UiPath Inc. PATH

Key data

The business

UiPath sells robotic process automation. A customer records the steps a clerk takes across ordinary business software, logging into a portal, pulling a number, pasting it into an accounting system, and a UiPath robot then runs those steps unattended, thousands of times a day. Banks, insurers and manufacturers use it to drain the back-office work that never justified a custom application. The company charges by annual subscription, so the number that matters is annual recurring revenue: about $1.9 billion, up 12% year over year. Roughly 43% of revenue is United States, 11% Japan, no single customer above 10%. The moat is embedment: once a few hundred robots run a company's month-end close and its claims intake, replacing them means re-mapping every process, which is why customers rarely leave even as they add slowly. The live question is whether AI agents are the next thing UiPath sells or the thing that makes the robots unnecessary.

The numbers

Revenue grew every quarter, but the earnings line is noisy and needs reading, not glancing.

The $198.8M in Q3 FY26 came on just $13.1M of operating income: it is a one-time release of a tax reserve, not the business earning that. Treat it as a paper gain. Q2 FY27, reported September 3, is the freshest print: GAAP operating income of $32M, operating cash flow of $31M, and net new recurring revenue of just $37M on a $1.9B base, with net retention at 109%. That is the tension in one line, real profit now, thin new growth.

FY2026 was the first year of positive operating income, $56.8M, and that figure, not the tax-boosted $282.3M, is the honest earnings base. What makes the story work is cash, not accounting profit.

The company throws off over $350M in cash on almost no capital spend, and has spent it buying back stock faster than it issues shares to staff: diluted share count fell to 527.8M from 548.5M a year earlier, about 3.8% fewer. Cash, restricted cash and marketable securities total near $1.4 billion. So the model prints money and shrinks the count. The variant here is narrow: the market treats decelerating growth as a permanent glide toward zero, while a profitable, net-cash software company at a trough multiple only needs net new recurring revenue to stop shrinking to re-rate. The print that settles it is net new ARR over the next two or three quarters.

Management

Founder Daniel Dines runs the company again as CEO, and the record cuts two ways. On the company's side of the ledger, capital allocation is disciplined: hundreds of millions returned through buybacks while shares outstanding actually fell, rare for a software firm still handing out stock. On the insider side it is one-directional. Over twelve months insiders sold about $57M across 50 sales and bought nothing, and Dines himself sold $22.5M on August 19, two weeks before the earnings print. Plan status is not disclosed in the filings I read, so whether that was pre-scheduled or discretionary is unknown, and unknown is not the same as fine. The September 3 filing also reshuffled the executive bench, moving Ashim Gupta to focus solely on the operating role. Founders sell for many reasons, but a founder selling into his own strength while the company buys the shares back is a mixed signal, not a clean one.

How it fails or surprises you

Net retention keeps sliding. Existing customers now spend 9% more than a year ago, down from the higher levels this company posted in earlier years. If net retention drops under 100%, the installed base shrinks on its own and no amount of new logos offsets it. Watch the quarterly retention figure and net new ARR: another quarter near $37M points the wrong way.

AI agents eat the task, not extend it (right tail inverted). This is the fact the read explains least well: automation budgets are rising, yet UiPath's growth halved. If native agents inside Microsoft and OpenAI absorb the click-work UiPath charges for, growth goes to low single digits and 4 times sales proves generous. The tell is net new ARR failing to reaccelerate through FY2027.

The trough multiple re-rates (right tail). At 4 times sales, the cheapest since its 2021 IPO, and 19 times forward earnings, the market pays nothing for agentic automation working. If net new ARR turns back toward $60M-plus and retention firms, a profitable, net-cash compounder shrinking its share count can re-rate to 6 to 8 times sales while earnings compound underneath.

Closing thoughts

This is the middle world, where a specific number decides it rather than a mystery or a coin flip. The market has priced UiPath as a decelerating RPA vendor about to be commoditized by AI, and at 4 times sales with cash and marketable securities near $1.4 billion plus real free cash flow, the left tail is dead money, not ruin: survival is not the question. On my read the range skews to the upside because the multiple already sits at an all-time low against a business that is now profitable and buying itself in, and that is judgment, not arithmetic. The fatter risk is time, several more quarters of $37M net new ARR while the story stays cheap and boring, against a genuine re-rate if agentic automation converts even modestly.

The bet is still that big companies keep paying UiPath's software robots to do their dull computer work, and that the new wave of AI agents makes those companies buy more UiPath, not rip it out for something built into Microsoft or ChatGPT. What breaks that is net retention falling below 100% while net new ARR stays stuck near $37M a quarter: those two numbers, watched together, tell you before the revenue line does whether the installed base is still growing or quietly rolling over.