Freshworks

IThe business

Freshworks makes software that companies use to answer requests. A customer asks for help, or an employee asks for a laptop, and the software logs the request, routes it, tracks it and, more and more, answers it. It has two families of products. Employee experience, which the company calls EX, is Freshservice, an IT and enterprise service management system, with Device42 for discovering and mapping IT assets and FireHydrant for incident response. Customer experience, CX, is Freshdesk and its relatives for support, chat, calls, sales and marketing.³ Nearly 75,000 companies pay for something, from about 170 countries. At the end of 2025 over 60% of annual recurring revenue (ARR) came from customers with more than 250 employees, and no customer was over 1% of ARR.³ It is headquartered in San Mateo, California, and most of its roughly 4,500 employees are in India.³

Revenue was $838.8M in 2025, up 16%, and is on course for about $965M in 2026, up 15%, using nine months reported and the middle of the fourth-quarter guide C (calculated).³² Almost all of it, 99%, is subscription, software licence and maintenance; services are about 1%.³ The company said after the third quarter that ARR had passed $1B, ahead of its own year-end estimate.² By region, in 2025, 46% of revenue was North America, 39% Europe, the Middle East and Africa, and 15% the rest of the world, and about 28% was exposed to the euro or the pound.³

The two halves

EX is the engine. Its ARR was $510M at the end of 2025, up 26%, and the company says it passed $540M in the first quarter, up 27%, and grew 24% in the second.⁹⁷¹¹ Freshservice is the main product in it. Management expects it to pass $600M of ARR by the end of 2026 and about $1B in 2028.⁷ EX was 37% of ARR in 2021 and the company puts it at about 60% in 2026 and 70% in 2028 E (estimate).⁷ CX is the older business, built on Freshdesk. Management expects it to grow low to mid single digits, and it runs it for profit: CX is sold almost entirely inbound, through the website and free trials, and the company says its operating margin is above 30%.⁷ I have no split of reported revenue by EX and CX, so I work from ARR and from the company’s own targets, and I treat the CX figure as the company’s claim.

The split matters because the two halves face different questions. In EX, the questions are whether Freshworks can win larger customers away from ServiceNow and Atlassian and whether AI agents add or remove seats. In CX, the question is price: Freshdesk competes with Zendesk, Salesforce, Intercom and HubSpot, and the AI tools being sold into customer service are the most visible in software.³

How customers pay

Most revenue is a subscription per seat, billed in advance, monthly, annually or over several years.³ Monthly plans were 13% of ARR at the end of 2025, down from 17% in 2023, and the company says its retention on monthly plans is lower than average.³ Freshservice was first sold per agent and per user. Management says that by 2028 it will price in several ways: per agent, per asset, per alert, per AI session and per employee, and it already sells an AI Copilot add-on for $29 a month per agent and plans $0.49 per AI Agent session.⁷ That shift is the company’s answer to the central worry in this report: if an AI agent answers the ticket, nobody needs a seat, so the price has to move to the work done and not to the person doing it. I come back to it in Valuation.

The software is built on Ruby on Rails, Java and MySQL, hosted on Amazon Web Services, and it runs as many independent “pods” so that customers’ data can sit in the region they choose.³ The company has a commitment of $255.7M to cloud providers and other vendors through 2028, which is small next to revenue.³ There is no debt.²

The shape is clear in the table. Revenue growth has slipped from 16% to 14%. Net dollar retention has fallen from 108% to 104%, though the fall is mostly the currency effect, since at constant currency it has stayed between 104% and 105%. The number of larger customers is growing quickly, up 25% a year for those above $100,000, while the number above $5,000 grows only 5%. So the business is getting bigger customers and fewer new small ones. That is what the company intends. It also means growth from here depends on the larger deals landing.²

Where Freshworks is from

Girish Mathrubootham founded Freshworks and ran it as chief executive from 2010 until May 2024. The first product was Freshdesk, a support desk that small companies could try free for 14 days, and the company still sells the CX side mainly by search and word of mouth.⁹³ It listed on the Nasdaq on 22 September 2021, opened at $43.50 and closed its first day at $47.55.¹ Dennis Woodside took over as chief executive in May 2024, and Mr Mathrubootham stayed as executive chairman until he retired on 1 December 2025.⁹¹⁴ In June 2024 Freshworks paid $238.1M for Device42, in January 2026 $88.7M for FireHydrant, and in October 2026 about $102.5M for two more companies, Torii and Robusta.³⁴ The company calls itself an EX-first business now. It reduced headcount by more than 20% from the third quarter of 2024, and it says it stopped investing in non-core products, focused its sales on eight countries and cut its engineering to three priorities, EX, AI and CX.⁷ That is a reasonable account of a business that grew fast, spent heavily, and changed course.

The latest quarter

The third quarter of 2026 beat the guide. Revenue was $245.4M, up 14%. GAAP operating income was $28.2M, a margin of 11.5% against a loss of 3.5% a year before. Non-GAAP operating income was $73.2M, a margin of 29.8%. Net income was $18.3M, or $0.07 a share, and adjusted free cash flow was a record $73.9M.² The company raised its full-year guide to revenue of $965M to $966M and non-GAAP operating income of $238M to $240M, from $963.5M to $966.5M and $222M to $228M in August, and from $958M to $964M and $207M to $215M in May.²¹¹⁷ It guided the fourth quarter to about 14% growth.² The company reported after the close on 3 November, when the stock stood at $14.15, and held its call at 5:00 p.m. Eastern. It closed at $13.26 on 4 November on 11.4M shares, about twice its average volume, then at $12.70, $12.11 and, four trading days after the report, $12.01, 15% lower, so the market did not like something it saw.²¹¹⁵¹⁶ The company gave no new reason that I can read. The numbers I can point to are the slowing of growth from 16% to 14%, a net dollar retention rate of 104%, and a fourth-quarter guide that does not speed up. I cannot tell which of these, or what else, moved the price.

IIThe moat

A service desk is not a product people choose twice. A company picks one, loads its tickets, its assets, its approval rules and its staff, and then lives in it. That is the protection Freshworks has, and it is stronger at the top of its range than at the bottom.

What holds customers in

In EX the protection is real. A mid-sized company’s IT team runs on the desk: every incident, change and request sits there, the asset list is built from it, and hundreds of small automations and approvals depend on it. Changing the system means moving history, retraining agents and rewriting workflows. That is why a company keeps the tool through a bad year. The numbers agree. In the first quarter, as the company reported them, EX customers retained and expanded at 111% a year and customers using more than one EX product at 119%.⁷ The 10-K says the top ten customers are under 5% of ARR and no customer is above 1%, so no single loss can hurt.³ Device42, which maps what is in a customer’s IT estate, makes the system more central, and the company says about a third of its large new lands include advanced asset management.³⁷

In CX the protection is thin. A support desk with a free trial and a credit card is easy to start and not much harder to leave. Overall net dollar retention was 106% in the first quarter and EX was 111%. If EX was about 58% of ARR, the rest, which is CX, kept about 99% of its dollars E (estimate).¹³⁷ I do not have the CX figure from the company, so that number is arithmetic on two reported ones and one assumed share. Month-to-month plans, which are the easiest to cancel, were 13% of ARR at the end of 2025.³

What does not protect it

There is no law, no network effect and no scale advantage here. Freshworks has about $1B of ARR. ServiceNow, which I cover on this site, has a contract base near $30B, grew subscription revenue 21% in its latest quarter and keeps 98% of contract value each year.¹⁷ The market for IT service software is large, about $45B by the company’s estimate, and fragmented; the company says no vendor has more than a 20% share.⁷ That is room to grow, and it is also the sign of a market without a lock.

The AI question

The new threat comes through a different front door. If an employee asks an assistant for a new laptop and the assistant files the request, resolves it or routes it to a tool nobody sees, the help desk becomes a database behind the assistant, and databases are worth less than the screens people use. Freshworks is aware of it. Its answer is to be the system behind the assistant: Freshservice has an MCP gateway so that outside AI tools can read it and act on it, and the company says its own AI agents already deflect about half of the EX tickets in its base, with Copilot users handling 52% more tickets per agent.⁷

Take the company’s own numbers. A customer whose agents handle 52% more tickets, and whose bots remove half of the rest, needs fewer agents. If the product is priced per agent, fewer agents means less revenue unless the price per agent goes up, or the pricing moves to something else. The company knows this. It sells Copilot at $29 a month per agent and plans a charge per AI session. It reports that AI ARR passed $25M at the end of 2025, nearly double the year before, and aims for $100M in 2028 from a base of 8,000 AI customers.⁹⁷ Net dollar retention for customers with AI is about 118% by the company’s account.⁷ Those are good signs and small ones: $25M is 3% of revenue. The question is whether AI becomes a layer that adds dollars to every seat, or a substitute that removes seats faster than the new charges replace them. I do not know, and the data for the answer does not exist yet. I take it up again in the next three sections.

The AI row matters most. Every other protection on this list is old and measurable, and I can say what it is worth. This one is new, and I cannot yet. It is also the idea the market is trading on. I come back to it in Valuation, because the value depends on it more than on anything else.

IIIThe competitive advantage

The protection says what keeps customers. This section is about why they choose Freshworks to begin with.

It is fast. The company says 80% of deployments go live in under 12 weeks, against what it calls the long, expensive projects of the larger vendors, and cites a customer that moved 30,000 users off a 14-year-old system in three months.⁷ It is simple: a mid-market IT team can set it up with no code, which keeps the cost of owning it low. And it prices in the open, which the company counts as an advantage.⁷ Those claims are the company’s. The evidence behind them is the buying pattern. Customers above $100,000 of ARR are 1,821, up 25%. Over 40% of EX ARR comes from customers above $100,000. The company won about 250 new customers above $30,000 of ARR in the twelve months to March, and in the first quarter it signed the two largest deals in its history, including its first with $1M or more of ARR.²⁷¹³

The cost base

Freshworks builds in India. A majority of its employees are there, including engineering, support and a large part of inbound sales.³ That lowers cost per head and it is one reason non-GAAP gross margin is 86.6% and non-GAAP operating margin is now 29.8% in the best quarter.² The company’s non-GAAP sales and marketing was 34.7% of revenue in the third quarter against 40.3% a year earlier, and research and development was 13.3% against 15.1%.² Management says it expects sales and marketing to reach 36% to 37% of revenue by 2028, a level it reached in the third quarter alone.⁷ The advantage is real. It is also open to any rival that hires there.

Evidence the advantage is widening

• EX ARR was $510M at the end of 2025, up 26%, and grew 24% in the second quarter, much faster than the whole company’s 14% revenue growth.⁹¹¹²
• Customers above $100,000 of ARR rose 25% in a year to 1,821, and the company says 40% of EX ARR comes from them.²⁷
• Non-GAAP operating margin was 21.2% in 2025 against 13.8% in 2024 and 7.5% in 2023 C (calculated), and 29.8% in the third quarter of 2026.³²
• More than 60% of gross bookings now come from existing customers, which are cheaper to win.⁷
• Freddy AI Copilot is attached to over 71% of new enterprise deals, and paying Copilot customers doubled in a year.¹¹⁷

Evidence it is narrowing

• Revenue growth fell from 16% to 14% in two quarters, and the guide for the fourth is about 14%.¹¹²
• Net dollar retention is 104%, down from 108%, and the customer count above $5,000 grows only 5% a year.²
• Leadership has turned over: a new chief revenue officer in March 2026, a new chief product and technology officer in August, and the former head of customer growth left in March, with an advisory role to April.¹⁸¹⁹
• Bought products are part of the story: Device42, FireHydrant and the two October deals cost $429M, seven times the $63M of free cash flow left after stock pay since the start of 2024, or 18 times the $24M left without the 2025 stock pay credit C (calculated).³⁴¹²
• Rivals that offer the same idea to larger customers are adding AI to a bigger base.³

Where it wins and where it loses

Freshworks wins where a company of 250 to 20,000 people wants a real service desk without a long project: it is the standard answer for the mid-market. It wins when the buyer is the IT lead, not a procurement committee. It loses at the top, where ServiceNow is the default, and at the bottom, where a free tool or a general assistant does the job. In CX it competes on price against larger suites. The company has chosen to fight where it can win and to run the rest for profit, which I think is right. The risk is that the middle it has picked is the part of the market where an AI-built tool is most likely to be good enough.

IVManagement

Who runs it

Dennis Woodside has been chief executive since May 2024 and a director since 2022. Before that he ran Impossible Foods as president, was chief operating officer of Dropbox from 2014 to 2018, chief executive of Motorola Mobility from 2012 to 2014 and spent eight years at Google. From 2018 to 2022 he sat on the board of ServiceNow, a company Freshworks competes with.⁹ Tyler Sloat has been chief financial officer since April 2020, after ten years as chief financial officer of Zuora, and chief operating officer since August 2024.⁹ Roxanne Austin, a director since 2021, became chair on 1 December 2025 when the founder retired; she also sits on the boards of CrowdStrike, Verizon and AbbVie.¹⁴ The board has eight members.⁹ The sales and product leadership is new: Ian Tickle became chief revenue officer in March 2026 and Ryan Manning chief product and technology officer in August.¹⁸¹⁹ This is an experienced team that has been in place for two years and has changed two of its senior roles in the last eight months.

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

Mr Woodside’s pay in 2025 was $15.6M: $0.63M of salary, $0.69M of bonus and $14.3M of stock awards, or 92% in stock. Mr Sloat’s was $6.9M, 86% in stock. The pay ratio to the median employee was 354 to 1.⁹ The annual bonus is paid quarterly and weighted 70% on net new ARR and 30% on non-GAAP operating margin.⁹ The stock awards are 70% time-vested units that vest quarterly over four years in 2025, and 30% performance units; in 2025 those were earned on revenue (70%) and free cash flow (30%) and paid 105.4% of target. For the 2026 grants the performance part is 40% of the value, measured on revenue (70%) and non-GAAP operating income (30%), with a maximum of 173.6%.⁹⁴

The plan pays them to grow revenue and ARR, to widen the non-GAAP margin and to produce free cash flow. Every one of those measures is taken before stock pay. No measure charges for the dilution the awards cause, for the price paid on a buyback or for the cost of an acquisition. Revenue can be bought, and a deal that adds $50M of ARR for $90M of cash raises every target on the list. The headline measure management has chosen for the owner’s view, free cash flow per share, is also helped by buybacks, which shrink the share count whether or not the stock was cheap.⁷ It is a standard plan for a software company, and I do not call it wrong. It does not tell the executives to care about what this page cares about, cash after stock pay and after deals. In May 2026 shareholders gave the pay plan 378.2M votes for, 57.1M against and 45.7M abstaining, a support of 79% of the votes cast including abstentions C (calculated).²⁰

How much of their own money is in

The two executives who run the business own real stock. Mr Woodside owned 2,725,256 Class A shares on 2 October, worth $32.7M at $12.01 C (calculated), and Mr Sloat about 2.27M, worth $27.3M, counting Class B. Both also bought on the open market. Mr Woodside paid $11.31 for 176,100 shares on 10 November 2025, and $7.95 for 125,000 on 2 March 2026, when the stock was close to its low. Mr Sloat paid $11.62 for 171,615 shares on 11 November 2025. That is $4.98M of purchases in all, three trades by two people.¹⁰¹² Across the twelve months to November 2026 I count eleven open-market sales, $1.57M in all, by four people. The largest seller was the chief accounting officer, who sold 77,514 shares for $0.96M; a former executive and two directors sold the rest. Many more shares are withheld to pay tax when awards vest. That is a payment to the tax authority and not a sale in the market.¹⁰ I read the record as modest: the senior people have bought, they have not sold, and the sums are small next to a stock-based pay of $15.6M for one of them.

The company has two classes of stock. Class B has ten votes a share and Class A one. At the end of March the founder, Mr Mathrubootham, held 10,973,112 Class B shares, 18.5% of all votes, though he has left the company; funds linked to Accel held 23.6% of the votes and funds linked to CapitalG, Alphabet’s investment arm, 16.7%, based on filings made in 2024 and 2025.⁹ By 29 October 2026 Class B had fallen to 21.9M shares, from 58.4M at the end of 2024, as holders converted or sold, and I calculate it carries 48% of the vote C (calculated).⁵⁴ Under the charter all Class B converts to Class A no later than the last trading day of the fiscal year after the seventh anniversary of the listing, which I read as the end of 2028 or of 2029.³ CapitalG is part of Alphabet, and Alphabet’s Form 13F lists the stake. For the quarter ended 30 September 2026, filed on 6 November, it lists 14,336,643 Freshworks Class A shares in two lines, 12,457,273 and 1,879,370, worth $183.4M at the 30 September close of $12.79. The smaller line matches the CapitalG II LP holding in the proxy. The larger line was 14,327,273 at the end of March and June, so it fell by 1,870,000 shares in the third quarter, from a combined 16,206,643, the count the proxy gives for the CapitalG funds in a filing of November 2024. A 13F does not say whether shares were sold, given away or moved. The funds were already holders under an investors’ rights agreement dated December 2019, so this is an old stake, not a new purchase. I draw no conclusion. A 13F covers listed shares only, so it leaves out Class B.^2122239

What they did with capital over a full cycle

The company raised about $1.1B in its offering in 2021 and ended that year with $1.32B of cash and securities C (calculated).²⁴¹² It has since booked stock pay of $902M, net of the 2025 credit (2022 to the third quarter of 2026), bought back $635M of shares, paid $381M in tax at vesting and spent $289M on acquisitions, with $102.5M more in October C (calculated).^1224534 It had $697M of cash and securities on 30 September 2026 and about $595M after the October deals. Free cash flow was positive over the whole period but small compared with the stock pay: $610M against $902M C (calculated).

Read the table as an owner. Without the interest on its cash, the business left nothing over after stock pay until 2025, when it left $36M, and then $40M in nine months of 2026. The 2025 figure is helped by the $38.7M credit; without it 2025 was a loss of $3M, so 2026 is the first clean year. In the same stretch it spent about $1.3B on buybacks, tax at vesting and acquisitions. That money was the cash raised at the listing, not what the business earned. Cash and securities fell from $1.19B at the end of 2023 to $595M after the October deals. That does not make the buybacks a mistake. It answers whether they came from earnings, and they did not.

The count of shares tells the rest. Shares outstanding rose from 273M at the end of 2021 to 303M at the end of 2024, an 11% increase after the offering, as the stock awards vested. Then buybacks cut it by 13.5% to 262M. Over the whole period since the listing, the count is down 4.0% C (calculated). In the nine months to September, about 7.6M units vested, of which 2.9M were withheld for tax, so 4.7M new shares were issued against 26.8M bought back.⁴ Buying back stock is only worth it if you pay less than the stock is worth. The next table shows what they paid.

It is a mixed record. The company bought the same number of shares in both periods, 26.9M and 26.8M. In 2025 it paid $14.29, 19% above today’s price, and has lost $61M on those shares. In 2026 it paid $8.65, and has gained $90M. Together it has paid $11.48 and is up $28M C (calculated). That is luck as much as skill: the stock fell to $6.88 on 23 February, after the 2025 buying was done, and rose 125% to its October high. Mr Woodside said in February that the company would buy shares “that we believe the market substantially undervalues.”²⁷ The February price was low. Whether $14.29 was a good price in 2025 depends on what the shares were worth then, and I did not estimate that.

Acquisitions are harder to judge, because the company does not report how much ARR it bought. Device42 cost $238.1M, 5% of it in stock, and the 10-K describes it as IT discovery and asset management; FireHydrant cost $88.7M and the 10-Q calls its results not material; Torii and Robusta cost $102.5M.³⁴ I cannot see a return on any of them. The company says it plans one to three tuck-in acquisitions a year.⁷ At the pace of the last three years, about $143M a year, that is a significant claim on cash and I charge for it in Valuation. I regard the buyback record as more disciplined than the deal record, because for a buyback I can see the price.

VFinancials

Freshworks listed in September 2021, so the annual reports give me seven years, 2019 to 2025, and not ten. The 2019 and 2020 figures come from the 10-Ks for 2021 and 2022. The share counts before the listing are not comparable with the ones after it, because the preferred shares converted. Read the table as a company that grew fast, lost money, then stopped.

Three points from the table. Revenue grew 4.9 times in six years, from $172M to $839M, and the rate of growth has fallen from 45% in 2020 to 16% in 2025. Gross margin rose from 79% to 85%. And the loss has turned: operating income before stock pay went from −$26M in 2022 to $160M in 2025, which is 19% of revenue, but after stock pay the 2025 figure is $13M, 1.6% of revenue, and a $38.7M credit described below helps that figure C (calculated).¹²³

That gap is the whole story of this company’s profit. Stock pay was 47% of revenue in 2021, 35% in 2023, 30% in 2024 and 17.5% in 2025, which is 22.1% without a one-off credit described below, and is on course for about 17% in 2026 C (calculated).¹²³² The fall is real and large. The 2025 figure needs a note, though. When the former Executive Chairman retired in December, awards he had not earned were forfeited and $38.7M of expense already booked was reversed. Without that credit, stock pay in 2025 would have been $185.5M, or 22.1% of revenue. The 2024 figure of $216.7M carried $24.0M of expense on the same awards C (calculated).³ The 2026 figure has no such effect, and it is the first clean one.

Where the margin came from

Almost all of the margin came from selling and overhead. From 2022 to the first nine months of 2026 the GAAP operating margin moved about 50 points, from −47% to 4%. Sales and marketing gave 24 of them, general and administrative 15, research and development 8 and gross margin 4, less one for restructuring C (calculated).¹²² Growth slowed over the same stretch from 34% to 15%. The company says it cut headcount by more than 20% from the third quarter of 2024, moved sales upmarket and stopped investing in non-core products.⁷ On the company’s non-GAAP basis, research and development (14.7% in the nine months) and general and administrative (9.3%) are already inside its 2028 targets of 14% to 15% and 9% to 10%, and sales and marketing (38.5%) is within about two points of its 36% to 37%.²⁷ The 2028 operating margin target of 24% to 26% is the margin it guides for 2026, 24.8%. So the cost cuts behind the last four years are mostly made, and what is left in the plan comes more slowly. The second table shows how much of the company’s own profit is stock pay. It is between 84% and 91% of what the company adds back to reach its non-GAAP figure in 2024, 2025 and the nine months of 2026, so the gap between the two profits is mostly one item C (calculated).

What an owner could really take out

Over the seven years Freshworks produced $433M of free cash flow in the company’s sense and booked $999M of stock pay, net of the 2025 credit, so on the measure I use the owner got −$566M C (calculated).¹²³ Over the last two years that flipped. Free cash flow less stock pay was $74M in 2025, or $35M without the credit, and $59M in nine months of 2026. For 2026 I put it at about $68M once I take out the interest the company earns on its cash, which I value separately.

Why take the interest out? The cash is already in the value, counted in full. Interest from it would count twice. It is also falling, from $51.7M in 2024 to $38.2M in 2025, and $19.0M in nine months of 2026, as the company spends the money on shares and deals.⁵³⁴ Interest was 17% of 2025 free cash flow. A simple price over free cash flow is 12.5 times, which looks low. On owner cash before deals it is 46 times C (calculated).

Quality of earnings

• The tax benefit. 2025 net income of $183.7M includes $151.7M from releasing a valuation allowance on US deferred tax assets. Without it, net income would have been about $32M C (calculated).³ In nine months of 2026, GAAP net income was $16.7M, or $0.06 a share.²
• Stock pay. The company’s non-GAAP profit leaves it out. It was $120.6M in nine months of 2026, and $311.9M of expense on unvested awards is still to come over 2.7 years.⁴ The company granted 22.2M units in nine months at a fair value of $11.16, which is $247M, against $121M of expense in the same period C (calculated). Part of the gap is timing, since awards vest over four years, and part is the low share price, which means the same dollars buy more shares; I cannot tell how much is which. The unvested balance rose to 28.5M units, 10.9% of shares outstanding.⁴
• Adjusted free cash flow. The headline figure adds back acquisition and restructuring costs paid, $7.5M in nine months of 2026, and so reads $187.4M against $179.9M on the plain definition.² Restructuring happened twice in two years, in November 2024 and May 2026.³⁴
• Working capital. Deferred revenue grew $61.2M in 2025, receivables took $28.1M and commissions paid exceeded commissions expensed by $8.4M, a net gain of $24.7M, or 21% of the year’s new revenue C (calculated). That is what I use. Customers pay in advance, and growth brings cash early; a slowdown would take it away.³
• Cash tax. Taxes paid were $14.4M in 2025, 8% of non-GAAP operating income. The company has US net operating losses of only $67.3M, and from 2026 it uses a long-term non-GAAP tax rate of 24%, so cash tax is likely to rise.³²
• Currency. About 28% of revenue is in euros or pounds, and nine months of 2026 include a $10.3M net foreign currency loss in the cash flow.³²
• Lawsuits. A securities class action over the offering, filed in November 2022, ended in a judgment for the company on summary judgment in April 2025, and the plaintiff has appealed. A derivative case filed in March 2023 is stayed behind it.³⁴

The balance sheet

On 30 September 2026 Freshworks had $466.7M of cash, $1.1M of restricted cash and $229.5M of marketable securities, $697.3M in all, and no debt.² Deferred revenue was $404.4M. Goodwill and acquired intangibles were $286.2M, up from $223.7M at the end of 2025 with FireHydrant.² In October it paid about $102.5M for Torii and Robusta, and $8.5M of indemnity holdbacks for FireHydrant are unpaid, which leaves about $586M, or $2.23 a share C (calculated).⁴ Purchase commitments run to $255.7M through 2028, nearly all cloud and software services.³ The company has $168.3M left on its buyback, and has said it plans to retain earnings and pay no dividend.⁴³

VIValuation

First, how this goes badly wrong

Before the numbers I asked what would most likely hurt an owner here, and worked back. I found three.

The seats go. Freshworks sells mostly by the seat, and its own figures show the seat is at risk. Its AI agents deflect about half of the IT tickets they see, and its Copilot lets an agent handle 52% more. If customers need fewer agents, per-seat revenue shrinks unless the new charges per AI session, per asset and per alert arrive faster than the seats leave. In my bear case growth falls to 2.5% and the shares are worth $3.97. Each point of revenue growth in every year is worth about $1.13 a share.

Stock pay stays high. It is 16.8% of revenue now. If it stayed there for good, my base falls from $10.34 to $6.72. Each point of revenue it comes down in every year is worth about $0.69 a share. The grants are the warning: 22.2M units went out in nine months, $247M at the grant price, against $121M of expense.

The cash goes on deals. The company has spent $429M on four companies in two and a half years, 5.8 times the free cash flow it kept after stock pay in 2025, or 12 times without the 2025 credit. Each extra point of revenue spent on deals in every year takes about $0.91 off my base. At the 2024 to 2026 pace, about 15% of revenue, with no extra growth credited, the base would be close to nothing. I do not think that is the right case, since deals buy revenue as well, but I cannot see how much, because the company does not say.

No part of this needs anything strange. They set my bear case, and they are why I start from the cash that reaches an owner, not from earnings or from the company’s free cash flow.

What an owner could really take out

I value cash after stock pay and after the capital needed to stand still. For a software company that means taking the pre-stock-pay operating cash, subtracting the stock pay as if it were paid in cash, subtracting cash tax and capital spending, adding the cash that comes early from customers who pay in advance, and then subtracting a charge for deals. I charge for deals because Freshworks has used them to build the asset and operations side of Freshservice, and its own plan says one to three tuck-in deals a year.³⁷

On this basis 2026 owner cash is about $68M, against free cash flow of $252M, non-GAAP operating income of $239M and GAAP net income of $17M for nine months E (estimate). That is 27% of the company’s free cash flow, and the $191M spent on deals is nearly three times the amount. A shareholder who looks only at the reported figure sees a stock at 12.5 times free cash flow. One who counts stock pay sees 46 times, and one who also counts the deals sees a negative number for 2026. Both are right about different things.

What the price implies

Put the other way, what does $12.01 need? At a 10% discount rate and 3.5% growth after 2041, it needs revenue to grow about 10.5% a year for ten years, from $965M to $2.61B, with my margins and stock pay. My base has 9.0%. Or it needs margins 2.4 points above my base every year. Or stock pay 2.4 points lower. The company’s own 2028 targets are revenue above $1.3B, a non-GAAP operating margin of 24% to 26% and stock pay of 13% to 14%; my base has $1.23B, a margin of 25.5% and stock pay of 15.5%.⁷ So the price needs the company to do a little less than it says, and a little more than I expect, for ten years.

What I can’t judge

• Whether AI adds or removes seats. The company’s numbers are encouraging and small: AI ARR was above $25M at the end of 2025, about 3% of revenue. Nobody can tell yet which way it breaks.⁹
• The stock pay path. Management targets 13% to 14% of revenue in 2028. Grants in 2026 are running at twice the expense. I allow 15.5%, but the range of honest answers runs from 12% to 18%.⁷⁴
• How management’s cash flow target is built. The 2028 target is a free cash flow margin of 32% to 34% from a non-GAAP operating margin of 24% to 26%, a gap of about eight points. In 2024 and 2025, once I take out interest, the gap was under one point. The deck does not explain the difference and interest income is falling. I do not use the target in the base.⁷³
• What the bought products earn. The company does not give the ARR it bought.
• Cash tax. I use 24% on profit after stock pay, with a floor of 8% of non-GAAP operating income, the 2025 ratio.³

Because I can’t know these, I show a range and the price’s place in it, and I do not treat $10.34 as more exact than it is. A fair reading of my work is that the base lies somewhere from $8 to $13. The two sensitivity tables below show how far one input moves it.

The discount rate

I ask 10.0%, the same as on my ServiceNow and Adobe pages. The ten-year Treasury yields 4.88%, so I am asking for about five points over it.²⁸¹⁷⁸ For a company of this size, with a controlled share structure and a question about its product, I could argue for more. At 9% my base is $12.24 and at 11% it is $8.96. At 11% the price is 34% above the base value.

Three cases

Bear

AI agents resolve more tickets than the new per-session charges replace, and seats stop growing. Revenue growth falls from 9% in 2027 to 2.5% by 2031 and stays there. Margin slips from 24% to 21% as AI compute and price cuts bite. Stock pay stays near 15% of revenue, because staff paid in a falling stock need more shares. Deals take 3% of revenue. Owner cash is $20M in 2030, and the business is worth little more than its cash. Value $3.97.

Base

Growth is 13% in 2027 and fades by about a point a year to 5% in 2036. Non-GAAP operating margin edges up from 25% to 26.8% in 2030 and 28.7% in 2036, inside the company’s 24% to 26% target for 2028. Stock pay eases from 16.5% to 13.5% of revenue in 2030 and 10.5% in 2036, slower than management’s plan. Deals take 8% of revenue to 2030, then 5%, then 3%. Revenue is $1.09B in 2027, $1.51B in 2030 and $2.28B in 2036. Value $10.34.

Bull

The company’s plan works and keeps working. Growth is 15.5% for two years, close to the 16% a year the 2028 revenue target needs, and margin reaches 30% in 2030 and 34% by 2036. Stock pay falls to 11% of revenue in 2030 and 8% later. Deals take 6% of revenue, then 3% and 2%. Revenue is $1.66B in 2030 and $2.86B in 2036. Value $21.37.

Other uses of the same dollar

The first alternative is a Treasury, at 4.88% for ten years with no risk to the coupon. Freshworks’ owner cash yield before deals is 2.2%, less than half of that, so the case rests on growth. On my base it returns 9.1% a year, about four points over the Treasury and a point short of my 10%.²⁸

The best comparisons on this site are two. ServiceNow is the closest business: it runs the IT service desk for large companies and asks the same question of AI agents.¹⁷ Adobe asks the same question about AI and has the best cash after stock pay of the software pages I compared.⁸ By my numbers, ServiceNow at its price returns 11.2% a year on my base cash flows and Adobe returns 15.0%, against 9.1% for Freshworks. I do not lean on the comparison, since they are larger, in some ways different, and each has its own risks. It says only that, by the standard I use, the Freshworks price is a less generous one than theirs.

The last two rows are the sharpest. Adobe and ServiceNow both spent heavily on shares. Adobe’s count fell, funded mostly by its large owner cash. ServiceNow’s did not, because stock pay absorbed the buying. Freshworks’ count fell 13.5% in under two years, but it paid for the buying with the cash pile from the listing, at least 4.7 times what it made after stock pay (6.6 times without the 2025 credit). A company can do that once.

What the market paid before

At the end of 2021 the market valued Freshworks at 15.8 times revenue, on growth of 49%. It is now 2.7 times, on growth of 15%. The multiple has fallen by a factor of six while revenue has risen 2.6 times. Growth fell from 49% to 15% over the same time.

The share price tells the same story with less arithmetic. It was $26.26 at the end of 2021, $12.25 at the end of 2025, as low as $6.88 on 23 February and as high as $15.49 in late October. In the four trading days after the third-quarter results it lost 15% while the company raised its guide.¹

The margin of safety

I set it at $7.24, 30% below my base value. That is more than the 25% I use on steadier businesses with net cash, for the same reason as on my ServiceNow page: the largest input, whether AI adds or removes seats, can’t be known, and being wrong about it costs more than being wrong about the discount rate.¹⁷ At $7.24 my bear case would still lose 45%, and my base would return 12.8% a year on the cash flows E (estimate). The price is 1.7 times that. The stock closed under $7.24 on 24 February 2026, at $7.16, and has not since.¹ It may never go back there. At $12.01 my own rule says I am not paid enough for the things I can’t see, and I would be depending on stock pay and the deals going well.

Two different questions

The first is how good the business is. It is a good one, and better than it was: gross margin 85%, retention in EX above 110%, a customer base moving to larger buyers, no debt, and a non-GAAP operating margin that went from 7.5% for 2023 to 29.8% in the third quarter of 2026.²³⁷ It is not yet a business that leaves much for its owners after it pays its staff in stock. The second question is how good the price is. At $12.01 the price is 16% above my base and 44% below my bull, and it needs revenue to grow about 10.5% a year for ten years on my margins, or the equivalent in margin or stock pay. Those are different answers. A business can be improving and the price can already know it. This looks like that case, and the market may see something I do not. The strongest form of that is that management delivers its plan for stock pay and margin. The weakest part of my own case is that I have counted the deals as a cost at a time when management says they are a plan.

Freshworks shares one idea with the pages for Adobe, Intuit, GoDaddy and Booking: the fear that AI will take the customer’s first click, or the seat, away from software. If the fear is right, all five are priced fairly or too high together. If it is wrong, all five look low together. This page differs on one point, since its price is above my base value and the others are below theirs. The question is the same and the answer moves all five at once, so I count them as one bet, not five, and I do not add them up.^8293031

VIIThe case for the price

What most people miss

Three facts cut in Freshworks’ favour. First, the core product keeps its customers better than the headline shows. In the first quarter EX customers retained and expanded at 111% and those with more than one EX product at 119%, while the blended rate was 106%.⁷¹³ Second, the cash. The company has about $586M after the October deals and no debt, or $2.23 a share, 19% of the price C (calculated).²⁴ Third, the two people who run the company bought stock in the open market, $4.98M in three trades since November 2025.¹⁰

The bull case

I found no named investor or analyst with a published view on Freshworks that I could check against a dated document. Nasdaq shows a one-year target of $16.00 without naming anyone, and the company’s slides of 14 May quoted an average analyst target of $12 with the stock at $8.50 on 12 May.¹⁵⁷ I could not read any broker’s note, so I give management’s case. In the third-quarter release of 3 November, Mr Woodside says the upmarket strategy is working, that customers want one platform and one data model for service operations, and that new multi-product EX deals rose over 35%.² In May its slides, under “Valuation discount,” showed the stock at 1.6 times revenue and 5.9 times free cash flow and a path to free cash flow per share of about $1.35 by 2028.⁷

In plain words: a company growing in the mid-teens, whose better half grows in the mid-twenties, whose non-GAAP margin rose by six or seven points a year in 2024 and 2025, with about a fifth of its value in cash. If stock pay keeps falling and the AI products add to the seat instead of replacing it, owner cash in 2028 is more than three times today’s.

Against reality

If Freshworks hits its 2028 targets, the price is possible. They are revenue above $1.3B, free cash flow above $425M, stock pay of 13% to 14% of revenue and net dollar retention above 108%.⁷ Then an owner would get about $230M a year after stock pay and without interest, against about $68M this year E (estimate). My bull case, $21.37, takes the plan’s revenue, margin and stock pay but not that cash target, and still charges for deals; the price is 44% below it. Growth is 14% where the revenue target needs about 16% a year. Net dollar retention is 104%. Free cash flow is on course for $252M, against the $265M the May slides showed for 2026, and the target is 69% higher in two years. And the plan puts free cash flow eight points of margin over operating margin, which it has not been, so the cash target is the part I trust least.²⁷

The plan needs four things together: EX growth in the mid-twenties through 2028, about 60%; stock pay down to 13% to 14%, about 50%; the margin holding at 24% to 26% while it spends on AI, about 80%; and AI adding to seats, about 70%. Multiplied, that is 17%, about one chance in 6. They are my guesses, not independent, and not used in the values E (estimate).

What a 10x would take. Ten times $12.01 is $120.10, or $31.5B of market value on today’s shares. At 25 times owner cash, that needs about $1.26B a year, and at an owner cash margin of 20%, about $6.3B of revenue. That is 6.5 times this year’s, or growth of 20.6% a year for ten years C (calculated). My bull case has 11.5%. Freshworks grew 30% a year from 2019 to 2025, starting from a fifth of today’s revenue, and is growing 15% now. Its stock has never closed above $50.25. ServiceNow shows it can be done: revenue went from $1.4B in 2016 to $13.3B in 2025, 28% a year, and it is the vendor Freshworks meets in large accounts.¹⁷ I could not open a published table of how often companies of this size keep up 20% growth for ten years, so I give no rate. A tenfold price within ten years is not plausible for a company growing 15% a year, and I would say so of any like it.

Where I come down

Three things moved me: gross margin, which rose from 79% to 85% in five years, the retention in EX, and the open-market purchases by the two executives. The free cash flow headline did not, because stock pay and the deals take most of it. The strong third quarter moved my 2026 numbers and not my base. What would raise my values is stock pay under 14% of revenue in 2027 with growth holding at 13%, or a disclosure of the ARR the bought products brought in that shows the price was low. Until then the price sits between my base and my bull.

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 idea. This page shares one idea with Adobe, Intuit, GoDaddy and Booking, the fear that AI will take away what these businesses sell. If my view of that fear is wrong, I am wrong on all five, so these are not five independent bets.^8293031
• India. A majority of employees are in India. Tighter US immigration rules, or a move in the rupee, would hit costs, and the 10-K lists both as risks.³
• Currency. About 28% of revenue is in euros and pounds, and a net foreign currency loss of $10.3M ran through the cash flow in nine months.³²
• Cloud supplier. Freshworks runs on Amazon Web Services, and a long failure there would stop the service.³
• Security. A breach of customer data would cost trust and money, and the product holds a company’s IT estate.³
• Lawsuits. The company won the securities class action filed in 2022 on summary judgment in April 2025, and the plaintiff has appealed. The derivative case of 2023 is stayed behind it. I do not know what an appeal could cost.³⁴
• Leadership. The chief revenue officer and the head of product are both new in 2026, and the founder has left the board.¹⁸¹⁹¹⁴
• Control. Class B, with ten votes a share, carries 48% of the vote, and the holders have been converting or selling. A few holders can still swing a vote until it ends in 2028 or 2029.³⁴
• Tax. The 2025 profit includes a $151.7M one-time tax benefit, and federal tax rules in the US may change.³
• Macro. IT budgets shrink in a downturn and the company’s own filing says customers have delayed or shortened contracts before.³