From Madhur Jain | Product & Market Analysis

SaaS Buybacks vs Stock Pay: $7.2B Spent, 37% Only Offset New Shares

On this page

Eight public software companies spent $7.18 billion on share repurchases in their latest year-to-date filings through 30 September 2026. Roughly $2.66 billion of that, about 37%, did not shrink anyone's share count. It only replaced shares issued to employees and others over the same months. SaaS buybacks are real, but the share count tells you how much of each dollar actually came back to you.

Key takeaways

  • Paycom retired about 96 shares for every 100 it bought back. Its count fell from 54.8 million to 44.3 million in nine months, and its stock pay was only 3.7% of its buyback spend.
  • Twilio's share count rose through a $448 million buyback. It bought 3.1 million shares in nine months, yet shares outstanding grew by 2.34 million, so issuance outran repurchases.
  • ServiceNow spent $2.23 billion to cut its count by 9.4 million shares. It bought 20.1 million shares, so more than half of the buyback only offset new issuance.
  • Stock pay as a share of buyback spend predicts the result. The four companies with stock pay below 40% of repurchases retired 68% to 96% of the shares they bought. The three above 60% retired 47% or less.

Short answer: Do SaaS buybacks reduce the share count? Partly. Across eight Q3 2026 filings, the share of repurchased stock that actually came off the count ranged from about 96% at Paycom to below zero at Twilio. The gap is explained mostly by how much stock each company issues to employees.

$7.18BCombined buyback cash across ServiceNow, Atlassian, Paycom, HubSpot, Dropbox, Twilio, Dynatrace and Qualys, latest year-to-date windows. Source: company 10-Q filings, 2026.
About 37%Share of that $7.18B that only offset new shares issued over the same months, roughly $2.66B. Source: Zan Digital calculation from the filings.
+2.34MNet rise in Twilio's shares outstanding in nine months of 2026, despite $448M of repurchase cash. Source: Twilio Q3 2026 10-Q.

The SaaS buyback ranking: net shares retired per dollar

The table below ranks eight software companies by one question. Of the shares each one bought back, how many actually left the share count? That is the retirement rate. It is the number a buyback press release never prints.

Every figure comes from each company's own quarterly filing for the period ending 30 September 2026. The window is the fiscal year to date, which is nine months for six companies. Dynatrace's fiscal year starts in April and Atlassian's in July, so their windows are six months and three months.

SaaS share repurchases ranked by retirement rate, year to date through 30 September 2026
CompanyWindowBuyback cashShares boughtNet fall in shares outstandingRetirement rateNet shares retired per $1BStock pay as % of buyback
Paycom9 months$1,390.9M10.96M10.5M96%7.5M3.7%
Dropbox9 months$847.3M31.6M26.1M83%30.8M26.6%
HubSpot9 months$973.7M4.60M3.72M81%3.8M38.1%
Dynatrace6 months$522.8M12.27M8.35M68%16.0M29.0%
Qualys9 months$185.6M1.63M0.81M49%4.3M31.4%
ServiceNow9 months$2,225M20.1M9.36M47%4.2M82.2%
Atlassian3 months$588.0M3.34M1.25M37%2.1M60.6%
Twilio9 months$448.3M3.1MRose 2.34MNegative (about -75%)Negative96.4%

Buyback cash is the repurchase line in each cash-flow statement. Shares bought come from the equity note or equity statement. Shares outstanding are period-end balance sheet counts against the prior fiscal year end. Paycom's counts are rounded to 0.1 million in its filing, so its rate is approximate. Stock pay is the stock-based compensation add-back in the cash-flow statement for the same window.

The ranking by retirement rate and the ranking by "net shares retired per $1 billion" are not the same list. The second one depends on share price. Dropbox bought stock at an average of about $27 a share, so each billion buys far more shares than at HubSpot's average of about $213. That is why the retirement rate is the fairer comparison, and it is the one this post ranks on.

Of every 100 shares bought back, how many left the count? Retirement rate, year to date through 30 Sep 2026. Bars left of the line mean the count rose. Paycom96% Dropbox83% HubSpot81% Dynatrace68% Qualys49% ServiceNow47% Atlassian37% Twilio-75% Source: company 10-Q filings for the period ending 30 Sep 2026. Zan Digital calculation.
The dark bars are the four companies whose buybacks mostly shrank the count. The red bar is the one where the count grew anyway.

How the SaaS buyback calculation works

A buyback reduces the share count only if the company issues fewer new shares than it buys. Software companies issue new shares constantly, mainly to employees. So the honest measure of a buyback is the net change, not the gross purchase.

The formula, step by step

Take shares outstanding at the end of the prior fiscal year. Take shares outstanding at 30 September 2026. The difference is the net reduction. Divide it by the number of shares the company says it repurchased in the same window. That ratio is the retirement rate.

ServiceNow is the cleanest worked example. Its balance sheet shows 1,047.3 million shares at the end of 2025 and 1,037.9 million at the end of September. That is a net fall of 9.36 million. Over the same nine months it bought 18.5 million shares through an accelerated repurchase and 1.6 million more on the open market. So 9.36 divided by 20.1 gives 47%.

The dollar version follows from that. If 53% of the shares bought were matched by new issuance, then roughly 53% of the dollars spent went to standing still. For ServiceNow that is about $1.19 billion of its $2.23 billion.

Why period-end counts, not weighted averages

Earnings per share uses a weighted-average share count. That figure lags, because shares bought late in a period count for only part of it. Period-end shares outstanding show where the count actually landed on the last day.

I checked the weighted figures as a cross-reference anyway. They point the same way. Paycom's basic weighted-average count fell 21.1% between the third quarters of 2025 and 2026, and Dropbox's fell 16.4%. ServiceNow's fell 0.3%. Twilio's rose 0.8%.

Basic weighted-average shares, Q3 2025 against Q3 2026, with trailing 12-month buyback cash
CompanyBasic shares, Q3 2025Basic shares, Q3 2026ChangeBuyback cash, trailing 12 months
Paycom56.1M44.3M-21.1%$1.50B
Dropbox261.1M218.2M-16.4%$1.26B
HubSpot52.5M49.5M-5.8%$0.99B
Dynatrace301.6M288.0M-4.5%$0.91B
Atlassian263.0M252.1M-4.2%$2.14B
Qualys36.0M34.8M-3.4%$0.23B
ServiceNow1,038M1,035M-0.3%$2.82B
Twilio153.1M154.4M+0.8%$0.65B

Trailing 12 months is the prior full fiscal year's repurchase cash, minus the prior-year year-to-date figure, plus the current year-to-date figure. Atlassian and Dynatrace use their own fiscal quarters ending 30 September. Basic counts are used because Atlassian reported a loss in the comparison quarter, which makes its diluted count equal its basic count.

Read the bottom two rows together. ServiceNow and Atlassian spent the most in the trailing year, $2.82 billion and $2.14 billion. They also show some of the smallest declines in weighted shares. Spending the most and shrinking the least is the pattern this post exists to surface.

Why SaaS buybacks leak: stock-based compensation

Stock-based compensation is pay delivered in equity rather than cash. Software companies use it heavily because it conserves cash and ties staff to the share price. Every vested award either adds a share to the count or forces the company to spend cash to avoid that.

That is why stock pay and buybacks belong on the same page. A company that pays its staff partly in new shares and then buys shares back is, in part, running money in a circle.

Stock pay as a share of buyback spend

Across the eight companies, stock-based compensation in the same windows totalled $3.48 billion, equal to 48% of buyback spend. The spread inside that average is wide. Paycom's stock pay was 3.7% of its repurchases. Twilio's was 96.4%.

The relationship in one chart

Plot each company's stock pay ratio against its retirement rate and a clear line appears. The four companies below 40% retired between 68% and 96% of what they bought. Qualys, at 31.4%, is the exception, and it retired 49%. Above 60%, no company retired more than 47%.

More stock pay per buyback dollar, fewer shares actually retired X: stock-based pay as % of buyback cash. Y: retirement rate. Year to date through 30 Sep 2026. 100% 0% -80% 0% 50% 100% Paycom Dropbox HubSpot Dynatrace Qualys Atlassian ServiceNow Twilio Source: company 10-Q filings, 2026. Eight points, so read it as a pattern, not a model.
Paycom sits top left with little stock pay and almost full retirement. Twilio sits bottom right, where stock pay nearly matched the buyback and the count still rose.

Stock pay is an expense, not a share count

One caution on the chart. Stock-based compensation is an accounting expense measured at grant-date value. It does not convert one for one into shares issued this year. Awards vest over several years, and options and the employee stock purchase plan add shares on their own timing. The ratio is a good predictor here, not an identity.

ServiceNow: $2.2 billion of buybacks for 9.4 million fewer shares

ServiceNow is the largest spender in the set and the clearest case of buyback leakage. It bought nothing in the third quarter itself. Its nine-month repurchase cash was $2,225 million, all of it before July, most of it a $2 billion accelerated repurchase in the first quarter at an average of $107.97 a share.

Over the same nine months it recorded $1,829 million of stock-based compensation. It also paid $456 million in taxes on behalf of employees whose awards vested, a figure the next section returns to. Put those numbers next to each other and the buyback looks less like a return of capital and more like a cost of running the payroll.

ServiceNow share count, 31 Dec 2025 to 30 Sep 2026 Millions of shares, split-adjusted. The axis starts at 1,020 million to make the movement visible. 1,047.3M End of 2025 -20.1M Bought back +10.7M Issued, implied 1,037.9M 30 Sep 2026 Source: ServiceNow Q3 2026 10-Q. Issued is the balancing figure: shares bought minus the net fall.
The red bar is what the press release does not show. Roughly 10.7 million new shares refilled more than half the hole the buyback dug.

The implied issuance of 10.7 million shares is a balancing figure, not a line the company reports. It captures everything that added shares: vesting awards, option exercises, the employee purchase plan and any stock used in deals. The filing does not split it for me, so I do not split it either.

My view is that ServiceNow's buyback is best read as dilution management with a margin of real shrinkage on top. That is a defensible choice for a company paying $1.8 billion of stock comp in nine months. It is not the same thing as returning $2.2 billion to holders. For what the market is pricing on the operating side, see the rebuild of the ServiceNow bear case from its own filings.

Twilio: the share count rose through the buyback

Twilio is the only company in the set whose share count went up. Its 10-Q reports 3.1 million shares repurchased for $446.7 million in nine months. Shares outstanding still rose from 152.4 million to 154.7 million.

That implies roughly 5.4 million shares added from other sources in nine months. Against that, stock-based compensation was $432 million, almost equal to the buyback cash. Twilio's own release says it repurchased $127.2 million in the third quarter alone, and roughly $700 million of a $2.0 billion authorisation remains.

A falling buyback pace makes the gap worse. Twilio bought 6.4 million shares in the first nine months of 2025 and 3.1 million in the same period of 2026. Dollars fell less than shares, which says the average price paid rose sharply. A buyback budget buys fewer shares as the stock climbs, while award grants keep arriving.

I would not call Twilio's buyback wasted. Without it, the count would have risen by roughly 5.4 million shares instead of 2.3 million. But a holder who reads "$448 million returned" and expects a smaller company is reading the wrong number. Palantir shows the same tension from the other direction in the comparison of its revenue growth against its stock pay growth.

Paycom, Dropbox and HubSpot: SaaS buybacks that shrank the count

Three companies converted most of their buyback spend into a smaller count. They share one trait: stock pay that is small relative to the cash they put into repurchases.

Paycom is the extreme. It spent $1,390.9 million in nine months, repurchasing about 10.96 million shares at an average cost of $128.56. Its shares outstanding fell from 54.8 million to 44.3 million, about 19% in nine months. Stock pay was $52.0 million over the same window. Little was issued, so nearly everything bought stayed retired. It still holds $1,652.7 million of authorisation.

Dropbox bought 31.6 million shares for $847.3 million, at about $27 a share. Its count fell by 26.1 million, a retirement rate of 83%. Dropbox has run this playbook for years: in 2025 alone it repurchased about 60.4 million shares for $1.7 billion. The trade-off is visible in its revenue, which grew 0.5% year on year in the quarter. A company returning capital this aggressively is telling you where it thinks the best return sits.

HubSpot bought 4.6 million shares for $980.1 million at an average of $213.18. Its count fell by 3.7 million, an 81% retirement rate. The interesting detail is price. A year earlier HubSpot paid an average of $501.67 a share. The buyback stepped up after the valuation reset, which is the textbook case for repurchasing. How the business is holding up beneath it is covered in the HubSpot Q3 agent seat test.

Dynatrace and Qualys sit in the middle

Dynatrace retired 68% of the 12.3 million shares it bought over six months, and announced a further $1 billion authorisation with its results. Qualys retired 49% of 1.63 million shares. Its stock pay ratio of 31% is similar to Dynatrace's, so its lower rate shows the ratio is a pattern, not a rule. With a buyback this small, a few hundred thousand shares of issuance moves the rate a long way.

The hidden SaaS buyback: tax withholding on vested shares

There is a second channel for buying back stock that never appears in the buyback headline. When restricted stock units vest, many companies withhold some of the shares and pay the employee's tax in cash. Economically, that is a buyback. The shares never reach the market, and cash leaves the company.

This channel is large at some companies in the set. ServiceNow paid $456 million in net share settlement taxes in nine months, on top of its $2,225 million repurchase line. Dropbox paid $103.3 million. Qualys paid $19.5 million and Paycom $18.2 million.

Add ServiceNow's two lines together and it spent about $2.68 billion in cash in nine months on managing its share count. Of that, the count fell by 9.4 million shares. The cash figure most investors quote is the smaller one.

Atlassian's filing shows no separate net share settlement line, so its true cash cost of dilution is not visible from the cash-flow statement. That is a disclosure gap worth noting rather than guessing around. The rest of Atlassian's quarter is covered in the Atlassian Q1 FY27 cloud and Rovo breakdown.

Where this buyback analysis is weakest

Every number above traces to a filing. The way they are combined still carries real limits, and three of them matter.

The windows are not the same length

Six companies report nine months. Dynatrace reports six and Atlassian three. A single quarter is noisy. Atlassian's 37% rate rests on 3.34 million shares bought and a net fall of 1.25 million. A big vesting date inside one quarter can move that a lot. I would not rank Atlassian below Qualys with confidence on this data alone.

There is also a data quality wrinkle. Atlassian's 10-Q balance sheet and its equity statement print different Class A share counts. Both show a fall of about 1.24 million shares over the quarter, so the rate holds, but the levels do not reconcile in the filing.

Issuance is not all stock pay

The implied issuance figure catches everything that adds shares, including acquisitions paid in stock and convertible notes. I did not decompose it, and the press releases do not either. If one of these companies issued shares for a deal in the window, part of its "leak" is an acquisition, not payroll. That would change the story for that company, though not the arithmetic.

The case for buybacks that leak also deserves a fair hearing. Buying shares to offset dilution keeps the per-share value of the business steady while the company pays staff in stock. If the alternative is paying the same people in cash, the buyback is simply the cash cost of that pay, deferred. The flaw is not the buyback. It is calling it capital return.

Frequently asked questions

Do share buybacks reduce share count at SaaS companies?

Partly, and the amount varies widely. In eight filings for periods ending 30 September 2026, Paycom's count fell by about 96% of the shares it repurchased, while Twilio's count rose despite buying 3.1 million shares. ServiceNow retired about 47% of what it bought. The difference comes from how many new shares each company issued over the same months, mostly through employee equity awards.

How much stock-based compensation do SaaS companies pay compared with buybacks?

Across ServiceNow, Atlassian, Paycom, HubSpot, Dropbox, Twilio, Dynatrace and Qualys, stock-based compensation totalled about $3.48 billion in the latest year-to-date windows. Buybacks totalled about $7.18 billion, so stock pay equalled roughly 48% of repurchase spend. The range was 3.7% at Paycom to 96.4% at Twilio, and companies with lower ratios retired more of the shares they bought.

Why did Twilio's share count rise despite buybacks?

Twilio repurchased 3.1 million shares for $446.7 million in the nine months to 30 September 2026. Over the same period, shares outstanding rose from 152.4 million to 154.7 million. That implies about 5.4 million shares were added from other sources, such as vesting equity awards. Its stock-based compensation of $432 million nearly matched its buyback cash, so issuance outran repurchases.

What is the difference between basic and diluted share count?

Basic shares count stock actually outstanding, weighted across the period. Diluted shares add stock that options, unvested awards and convertibles would create if exercised. A company reporting a loss shows the same number for both, because adding shares would make the loss per share look smaller. That is why this analysis compares basic counts year on year.

How do you calculate net share reduction from a buyback?

Subtract shares outstanding at the end of the period from shares outstanding at the start. That is the net reduction. Divide it by the number of shares the company says it repurchased in the same period to get the retirement rate. Both inputs are in the 10-Q: the balance sheet or cover page for counts, and the equity note for shares bought.

Which SaaS company had the most efficient buyback in Q3 2026?

On retirement rate, Paycom ranked first of the eight companies reviewed. Its shares outstanding fell from 54.8 million to 44.3 million in nine months after it repurchased about 10.96 million shares, a rate near 96%. Its stock pay was only 3.7% of its buyback cash. Dropbox and HubSpot followed at 83% and 81%.

Where to start: test your next buyback headline

Pick one software stock you hold or follow. Open its latest 10-Q and find two numbers: shares outstanding on the balance sheet at the start and end of the year, and shares repurchased in the equity note. Divide the change by the repurchases. It takes about ten minutes, and it tells you more than the press release does.

Then add the stock-based compensation line from the cash-flow statement and divide it by the buyback cash. If the result is above 60%, treat the next buyback announcement as a payroll expense rather than a return of capital. On this set of filings, that rule would have sorted every company correctly except Qualys. For the wider context on why software valuations reset in the first place, see the explanation of the SaaS selloff.

References

  1. ServiceNow, Form 10-Q for the quarter ended 30 September 2026, with Q3 2026 earnings release (Exhibit 99.1, 28 October 2026). Used for repurchases, stock pay, withholding taxes and share counts.
  2. Twilio, Form 10-Q for the quarter ended 30 September 2026, and Q3 2026 earnings release (29 October 2026). Used for repurchases, stock pay and Class A share counts.
  3. Paycom Software, Form 10-Q for the quarter ended 30 September 2026, filed 5 November 2026. Used for repurchases, average cost, stock pay and share counts.
  4. HubSpot, Q3 2026 earnings release, Exhibit 99.1, 4 November 2026, and Form 10-Q. Used for repurchases, average prices, stock pay and share counts.
  5. Dropbox, Q3 2026 earnings release, Exhibit 99.1, 5 November 2026, and Form 10-Q. Used for repurchases, stock pay, withholding taxes and share counts.
  6. Atlassian, Form 10-Q for the quarter ended 30 September 2026, and Q1 FY2027 earnings release (5 November 2026). Used for repurchases, stock pay and Class A and B counts.
  7. Dynatrace, Form 10-Q for the quarter ended 30 September 2026, and Q2 FY2027 earnings release (3 November 2026). Used for repurchases, stock pay and share counts.
  8. Qualys, Form 10-Q for the quarter ended 30 September 2026, filed 9 November 2026. Used for repurchases, stock pay, withholding taxes and share counts.

The weakest part of this source base is the derived figures. Retirement rates, implied issuance and the 37% estimate are Zan Digital calculations from reported lines, not numbers any company publishes, and the year-to-date windows differ in length across the eight companies. All filings retrieved 12 November 2026.

AV
Madhur Jain
Writes for Zan Digital about AI product economics, B2B software markets and what the numbers behind vendor claims actually say.

Related reading