 target by about three years: 11.7% in 2029, and under 10% only in 2032.

Bull, $178. This is close to the company’s own plan. Subscription revenue of $31.0B in 2030, stock pay under 10% by 2029, and operating margin of 39.5% in 2030. Even so, growth plus company-style free cash flow margin in 2030 comes to about 56, not the 60+ the company targets, because I tax it more fully.⁵ Hitting 60 in full would need about four more points of margin and would be worth about $198, which I show only for scale E (estimate).

Bear, $54. Growth halves by 2029 and margin never rises from 31.5%. It isn’t a collapse. Revenue still reaches $34B in 2036. It is what a good business looks like when the market stops growing under it. Salesforce today looks a bit like this, though with better owner margins.

I give the cases weights of 30%, 50% and 20%, which comes to $111 E (estimate). I don’t pretend those weights are measured.

What I can’t judge

The biggest input here is the net effect of AI on seats, and I can’t know it. No one can. The company says active seats are rising and half of net new business is non-seat.¹⁹ Investors reasonably suspect a lag. The evidence that would settle it, a large customer’s seat count at renewal, isn’t published. The company reports AI ACV, but that number counts only the increase, and it doesn’t report what was lost in the same accounts.⁸

So I show the range and not a false point. Each row below moves growth in every year. Each column moves margin in every year.

Values run from $65 to $209. That width is the honest answer. Two points of growth for fifteen years matters more than five points of margin. The market’s fear is about the right variable.

What the price implies

A week earlier, at $158.96, the same model needed growth about 2.9 points above my base every year, and subscription revenue of about $31B in 2030, close to the company’s plan. So in a week the market went from pricing most of the plan to pricing a little less than my base. One guide doesn’t make a business worth a third less. Either the price asked too much before, or it asks too little now, or both.

The discount rate and long-run growth

The margin of safety

I set it at $83, 30% below my base value. That is a larger cushion than the 25% I use on steadier businesses such as Visa, for one reason: the biggest input can’t be known, and being wrong about it costs far more than being wrong about the discount rate. At $83 the owner cash implied by 2026 guidance is about 3.9% of the market value, a point below a Treasury before any growth. At that price my bear case would lose about 35%, and my base would return about 14% a year. The stock closed at $83.00 on 10 April 2026, so that price has been available this year.¹

Quality and price are two questions

How good is the business? Very good. The renewal rate, the contract length and a decade of revenue growth say so. It has weaknesses I’ve listed, but few software companies of this size have a better record. How good is the price? Fair, with a small discount to my base value, on a number that depends heavily on something I can’t see. It isn’t the obvious bargain that a 27% daily fall suggests, because it was far from cheap before. The business is excellent, the price is fair, and the range around that price is wide. At a return of 11.2% on my base, the margin over my 10% hurdle is thin payment for that range.

The next best use of the dollar

A ten-year Treasury pays 4.94% with no risk to the coupon.¹⁵ ServiceNow’s owner cash yield is 2.2% today and about 3.1% on 2026 guidance, so the case depends on growth, as it does for all software priced for growth. On my base case the return is 11.2% against that 4.94%. The extra 6 points or so is pay for accepting a 3.0% return if the bear arrives.

The closest company on this site is Microsoft, which sits at 44% above my base value there. ServiceNow is 11% below its base, with the disadvantage that Microsoft owns the desktop and the cloud, and the advantage that ServiceNow needs almost no capital. Capital spending was 3.8% of ServiceNow’s revenue and about 39% of Microsoft’s in the last twelve months C (calculated).²⁷¹⁷ Snowflake and Palantir are 140% and 122% above their base values on my pages. Among software names here, ServiceNow is the one where the price is below my value, and I think that reflects the market’s being more afraid of seat-based models, not any superiority of mine.

Outside my pages, Adobe’s 8.6% and Salesforce’s 6.1% owner yields ask a fair question. If the AI fear about software is overdone, those two gain more, since they start with more cash per dollar of price. If it is right, ServiceNow’s better position has to carry a higher price through a slower future. I haven’t valued them in depth, so I don’t say which is better.

The price and why the market may be right

The price is below my base value. Here is why the market may be right and I may be wrong.

• Growth bought is not growth earned. About 2.5 to 3 points of third-quarter growth came from deals, built from Armis at 125 basis points in the second-quarter guide, Moveworks at about 1 point and a smaller piece for Veza; the company does not split it E (estimate). Take them out and subscription growth is about 18%, not 21%, and my base already assumes 17.5% next year with little room if AI adds nothing.
• Gross margin has turned. GAAP gross margin rose from 71% in 2016 to 79% in 2024, and was 73% in the last twelve months C (calculated). I assume it stabilizes. No filing says it will.
• My base margin is a guess about 2036. 46% of my value comes after 2041. A buyer today pays for cash a long way off, from a business whose product category may look different by then.
• Management’s language has run ahead of its numbers. “One of one” and “Rule of 57” are slogans. A guide cut by a point or two is a number. When the two disagree, I go with the number.

Where this could be in 5 and 10 years

For the base to hold at five years, subscription revenue has to reach about $31B in 2031, with growth fading to 12%. Operating margin has to be 35% and stock pay 10% of revenue. Owner cash, which is what is left after stock pay, capital spending and tuck-in deals, is then $6.6B. At ten years the base needs revenue near $49B in 2036 and a 40% margin. Cash taken out along the way adds $71 a share, so the 11.4% return leans on the business paying out. It differs from the 11.2% above because it sells at my 2036 value, not in 2041.

Bear and bull differ on the same numbers. In the bear, growth is 7% by 2031 and margin stays at 31.5%, so the 2036 value is $72, below today’s price. In the bull, growth is still 15.5% in 2031, margin reaches 44% and stock pay falls to 7.5%. The discount rate sits under all three: 46% of my base value today comes from after 2041, so a small change in the rate matters.

The most likely way the base is wrong is growth. If AI cuts the seats customers buy and growth runs 3 points a year below my base, the value is $109 in 2031 and $130 in 2036 E (estimate). That is the input I can’t yet measure.

A reader’s guide: these are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three things cut in ServiceNow’s favour and got lost on 29 October. The first is arithmetic. Guided cRPO growth falls 2.5 points, from 21% to 18.5%. Moveworks closed in December 2025, so its roughly 1 point stops counting as new E (estimate). Management says early renewals moved about half a point into the third quarter and out of the fourth. It also names 0.25 points of on-premises mix and 0.5 points of backlog excluded by termination clauses.⁷¹⁶ Together that is about 2.25 points E (estimate). So the drop from one quarter to the next is mostly mechanics. The level is a different matter: those same items sit inside my organic estimate of about 17.25% to 18%, down from about 20% a year ago E (estimate).

Second, RPO growth of 22% runs ahead of cRPO’s 20.5%: customers sign for longer.² Third, capital. ServiceNow spends 3.8% of revenue on equipment, against about 39% at Microsoft, and its release claims “no capital or compute constraints”.²²⁷

The bull case

AI agents need governing, and the company that already holds the map of systems and approvals is best placed to do it. McDermott says token use can raise revenue per user five times.¹⁶ The AI ACV target rose from $1B to $1.7B in ten months.² The 2030 plan is $30B+ of subscription revenue, 30% of ACV from AI and a Rule of 60+.⁵

Polen Capital’s 9 October letter, written before the report, saw ServiceNow “showing that AI can be a revenue opportunity rather than a headwind”.³⁵ At 30 June, the latest 13F date, Polen held 5.37M shares, down from 7.27M. Akre had bought 1.25M since the start of the year. Tiger Global had cut to 0.87M from 2.13M.³⁶³⁷³⁸ Needham and Bank of America were reported to see the quarter as the growing pains of a new AI line.²⁰

Against reality

If ServiceNow gets cRPO back near 20% in 2027, holds subscription gross margin at 80% and meets its stock pay target, the plan is possible, and my bull case of $178 with it. If it only does what my base assumes, my base cash flows return about 11% a year at $105.12.

Today is some way from the plan. $30B of subscription revenue in 2030 needs 17.4% a year from the 2026 guide C (calculated). Organic growth is near 18% and slowing. AI ACV of $1.7B is about a tenth of a contract base near $17B, and must reach about $9B by 2030, about 52% a year E (estimate).

Multiply it out. Growth above 17% through 2030: 45%. Assists add more than seats lose: 60%. Gross margin holds near 80%: 60%. Operating margin reaches about 39%: 50%. Stock pay under 10% by 2029: 70%. Multiplied, my guesses give about 6% E (estimate). Allowing for the overlap, perhaps 15% to 20%, my bull weight.

What a 10x would take. Ten times $105.12 is $1,051, about $1.10T of market value. At 30 times owner cash that needs about $37B a year, and at a 30% owner margin about $122B of revenue, 2.8 times Salesforce today and 7.5 times 2026, growing 22% a year for ten years C (calculated).¹⁰ My bull has 14%. Of the companies on this site I checked, three did a 10x from about this size: Tesla in 1.8 years from January 2020, NVIDIA in 5.8 years from September 2017 and Broadcom in 7.8 years from March 2017.³⁹ For ServiceNow I think it is close to impossible.

Where I come down

The cRPO arithmetic moved me: I think the fall in the headline number overstates the fall in demand. The AI numbers did not, because no one reports what AI costs the seat business in the same accounts. A year in which organic cRPO holds near 18% as AI ACV passes $2.6B would raise my values most.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Federal concentration. One federal channel partner was 12% of revenue and 20% of receivables.⁴ A budget fight or a shift in procurement would hit fast, and 12-month federal contracts make this backlog shorter than the rest.⁴
• Refinancing. The $2.1B of commercial paper rolls every couple of months at about 4%.⁴ The credit is strong and this is a small risk, but a company with no debt two years ago now has a calendar to manage.
• Security. A company that says it governs every agent in the enterprise becomes a target. A serious breach through the platform would damage the very claim the strategy rests on. The 10-K says as much in its risk factors.⁶
• Integration. Armis’s chief runs the whole security business now.⁸ Three acquired teams are being absorbed during a workforce reduction, with services losing money to get AI live.⁴²
• Talent and stock price. Staff have 41.8M unvested stock units granted at an average value of $126.67.⁴ At $105.12 those are worth 17% less than when they were granted. Pay may have to rise in cash or in shares.
• Geopolitics and timing. Delayed on-premises deals in the Middle East took about 75 basis points from first-quarter growth.⁸ Revenue recognised upfront on such deals adds noise to every comparison.
• Key person. The public case rests heavily on McDermott. His contract runs to 2030 at least, in one role or another.³⁰ If he moves to chairman, the case changes hands at a hard moment.