UiPath

IThe business

UiPath makes software robots. A robot opens the same screens a clerk would, copies a number from one, types it into another, checks a rule and moves on. That is robotic process automation, or RPA, and it is where the company started, in Bucharest in 2005, with a way to recognise what is on a screen.² A lender uses it to check a loan file against its rules. A hospital uses it to check a claim before it goes to a payer.² The work is dull, it is repeated thousands of times a day, and the old systems do not offer a clean connection, so a robot does it through the screen.

Today the company calls the product a “business orchestration and automation” platform. In March its release said “agentic automation”, and by May the wording had changed.⁹¹⁷⁴ The pieces are these. Maestro is the layer that decides which job goes to a robot, an AI agent or a person, and keeps track of the state of the process. Agent Builder makes AI agents that plan and act. The robots and API connections do the exact, rule-based steps. IXP reads documents and e-mails and turns them into data. Test Cloud tests other companies’ software with AI. Packaged “Solutions” for banks, insurers, hospitals and retailers sit on top.² The pitch, in the chief executive’s words, is that AI increases the need for “the orchestration, governance, and exactness that deterministic automation provides.”⁴ I will test that claim through Sections II and VII.

How it sells, and what it earns

UiPath sells term licences, support, and cloud subscriptions, mostly on one-year contracts, through its own sales staff and through resellers and consulting firms.² Revenue in fiscal 2026, the year to 31 January 2026, was $1,610.6M. I call the year to 31 January 2027 “fiscal 2027” throughout. Licences were 38% of it, subscription services 59% and professional services 3%.² The split matters because a licence is counted in full on the day it starts, while a subscription is spread over the term. A good licence quarter lifts revenue now and says little about next year.

The customer numbers are less happy than the revenue. UiPath counted about 10,747 customers at the end of fiscal 2026, against 10,753 a year earlier, 10,830 two years earlier and about 10,800 three years earlier.²⁷⁶⁸ Spending grew inside a customer base that did not. The group with more than $1M of ARR rose from 158 companies to 387 in four and a half years, and brings half the revenue.⁵³ So UiPath is selling more to the big accounts it already has, and not adding accounts. In fiscal 2026, 85% of the revenue growth came from existing customers.²

About 50% of revenue comes from the Americas (46% from the United States), 33% from Europe, the Middle East and Africa, and 17% from Asia-Pacific.² It employed 3,981 people at the end of fiscal 2026, 4,035 at the end of fiscal 2024.²⁶ Revenue per employee rose from $324,000 to $405,000 in two years C (calculated).

The measure the company runs on

UiPath’s own measure is ARR, annualized renewal run-rate: the yearly value of what customers have been invoiced for under subscriptions and support, with no growth or cuts assumed.² It was $1,937.7M at 31 July 2026, up 12%. Of that growth, 72% came from existing customers and 28% from new ones.³ Two things to know. ARR counts invoices, and a customer that cuts back at renewal shows up only when the renewal is invoiced. And the company does not publish a gross renewal rate, so the lost customers are mixed into the net figure.

Net new ARR, the amount added each quarter, was $37M in the second quarter, against $31M a year earlier and $49M in the first quarter. The first half added $85M against $57M a year before.³¹⁸¹⁹⁴ That is a good half. The guide for the rest of the year is where it gets harder to read. The company expects ARR of $1.992B to $1.997B at 31 October and $2.065B to $2.070B at 31 January 2027, which adds $127M to $132M in the second half.⁴ The second half of fiscal 2026 added $129M C (calculated).⁹¹⁹ So the company has guided the second half flat on last year after a first half up almost half. Its guide has been cautious in the past. It guided ARR of $1.929B to $1.934B for the second quarter and reported $1.938B.¹⁷⁴

The latest quarter

Second-quarter revenue was $410.3M, up 13.4% C (calculated), against a guide of $395M to $400M. GAAP operating income was $31.6M and non-GAAP operating income $89.0M, a margin of 22%. Gross margin was 80% against 82%, because professional services cost $39.4M to deliver and brought in $20.3M.⁴ Cash and securities were $1,405M and the company has no debt. It raised its full-year guide for the second time this year: revenue to $1.789B to $1.794B from $1.754B to $1.759B in March, and non-GAAP operating income to about $445M from about $415M.⁴⁹ The results came out after the close on 3 September. The stock had closed at $18.22 that day, up from $10.87 at the end of June. The next day it closed at $15.19, down 16.6% C (calculated), and it is $12.98 now.¹ I can’t tell you why it fell. One candidate is the arithmetic of the guide: it implies second-half revenue of $963M, 7.9% above the second half of fiscal 2026, after 15% in the first half C (calculated).⁴⁹ Revenue is lumpy because licences are counted up front, so the half-year swing in growth says less than it seems. It is still what the guide says.

Where UiPath is from

The company was founded in Bucharest in 2005, and was incorporated in Delaware in 2015.² Daniel Dines, the founder, is its chief executive and chairman. It listed on the New York Stock Exchange on 21 April 2021, selling 13.0M shares at $56.00 and raising $692M, and the stock closed that day at $69.00.⁵¹ It reached $85.12 on 24 May 2021. The headquarters is One Vanderbilt Avenue in New York. A Romanian subsidiary shares the cost of development with the U.S. company under a cost-sharing agreement, which matters for its tax in Section V.² The price today is 77% below the IPO price and 85% below the peak C (calculated).¹

IIThe moat

A durable business keeps its customers and its pricing without having to spend more each year to do it. UiPath’s customers are tied to it in two ways, and AI bears on them differently. The first is what they have built. A company with 400 robots has 400 scripts, each tied to particular screens in particular applications, written and tested in UiPath’s tools. Moving them to another vendor means rewriting them. The second is control. A bank that lets software log in as an employee needs a vault for the credentials, a log of every step, and a way to switch a robot off. UiPath sells those, and says it is the first enterprise automation platform to earn the AIUC-1 certificate for AI agent security, a claim I can’t check.²⁹

Evidence it is wide

The $1M group is the best evidence. It grew from 320 customers a year earlier to 387 in July, up 21% C (calculated), and 88% of revenue comes from the 2,666 customers above $100K.³ Net retention, the growth of spending by customers a year ago, has recovered to 109% from 107% at January.³² The company has $1,474.8M of contracted revenue still to be recognised, of which 62% falls within a year.² And the work is the kind customers don’t switch casually. The 10-K lists customers in finance, insurance, health care and the public sector, where an audit trail is the product.²

Evidence it is narrowing

The first crack is in the chart above. Net retention was 145% in January 2021 and 2022, 123% in 2023, 119% in 2024, 110% in 2025 and 107% in 2026. A customer base that adds 9% a year from existing accounts is healthy. It is a fifth of what it was.⁵⁸⁶⁷²³

The second is the length of the contracts. The 10-K says the licence agreements “primarily have annual terms” and that customers “have no obligation to renew.”² That is an honest sentence, and it means the installed base has to be won again every year. The company does not publish a gross renewal rate, so I can’t see how many customers leave and how much the others expand to cover it. The net figure of 109% could be 95% kept plus 14% expansion, or 100% kept plus 9%. The answer matters a great deal if AI cuts the number of robots a customer needs.

The third is the list of rivals. In its fiscal 2022 annual report UiPath named ten: Appian, Automation Anywhere, Blue Prism, Celonis, Kofax, Microsoft, NICE, NTT, Pegasystems and WorkFusion.⁵ The fiscal 2024 report stopped naming them and spoke of categories.⁶ The fiscal 2026 report adds “proprietary and open-source AI model providers and coding agents” to the list.² Its business section says AI gives customers “the ability to build software internally that would have otherwise been purchased.”² WorkFusion, one of the ten, was bought by UiPath in February 2026 for $189.5M.³ NICE is on this site too, and is the one rival from that list I have a page on.²⁰

The fourth is price. Gross margin was 80% in the second quarter, against 82% a year earlier and 83% in fiscal 2026. Cost of professional services rose 53% in fiscal 2026 on the use of outside contractors, and in the second quarter the services line cost $39.4M to bring in $20.3M.²⁴ A vendor that must pay people to get the product working is not selling something customers can install themselves.

My judgement

The switching cost on the robots already built is real and will last a few years. What it protects is a base that is today growing 12% a year. It does not protect how many new robots customers want, or what they will pay for each. If an AI agent can use a screen the way a person can, then a customer no longer needs a script for every screen, and the script is the thing UiPath sold for fifteen years. Its answer is that large companies won’t let an agent loose without a record of what it did, and that is why the orchestration and control layer is the part to value. That argument may be right. The company’s own quarterly record is the test, and I follow it in Section VII.

IIIThe competitive advantage

The moat says where UiPath is protected. This section asks what it does better than the others, and what the others can do to it.

Where it wins

It is neutral. UiPath works through the screen and through programming interfaces, with a wide range of systems and models. Its May release lists new links to Salesforce’s AgentExchange, Microsoft’s security tools and Databricks, and an expanded alliance with Deloitte.¹⁷ A customer that runs several AI agents from different vendors needs something above them that assigns the work, and a vendor that competes with none of them has a reasonable claim to that position. It builds on other companies’ models: it says its Screen Agent, running on Anthropic’s Claude Opus 4.5, ranked first on the OSWorld-Verified test of computer-use agents.⁹ That is the company’s statement, about one benchmark, and I treat it as a sign the product works rather than as proof it wins customers.

It has the installed base and the channel. The platform sells through its own staff and through consulting firms, and a substantial share of revenue comes through partners.² It also has the discipline of a company that has had to shrink. Sales and marketing fell from 52% of revenue in fiscal 2025 to 42% in fiscal 2026 and 40% in the second quarter. Non-GAAP operating margin was 18% in the first half of fiscal 2026 and 22% in the first half of fiscal 2027, and the guide for the full year is 25%.⁴ Employees were 3,981, fewer than the 4,035 of fiscal 2024 on 23% more revenue C (calculated).²⁶

Growth is not cheap. Non-GAAP sales and marketing cost $584M in fiscal 2025 and $587M in fiscal 2026, while net new ARR fell from $202M to $186M, so each new dollar of ARR cost $2.88 of selling and then $3.15. This year looks better, but the fair comparison is half against half, because most new ARR arrives in the second half: $4.78 in the first half of fiscal 2026 and $3.44 in the first half of fiscal 2027 C (calculated).⁹⁴²⁷ The share of revenue spent on sales and marketing fell, as above, and the cost of each new dollar of ARR rose last year and has fallen this year.

Where it loses

The platform vendors are the first threat. The 10-K says plainly that competitors “already have close, integrated relationships” with UiPath’s customers, and that any of them could offer automation inside a product the customer already pays for.² A customer of a large business-software vendor has a reason to ask why it needs a separate orchestration product. The model makers are the second. They are partners today, and their agents can already click through a screen. The third is the customer’s own staff, now able to write automation with a coding agent. UiPath answers with a product, “UiPath for Coding Agents”, that lets those agents build into its platform.¹⁷ That is a sensible reply. It is also an admission that the building, once UiPath’s job, can be done elsewhere.

Against the neighbours

UiPath is a ninth of ServiceNow’s revenue and grows two thirds as fast. Its owner margin is half as high, 7.5% against 15.7%, and 4.9% without interest on its cash C (calculated). The owner yield, 1.9%, is a little under ServiceNow’s 2.2%, so the market asks slightly more per dollar of today’s owner cash from a company that grows more slowly and keeps half as much of each dollar of revenue.¹⁴ ServiceNow is the best comparison on this site. It makes the same claim, that it is the control layer for AI agents across a company’s systems, and it faces the same fear, that agents reduce the seats and steps it charges for. Its page values it at $118 against a $105.12 price, 11% below, and its base case returns 11.2% a year.¹⁴ UiPath’s returns 9.2%.

This page shares its central idea with the pages for Adobe, Intuit, GoDaddy and Booking. In each, the market fears that AI agents will take over something the company charges for: the creative subscription at Adobe, the first click at Intuit and Booking, the front door at GoDaddy. If the fear is right, they fall together. If it is wrong, they rise together. UiPath and those four are one bet, not five, and ServiceNow, which I use as the comparison, carries the same fear. UiPath’s version is the most direct of the five, because an AI agent that can use a screen does what a UiPath robot does, and the other four lose a front door or a seat rather than a product. It is also the smallest, with the thinnest owner cash, and the only one of the six priced above my base value.

IVManagement

Who runs it

Daniel Dines founded the company, is its chief executive and its chairman, and holds every Class B share. Those carry 35 votes each, so his 64.7M Class B shares and 37.5M Class A shares gave him 84.6% of the votes at 28 April 2026. His sale plan converts 5M Class B shares to Class A, which lowers that share a little.¹⁵²¹ Of the Class A shares, 9.6M belong to his former spouse and carry his vote under a divorce settlement, so he owns about 92.6M shares outright, 17.8% of the stock C (calculated).¹⁵ He was chief executive until 2022, shared the job with Robert Enslin from 2022, handed it to Mr Enslin alone on 1 February 2024, and took it back on 1 June 2024 when Mr Enslin resigned.⁷¹⁵ With the vote in one pair of hands, my reading is that whether a chief executive stays is mostly his own decision.

The finance chief since 2019, Ashim Gupta, became chief operating officer too in 2024. On 3 September 2026 the company said he would “focus exclusively” on operations, sales and delivery, and promoted Hitesh Ramani, its chief accounting officer since April 2021 and a former Deloitte partner, to chief financial officer.²¹²² Raghu Malpani, who came from Microsoft in May 2024, became chief product and technology officer in March.²³ Yazdi Bagli, an executive of Kaiser Permanente, joined the board on 3 September, which now has eight members.²¹ Two of the directors are partners at the venture firm Accel, one of them the lead independent director, and Accel funds hold about 9.0M shares.¹⁵ The company has not claimed the “controlled company” exemption, so a majority of the board is independent, but with most of the votes held by one person the independence is advisory.²

How they are paid, and what that pays them to do

Mr Dines took a salary of $6,017 in fiscal 2026, no bonus, and has had no equity grant since before the IPO.¹⁵ His interest is his stock, and the question is how much of it he keeps. Everyone else has a bonus built on two numbers: 60% on incremental ARR and 40% on adjusted free cash flow. For fiscal 2026 the targets were $200M of new ARR and $375M of cash. The company reached 84% of the first, which paid 67%, and 99% of the second, for an 80% payout.¹⁵ The annual stock awards pay on revenue, with the target at $1.59B and the maximum at $1.65B, and nothing above 100% unless new ARR reaches $200M. Revenue was $1.611B and new ARR was $186M, so the payout was 100%.¹⁵

Here is what that rewards. Both bonus measures treat stock pay as free. Adjusted free cash flow starts from operating cash flow, and stock pay is not cash, so the $291M the company booked for it in fiscal 2026 costs the bonus nothing.¹⁵² New ARR counts what customers are invoiced and does not say what it cost to win it, and the proxy does not say whether ARR bought in an acquisition counts. Neither measure falls when the company issues more stock or pays more for growth.

The September awards are different in kind. On 3 September the committee granted 5.3M performance units to senior managers, 3.075M of them to Mr Gupta (1,125,000), Mr Malpani (1,125,000), Mr Ramani (525,000) and Mr Brubaker (300,000).²¹³ They pay only if the average share price over 90 trading days reaches two hurdles, the lower by 31 July 2029 for 33% of the units and the higher by 31 July 2031 for 67%, and they also vest by service, a twelfth each quarter for three years.²⁴ The hurdle prices are not in the 8-K, the 10-Q and its exhibits, or the four Forms 4, and no 8-K has followed, so I cannot say what these awards reward. A hurdle near today’s price would pay for sitting still; a hurdle at twice today’s price would be a real incentive. If every unit paid out, 5.3M units at $12.98 would be worth $69M, and below the lower hurdle they pay nothing C (calculated). The 10-Q for the third quarter, due in early December, may give the grant-date value of the units, which would narrow the range.

How much of their own money is in, and what they have done with it

Mr Dines owned 92.6M shares outright on 28 April 2026, worth $1.20B at today’s price C (calculated).¹⁵ Since then he has sold 3.9M, so about 88.7M remain C (calculated).¹⁶ That is serious money, and he is the largest shareholder by far. The Forms 4 show a pattern of selling as well. From 1 May 2025 to 16 October 2026 he sold 10.93M shares for $156.0M, an average of $14.28, under Rule 10b5-1 trading plans. The latest plan, adopted on 15 July 2026, allows up to 5M shares until 1 February 2027, and 3.9M have gone.¹⁶²¹ Of his shares, 20M were pledged as collateral for personal borrowing at 28 April, under a waiver of the company’s ban on pledging that the board gave him after he showed he could repay without the pledged stock.¹⁵ The size of the loan is not stated. Four other officers sold a combined $17.0M over the same period. In the 185 Forms 4 filed from February 2025 to 26 October 2026, no officer or director bought a share on the market.¹⁶

The largest outside holders, per the proxy, are BlackRock with 8.7% of the Class A stock, Tetragon and an affiliate with 6.0%, and Vanguard with 5.8%.¹⁵ Alphabet Inc.’s 13F report for the quarter ended 30 September 2026, filed on 6 November, lists 7,034,337 UiPath Class A shares, worth $90.3M at the quarter end and about 1.5% of the Class A stock C (calculated). The report names CapitalG II LP as an included manager for the line, and the count is the same in each of its four earlier quarterly reports.²⁵²⁶

What the company has done with capital is clearer than what it has said about it. It raised $692M at the IPO in April 2021, and in the first year its stock pay was $516M, of which $233M was a one-time charge for awards earned before the IPO.⁵ It spent nothing on buybacks until September 2023. Since then it has bought 91.3M shares for $1.09B, at an average of about $11.90, so today’s price is 9% above what it paid C (calculated).⁶⁷²³ That is a reasonable record on price. The cost is that the buyback has done little more than undo the stock pay. The share count fell from 569.0M in March 2024 to 521.2M, a drop of 47.8M, while the company bought 85.5M shares in the period C (calculated). The difference, about 38M shares, is what it issued to staff, 7% of the count in two and a half years.⁶⁷²³

It has also paid for companies and stakes: $189.5M for WorkFusion in February 2026 ($160.0M in cash and $29.5M of contingent payments), $38.1M for Peak in March 2025, and $35.8M for convertible bonds of H Company.³² Together they come to about $263M. I can’t judge them yet, because WorkFusion has been owned for nine months.

Two securities class actions bear on how management has talked to investors. The first, filed in September 2023, alleges misstatements in 2021 and 2022 about competitive position; the court dismissed it in full on 2 October 2025 and the plaintiff appealed on 9 February 2026. The second, filed in June 2024, covers December 2023 to May 2024; the court dismissed it in July 2025, the plaintiffs refiled against Mr Enslin and Mr Gupta, and the company moved to dismiss again on 27 October 2025. A derivative suit in Delaware repeats the allegations and adds that the company bought back shares at inflated prices. The company has recorded no liability for any of them.³

VFinancials

Seven fiscal years of audited statements are public, because UiPath listed in April 2021; the S-1 adds revenue of $148.5M for fiscal 2019, unaudited.²⁷ A ten-year record does not exist yet. What the seven years show is revenue that grew from $336M to $1.6B, a loss at the operating line in six of the seven years, and, since fiscal 2024, a business that makes real cash before it pays its staff in stock.

Growth has slowed in steps: 81%, 47%, 19%, 24%, 9% and 13%, with the first half of fiscal 2027 at 15%.²⁸³ Gross margin has stayed in a band of 81% to 85% after fiscal 2021, and was 80% in the latest quarter. The line that moved is operating margin before stock pay, which went from 2% in fiscal 2023 to 16%, 14% and 22% in the next three years and is 23% on the company’s own non-GAAP basis, guided to 25% this year C (calculated).²⁸⁹⁴

What an owner could really take out

Operating cash flow was $371M in fiscal 2026 and $373M over the last twelve months.²³ The company’s “adjusted free cash flow” is nearly the same, $372M for the year.⁹ Neither pays for the staff. In fiscal 2026 UiPath booked $291M of stock pay, which is 78% of operating cash flow and 18% of revenue. An owner can’t spend the cash and also keep the shares that were issued, so I take stock pay out as if it were paid in cash. That leaves $62M for fiscal 2026 and $128M for the last twelve months C (calculated).

Even that flatters it. About $45M of the last twelve months’ operating cash flow is interest on the company’s own cash, and that cash is worth its face value in my model, so counting its interest as well would count it twice. Without interest the figure is $84M, or 4.9% of revenue C (calculated).³ Over the whole seven years the owner cash is negative: −$1.6B C (calculated). The bad years were the IPO year, when $516M of stock pay was booked, and the years when revenue grew without profit. Fiscal 2026 is the first year the number is positive. I put the full-year figure for fiscal 2027 at about $135M before interest, using the first half as reported and a second half in which operating cash flow is 10% above last year’s E (estimate).⁴⁹

Quality of earnings

• The tax line. Net income of $282.3M in fiscal 2026 contains a $181.7M tax benefit, mostly the release of a $204.9M valuation allowance on U.S. deferred tax assets. Operating income that year was $56.8M. I do not quote a price over earnings on the net figure.²
• Cash taxes. UiPath paid $9.8M of income tax in cash in fiscal 2026 and nothing to the U.S. federal government. It has $662.4M of U.S. federal losses that never expire and $697.3M in Romania, which expire from 2027. The shield lasts some years, not forever, so I raise the tax rate in my model from 10% to 20% by fiscal 2032.²
• Stock pay is falling, and why it falls matters. It was $358M in fiscal 2025, $291M in fiscal 2026 and $98M in the first half of fiscal 2027, 12% of revenue. In the second quarter it was $45.0M against $78.0M a year before, and $33M of the $52M improvement in GAAP operating income came from that drop.²⁴ Older awards granted at higher prices are rolling off, and new awards cost less in dollars because the shares are worth less. That helps the income statement. It does not mean the company is issuing fewer shares: the count issued is about 38M in two and a half years.
• Interest. Interest income was $44.5M over the twelve months, which is 29% of GAAP operating income and 35% of owner cash C (calculated).³²
• Licence timing. A term licence is counted as revenue when it starts, so the swing between quarters is larger than the swing in the business. The first half of fiscal 2027 grew 15% on revenue and 12% on ARR.⁴
• Commissions. UiPath capitalises sales commissions and expenses them later. In the first half of fiscal 2027 it capitalised $68.0M of commissions and expensed $52.2M of earlier ones, so cash went out $15.8M faster than the income statement shows C (calculated).⁴
• Professional services. They lose money. In the second quarter they brought in $20.3M and cost $39.4M, a loss of $19.1M C (calculated). Revenue from them rose 64% in the half.⁴
• Restructuring and deals. Restructuring cost $14.1M in cash in fiscal 2026 and $5.0M in expense in the first half. Acquisitions raised intangible assets from $20.0M to $101.5M in six months.²⁴
• Open tax matters. Romania assessed $14.3M of VAT, which UiPath paid and is appealing. A second audit could cost about $13.0M. India’s tax authority has made a preliminary inquiry into transactions carrying $48.7M of GST. The company has accrued nothing for any of them.²³

The balance sheet

At 31 July 2026 UiPath had $607.4M of cash and $797.6M of marketable securities, together $1,405.0M, and no debt. It owes $34.3M of deferred and contingent payments for deals. Deferred revenue is $615.3M and receivables $307.1M. Operating leases are $79.8M, which I leave inside operating costs.⁴³ Net cash is $1.37B C (calculated), or $2.62 a share, and 20% of the price. Cash has fallen from $1.69B at January: $268.5M of buybacks, $149.4M of acquisitions and $23.2M of tax paid on employees’ share vesting went out, against $162.6M of operating cash flow.⁴ If the company keeps buying stock at the pace of the last three years, the cash balance is a smaller cushion than it looks.

VIValuation

First, how this goes badly wrong

Before the numbers I asked what would most likely hurt an owner here, and worked back. Three things stand out, and each one costs between $1.60 and $2.00 a share on its own E (estimate).

The click layer becomes a feature. UiPath began by letting software use a screen. An AI agent can now do that without a script, and UiPath sells one itself. If customers buy fewer robots and pay less for each, and the large software vendors include orchestration for nothing, growth stops and the price per unit falls. If growth is half of my base in every year, the value is $9.85. If it stops, my bear case is $4.51. This is the failure the price is pricing.

Stock pay takes the cash. The company has spent $1.09B on buybacks since 2023 and booked $1.1B of stock pay, and the share count fell 8%. If stock pay stays at 11.3% of revenue and does not fall to 8%, the value is $10.21. If an investor ignores stock pay altogether, as “adjusted free cash flow” does, the same model gives $16.40. That is $4.58 above my base, and it is what you are asked to overlook.

Management buys growth. Since the start of fiscal 2025 the company has paid or promised about $263M for WorkFusion, Peak and convertible bonds of H Company, while the controlling holder sold $156M of stock. If acquisitions run at 5% of revenue and not 2.5%, the value is $10.13.³²¹⁶

These can arrive together, since slow 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 below my base value.

What an owner could really take out

I value cash after stock pay and after the capital needed just to stand still. For software the second is small: UiPath spent $19.0M on equipment in fiscal 2026. The first is large. The company’s non-GAAP operating income is about $420M over the last twelve months and its adjusted free cash flow is $371M. After stock pay, the cash an owner could take out was $128M, and $84M without the interest on the company’s cash C (calculated).⁴⁹³

So I start from $84M of owner cash before interest, about 20% of the company’s operating income, and from $135M for fiscal 2027, which is 7.5% of revenue E (estimate). My base reaches $204M in fiscal 2028, $282M in 2030 and $454M in 2037. The method: take operating income before stock pay, take away stock pay, a small charge for payroll tax on shares, cash tax, an allowance of 2.5% of revenue for small acquisitions, and capital spending and working capital of 1.0% of revenue. The acquisition charge is my estimate. The company spent $245M on deals and stakes from fiscal 2022 to the first half of 2027, which is 3.4% of the revenue of those years; before WorkFusion it was $96M, or 1.8% of the revenue of fiscal 2022 to 2026 C (calculated).⁵⁸²³

The step from this year to the next needs a note. Owner cash goes from $135M in fiscal 2027 to $204M in fiscal 2028, up 51% on revenue up 10%. In round numbers, $68M of the rise is more operating income before stock pay, net of stock pay, and $51M is a smaller gap between that profit and cash, less $49M for small acquisitions, which this year’s figure leaves out. The gap is 6.0% of revenue this year and 2.8% in my base, and it is the part I trust least. Counting capital spending, it averaged 1.6% of revenue across fiscal 2025 and 2026, and it was 5.9% over the last twelve months C (calculated).⁷²³⁹ Working capital and commission timing move it most: together they gave $50M in fiscal 2025, took $40M in fiscal 2026 and took $46M in the first half of this year C (calculated).²⁷³ I do not know which year is normal. Each extra point of revenue lost to the gap, in every year, takes about 68 cents off the base value, and if this year’s gap repeated in every year the base would be $9.72 E (estimate).

What the price implies

Put the other way, what does $12.98 need? On my base shape it needs revenue growth 1.6 points higher in every year, which is 11.6% in fiscal 2028 and 9.6% in fiscal 2030, or an operating margin 2.1 points higher before stock pay in every year. The company is guiding ARR growth of 11.6% for this year and revenue growth of 11% for this year and 8% for the second half. So the price asks for the top of what UiPath is delivering to last, not the bottom.⁴ At $12.98 the enterprise value is 3.0 times this year’s revenue, 2.8 times ARR and 40 times my owner cash for this year.

What I can’t judge

• What AI does to the product. UiPath does not report revenue from agents or from Maestro, and its pricing, per robot, per seat or per use, is not disclosed. Net new ARR is the one test, and it is a mixed one.²
• How many customers leave. There is no gross renewal rate. Net retention of 109% may be 95% kept with 14% expansion or 100% kept with 9%. I treat it as the first.
• Whether the second-half slowdown is timing or trend. The guide implies 8% revenue growth against 15%. Licences are counted up front, and the company has beaten its guide before. I model fiscal 2028 at 10%.⁴
• Whether stock pay stays low. It fell from $358M to $98M in six months of fiscal 2027 partly because older awards at higher prices rolled off. If the price rises, so will the dollar cost of new awards. The hurdle prices on the September awards are not in any filing I could open.³²⁴
• What Mr Dines will do. He controls the vote, had 20M shares pledged in April and has sold $156M. I can’t see his plans past 1 February 2027.¹⁵¹⁶

Because I can’t know these, I show a range and the price’s position in it, and I don’t treat $11.82 as more exact than it is. A fair reading of my work is that the base is somewhere from $9 to $15, and the table below shows how each input moves it.

The discount rate

I use 10.0%, as on the other pages of this site. As a check, the capital asset pricing model starts from the 4.88% ten-year Treasury yield and adds beta times Damodaran’s implied equity risk premium.²⁹³⁰ UiPath’s monthly beta against the S&P 500 is 1.02 over 60 months and 1.49 over 36.³¹¹ With the 4.41% premium for 1 November that gives 9.4% and 11.4% C (calculated); with the 4.20% for 1 October used on the other pages, 9.2% and 11.1%. The two betas disagree because the stock has swung with its own news, from $10.87 to $18.22 and back in four months. I use 10.0% so the AI-fear pages can be compared on one rate. The case for a higher rate is real, since this is the smallest and least profitable of the six. At 11% my base is $10.63, and at 12% $9.72.

Three cases

Bear

AI agents that use a screen make the robot layer a feature, and large software vendors give away orchestration inside what customers already pay for. UiPath still has its base, but revenue grows 5% in fiscal 2028, 2% in 2029 and then stops, with the price per unit falling. Operating margin before stock pay slips from 25% to 20%, stock pay stays at 11% of revenue, and the company keeps buying small companies. Value $4.51, of which $2.62 is cash.

Base

ARR growth eases from 12% to about 10% and then fades. Revenue grows 10% in fiscal 2028, 8% in 2030 and 3.5% by 2037. Operating margin before stock pay rises from 25% to 30% by fiscal 2032, stock pay falls from 11.3% of revenue to 8%, and taxes rise from 10% to 20% as the losses run out. Owner cash reaches $454M by fiscal 2037, 14% of revenue. Value $11.82.

Bull

Orchestration becomes the product enterprises buy and agents run through it. Revenue grows 14% for two years and 10% or more through fiscal 2033. Margin reaches 35% and stock pay 7%. Revenue is $4.8B in fiscal 2037. Value $20.09.

Weighted 30%, 50% and 20% across bear, base and bull, the value is $11.28 E (estimate). Those weights are judgement. Growth matters more than margin: one point of extra growth in every year is worth about the same as 1.3 points of margin. The market’s fear is about the right variable.

Other uses of the same dollar

A ten-year Treasury pays 4.88% with no risk to the coupon.²⁹ At $12.98 my base cash flows return 9.2% a year, which is 4.3 points above that and 0.8 points below the 10% I ask. The bear case returns 2.7%, below the Treasury. The bull returns 13.9%. The extra four points or so is the pay for accepting the bear.

The better comparison is on this site. ServiceNow returns 11.2% on its base cash flows at its price, which is 1.2 points above its bar. Adobe returns 15.0%, Intuit 11.7%, GoDaddy 13.4% and Booking 13.5%, each against 10%.^1410111213 On my numbers, UiPath gives the lowest return of the six, in the same fear. That is not an argument for any of the others. It means the price of UiPath already assumes more than my work does, while theirs assume less. The company’s own alternative is its buyback: at $12.98, against my base of $11.82, each dollar spent buys about 0.91 dollars of value E (estimate). I did not find a better use in the filings.

What the market paid before

In January 2022 the market paid 19.4 times ARR for the business, in January 2023 5.7, in January 2024 7.6, in January 2025 3.7, and in January 2026 2.6. It pays 2.8 now C (calculated).^1586723 The multiple has stopped falling, and the stock is back where it was in January 2026, $12.59, with the business a year bigger. Since January 2023 revenue growth has been 9% to 24% and net retention has fallen from 123% to 109%. The multiple that bought 80% growth at the IPO is not the multiple to expect for 12%.

The margin of safety

I set it at $8.27, 30% below my base value. That is the cushion I used for ServiceNow, for the same reason: the largest input can’t be known, and being wrong about it costs far more than being wrong about the discount rate.¹⁴ At $8.27 the enterprise value would be $2.95B, 1.6 times this year’s revenue, my base would return 14% a year and my bear case would lose 45%. The stock has never closed that low. Its lowest close was $9.38 on 10 April 2026, and the price is 38% above it.¹ I am not saying the price will fall to $8.27. I am saying that at that price cash would be nearly a third of what a buyer pays, and my base case would return 14% a year.

Two different questions

The first is how good the business is. By the numbers it is a decent one that has become a better one: gross margin of 80% to 85%, operating margin before stock pay of 22% and rising, $1.4B of net cash, and ARR growth that is steady at 11% to 12%. It is also a business whose customer count has not grown, whose retention has fallen by a quarter and whose biggest threat is the AI it now sells.²³ The second question is how good the price is. At $12.98 it is 10% above my base value, which is not a bargain by my arithmetic and is not far from the base either. Those are different answers. A fair business can be a poor purchase at a full price, and a poor business a good one at a low price. This looks like the first case, and the market may be seeing something I am not: that a company with the orchestration layer in banks and hospitals has more time than my base gives it.

VIIThe case for the price

What most people miss

Two facts cut in UiPath’s favour and get little attention. First, the cost base has shrunk while revenue grew. Sales and marketing fell from 52% of revenue in fiscal 2025 to 42% in fiscal 2026, headcount is 3,981 against 4,035 two years ago, and revenue is 23% higher C (calculated).²⁶ Second, customers above $1M of ARR grew 21% in a year to 387 and bring half of the revenue, even though the total count is flat.³

The bull case

The bull case here is the company’s own: I found no named investor or analyst view that I could check against a dated document. I opened eight growth funds’ 13F reports for the four quarters to 30 June 2026: ARK, Altimeter, Atreides, Pershing Square, Tiger Global, Dragoneer, D1 and Coatue. Only Coatue lists the stock, with 2.27M shares at 30 September 2025 and 31 March 2026, and 1.96M, worth $21.3M, at 30 June. A holding is not a view, and none had filed for 30 September.³² In the second-quarter release Mr Dines says AI is “increasing the need for” the control layer that UiPath sells, and the first-quarter release says its agent products are “moving from pilot to production.”⁴¹⁷ The argument: companies cannot let AI agents loose on loans, claims and payments without a record and a way to stop them; UiPath already sits in banks and hospitals with that record; so as agents spread, orchestration becomes the product, and the screen-clicking robot becomes one tool it directs. The evidence offered is net new ARR of $85M in the half against $57M a year before, retention back to 109%, a guide raised twice, and links to Salesforce, Microsoft, Databricks and Deloitte.¹⁷⁴ UiPath shares this bet with Adobe, Intuit, GoDaddy and Booking: one idea about whether AI agents take work from software or add to it, so the five pages are one position, not five.

Against reality

If ARR growth holds at 12% for several years, operating margin before stock pay reaches 30% by fiscal 2032, and stock pay falls to 8% of revenue, the price is about right. That is my base case, $11.82, with 1.6 points more growth in every year. To reach my bull of $20.09, growth has to rise to 14%, margin to 35% and stock pay to 7%. Today ARR growth is 12%, margin is 25%, stock pay is 12% of revenue and customers are flat. My base already asks margin to gain five points and stock pay to fall four, and the price needs a little more than that. The bull asks growth to gain two points and margin ten. None of it is out of reach, and none of it has happened.

Retention matters too: at 105%, new customers would have to supply most of 12% growth, and the count has been flat for three years.²³

What a 10x would take. Ten times $12.98 is $130, or $68B of market value on today’s shares. At 25 times owner cash, which is a high multiple for a company that has slowed, that needs about $2.7B of owner cash a year. An owner margin of 25% is below what Salesforce and Adobe make now, 26% and 33%.¹⁴ At that margin the revenue is about $11B, 6.1 times this year’s, which is 20% growth a year for ten years C (calculated).¹⁴ UiPath’s ARR is growing 12% and its owner margin is 7.5%. ServiceNow had $1.4B of revenue in 2016, close to UiPath’s $1.7B, after growing 38% that year, and compounded 28% a year to 2025.¹⁴³³ A tenfold is possible in arithmetic and I would call it very unlikely here. If I multiply it out as three needs, revenue growth of about 20% a year for ten years (8%), an owner margin of 25% (35%) and a market still paying 25 times owner cash at the end (40%), the odds are about 1%. They are my guesses and they are not independent, and I do not use them in the values E (estimate).

Where I come down

Three things moved me: the cost cuts, the growth of the $1M group, and the cash. A strong first half moved my estimate for this year and not my base: half a year of licence revenue is not a trend, and the second-half guide matches last year. What did not move me is the flat customer count and the fall in retention from 145% to 109%. What would raise my values is ARR growth above 12% for four quarters running, retention above 110% and stock pay at or below 10% of revenue by fiscal 2028; if growth ran two points above my base and stock pay a point below it in every year, the value would be about $14 E (estimate). Until then the price sits between my base and my bull, and it is the one price of the AI-fear group that asks for more than my work gives.

VIIIRisks, and what would change my view

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

• One holder controls the vote. Mr Dines held 84.6% of the votes at 28 April, and 20M of his shares were pledged against personal borrowing. If the stock fell far enough for a lender to ask for more, he could be forced to sell or to add collateral. I can’t see the loan terms.¹⁵
• Leadership has changed a lot. The chief executive changed twice in 2024, a new finance chief took over this September, and the old one now runs operations only. Each change may be right. Together they mean the team that guides the second half is not the team that guided the first.⁷²¹
• Big customers and partners. The top 10% of customers bring a substantial part of revenue, and the $1M group brings 50%. A substantial amount of revenue goes through partners who can also sell rivals.²
• Taxes and audits. Romania, whose subsidiary shares development costs with the U.S. company, has assessed VAT and audited income tax. India has a preliminary inquiry on $48.7M of GST and a transfer pricing request open. The U.S. losses that shield cash tax will run out.²³
• Dependence on other companies’ models. The platform relies on third-party model providers and cloud hosts, and the 10-K lists their pricing and access as a risk. If model makers raise prices or compete directly, the margin falls.²
• Deals. WorkFusion is nine months old, and its earn-out is a $30.4M liability. Integration problems tend to show after a year.³
• Currency and law. About 54% of revenue is outside the United States, and rules on AI differ across the European Union, South Korea and several U.S. states.²
• The courts. The first securities action is on appeal and the second awaits a ruling on a motion to dismiss. No loss is accrued.³