Apocalypse When
AI was supposed to be killing software. Revenue growth is running at a four-year high. The re-pricing that actually mattered happened in 2022 — and the market has been sorting the industry ever since.
Quick Summary
Growth accelerated while valuations collapsed. Software revenue grew 16.3% over the trailing twelve months, its fastest in four years, while multiples fell 38% between July 2025 and March 2026. The selloff repriced sentiment, not performance.
The change that mattered happened in 2022, not 2026. Profitable growth overtook unprofitable growth that year and never gave the lead back, widening the gap between the best and worst operating models from 2.3x to 5.6x — all of it before the AI selloff.
The industry has sorted into four operating models, and the market pays a ladder. Companies that grow and earn trade at 9.2x revenue; those that do neither trade at 1.8x. Forty percent of the companies in that bottom group in 2021 no longer trade at all.
The aggregate economics remain extraordinary — and almost nobody achieves them. The industry converts 47% of revenue into adjusted EBITDA and a fifth into free cash flow, but the average company earns a 4.2% net margin against the industry's 24.5%.
Today pundits, reporters, journalists, bloggers, podcasters, and analysts are reporting that the software industry is currently going through an AI-induced apocalypse — the SaaSpocalypse. Here we present various data on the state of the software industry, including growth, valuation, and operating characteristics to see if they suggest a massive cataclysm or doomsday, at least as of September 2026. This analysis is based on 250+ global software companies, representing $975B in trailing twelve-month (TTM) revenue [1].
Growth accelerated straight through the SaaSpocalypse
Thus far, the apocalypse is a story about 2025 and 2026, and it grew louder after the Anthropic coding releases of January and February 2026. Over precisely that window, industry revenue growth did not weaken. It accelerated — from 12.1% in 2024 to 13.7% in 2025 to 16.3% on a trailing-twelve-month (TTM) basis, its fastest since 2021 and above its ten-year average.
Valuations went the other way, and hard. Average multiples (market cap / revenue) held near 6.5x through mid-2025, turned down in August, fell steeply from January 2026, and bottomed in March at 4.1x — a 38% decline from July 2025. They have since recovered 28%, but at 5.3x remain roughly a fifth below where they started.
So far this is a repricing of sentiment, not of performance. The market marked software down by more than a third while the companies underneath grew faster than they had in four years. Nothing in the revenue data yet corroborates the thesis that AI coding tools are destroying software demand — which is not proof it will not happen but does mean it has not happened yet in the numbers. What it would look like if it were starting is set out at the end of this article.
Two qualifications. Excluding Microsoft, aggregate TTM growth drops to 15.6%, still the best in several years and above the ten-year average. And the aggregate conceals a sharp size gradient — the largest revenue quartile grows far faster than the smallest, so the acceleration is not evenly shared. And growth and market cap are accruing to the top 3 revenue quartiles.
The aggregate[2] and the average have come apart:
In 2021 the average company was valued above the revenue-weighted aggregate — small companies carried a premium. That has reversed and widened: the aggregate now sits 70% above the average. Large companies held their multiples; the typical company did not. The small-cap valuation premium has inverted outright.
The operating-model shift before any apocalypse
To explain the operating model shift we use the 2x2 shown below to segment the population of software companies.
Every company is placed by two numbers: revenue growth (1 Year), and adjusted EBITDA margin[3]. The grid is cut at the medians — 12.5% growth and 23.2% margin for TTM 2026.
Figures above are average market cap / revenue, TTM September 2026 within each quadrant.
"High" here means above the median — nothing more. A company growing 13% is therefore filed under "high growth," which is a generous use of the word. The labels are shorthand for above-median and below-median within this universe, not judgments about absolute performance.
Two consequences follow. Because the split is at the median, roughly a quarter of companies land in each quadrant by construction — so the interesting fact is never how many are in a box, it is what the market pays for each box. And because each period is cut at its own median, the bar moves: 2021's growth line sat at 25.5%, twice today's.
In 2021 the industry faced a real constraint: two-thirds of companies sat on the off-diagonal, growing or earning but rarely both. Today the four quadrants are almost exactly even. Companies did not shift in one direction; the population bifurcated, into those that learned to do both and those that do neither.
Read the 2021 column first. Unprofitable growth was the most valuable profile in software, at 18.0x — ahead of the 16.4x paid to companies that managed both. That is the bubble stated as a single number, and it has since reversed completely.
Then read the change column. Both high-margin quadrants fell by roughly 40%. Both low-margin quadrants fell by 69% and 73%. Profitability, not growth, determined who kept their valuation through the reset. Today growth still commands more than margin — 5.5 against 4.3 — but getting here required margin.
Sustained balance is worth more than newly found balance
Companies that never left the balanced quadrant are worth at least 50% more than those that arrived in it [4]. The market is pricing durability, not just the current-period combination.
Exits were concentrated, and they left the sample
Forty percent of the companies that were neither growing nor earning in 2021 no longer trade. Without the delisted companies restored to the dataset, those exits would have silently disappeared and the bottom quadrant would look far healthier than it was. One caveat matters throughout this section: these companies are classified by where they stood in 2021, not by where they stood when they were bought. Several had moved a long way in between.
It is not a failure rate. Nine of those 20 carried gross margins above 60%, eight had revenue above $500M, and the best of them left at a premium — CyberArk at 16.6x revenue and Nuance at 13.1x, against a median exit of 3.3x. What they had in common was a conclusion, reached by boards and buyers alike, that the transition could not be finished in public.
The buyers split evenly. Financial buyers took half — Thoma Bravo alone took Verint, Dayforce, PROS and Bottomline. Strategic buyers took the rest, and far more by value: Cisco paid $28B for Splunk, Microsoft $19.7B for Nuance. Three of the strategic acquirers were themselves private-equity owned, which puts the financial share nearer two-thirds.
CyberArk is the clearest case of a company that moved. It was in the bottom quadrant in 2021 on 8.3% growth and a 7.2% margin, left the following year, and never returned: growth ran 17.7%, 27.1%, 33.1% and 36.0% over the next four years while margin was rebuilt from 0.6% to 21.6%, and the multiple recovered from 9.0x to 16.6x. Palo Alto Networks paid roughly $25B for it in February 2026. Synchronoss is the control case — growth of −3.8%, −38.1%, −5.5% and 5.7%, a multiple that never rose above 0.66x, and an exit to Lumine Group at 0.6x. Same quadrant in 2021, nearly thirty times apart on the way out.
Depressed growth and margin during a business-model transition is not a death sentence. It is the price of the transition — and whether the exit comes at 16.6x or 0.6x depends on whether the transition finished.
One more structural note: new listings collapsed after 2021: 49 of today's companies listed that year, one in 2022, two in 2023. The window has been reopening since — five in 2024, seven in 2025 — but four thin years are a large part of why the population is smaller than it was.
When the re-pricing actually happened
The inversion happened in 2022. That is the year profitable growth (7.1x) overtook unprofitable growth (5.9x), and it has not reversed since. From there the gap between the best and worst quadrant widened every year — 2.3x to 5.6x by FY2025. All of that pre-dates the AI selloff.
The AI selloff did none of that work
The market sold growth, not weakness[5]. The two high-growth quadrants de-rated close to 40%; the two low-growth quadrants fell about 30%. The hardest-hit quadrants were the two high-growth ones, including the one carrying the best operating model. Whatever was being repriced, it was not current profitability; it was the durability of the growth.
The market re-priced software operating models between 2022 and 2025, widening the best-to-worst gap from 2.3x to 5.6x. The AI-driven selloff that followed did none of that work: it took roughly a third off everything, tilted slightly against high-growth names, and left the ranking it inherited intact. The drawdown looks indiscriminate because there was no sorting left to do — the panic could move the level, not the order.
The Rule of 40 is a weak screen but a strong classifier
Rule of 40 here is defined as the sum of 1-year growth rate and adjusted EBITDA margin[6].
Nearly half of all software companies clear the bar, so as a filter it separates very little. But it sorts valuation almost 3 to 1. The Rule of 40 is unselective and highly informative at the same time.
The obvious objection to the balanced version is that it is just a proxy for growth. It is not. Holding growth above 20% and splitting on margin:
Among fast growers alone, profitability nearly doubles the multiple.
The same result on free cash flow
Free cash flow is a stricter test than adjusted EBITDA — it charges capex, cash taxes, interest and working capital. The table below compares using free cash flow instead of adjusted EBITDA in the Rule of 40 equation.
Re-running every cut on FCF margin barely moves the answer.
Every conclusion survives on the industry's own cash metric, on a stricter bar that only 38% of companies clear rather than 48%.
FCF also sharpens the central finding. In FY2021 the market paid essentially nothing for cash generation among fast growers — 19.09x for those below the median FCF margin against 19.54x for those above it, a dead heat. Today the same comparison is 9.0x against 5.3x: a 70% premium for growth that pays its own way.
What this means going forward
Step back from the panic and the industry's aggregate economics are extraordinary — and largely unchanged by it. Software grows at roughly three times nominal GDP while generating adjusted EBITDA margin of 47% and converting a fifth of revenue into free cash. Keep in mind that aggregate numbers represent the operating characteristics of the entire industry.
The distribution is the story. The industry earns a 24.5% net margin; the typical company inside it earns 4.2%. Gross margin is nearly identical for everyone — 69.8% against 68.3% — so what it costs to run and deliver software is not what separates them; it's the operating costs that fall below gross profit.
The future is here; it is very unevenly distributed.
It’s important to note that the industry aggregate is essentially the target operating model for most software companies and is only achieved at significant scale. You will find that the largest companies in the industry operate more closely to this model than the smaller ones. The smaller companies have not achieved escape velocity and have to invest in order to achieve such velocity.
Since the 19th century, the word “apocalypse” has been used to mean cataclysm. That is not what it originally meant. It meant to reveal. On the older definition the last four years qualify: nothing ended, but a great deal was uncovered — which companies had built a business that funds itself, and which had been carried by a market that briefly stopped asking.
For operators
The old excuse is gone. Growth and margin stopped trading off against each other in 2023. "We could be profitable if we stopped growing" is no longer a structural claim about software; two thirds of the industry now grows faster than 5% while earning an adjusted EBITDA above 10% — up from less than half in 2021.
Growth is still the main engine — but it has to fund itself. Growth outranks margin in what the market pays, so this is not an argument for harvesting. It is an argument against growth that does not pay its way: that profile has gone from the most valuable thing in software at 18.0x to 5.5x, below companies that grow more slowly with cash behind them.
Arriving late earns less than never leaving. Companies that have held the top quadrant since 2021 trade at 14.3x; those that arrived recently trade at 6.7-9.5x. The market is paying for a track record it believes, which takes years to build and is the one input a turnaround cannot buy.
The bottom quadrant is not a resting place. Of the companies that were neither growing nor earning in 2021, 40% no longer trade independently.
For investors
Few industries post this combination at the aggregate level, and the recent drawdown has not changed the underlying economics — revenue growth accelerated while multiples fell by more than a third. But the aggregate is not purchasable. You cannot buy a 24.5% net margin; you buy individual companies, and the average one earns 4.2%.
Dispersion has widened accordingly. The gap between the best and worst operating model went from 2.3x to 5.6x between 2022 and 2025. Sector exposure captured most of the return in 2021, when the whole industry re-rated together. It captures much less now, when a nearly fivefold valuation gap separates the quadrants and 40% of one of them disappears within five years.
What would change this conclusion
The claim here is narrow: AI has not yet shown up in software revenue. Here is what would show that it had.
If AI is destroying demand, it appears first where products are least differentiated and most seat-based — at the small end. The signal is not the growth rate but the incidence of decline. Among the largest half of the industry, five companies out of 132 are shrinking. In the third quartile it is nine of 66. In the smallest quartile it is twenty of 66 — nearly a third.
That is not yet evidence of an apocalypse. Small software companies were more fragile than large ones long before AI, and the seventy percent of the smallest quartile that is still growing is growing at 14%. What would be new is the pattern moving up-market.
Three markers, then. If the share of shrinking companies rises materially in the second and third quartiles, where scale should still offer protection, the thesis is working as its proponents describe. If aggregate growth breaks below the low double digits after three consecutive periods of acceleration, that is a reversal rather than noise. And if the valuation gap between quadrants compresses, it will mean investors have stopped believing that operating model predicts survival — the market abandoning the sorting it has done since 2022.
None of those has happened. The first one that does is the one to take seriously.
This is industry analysis, not investment advice, and past dispersion is not a forecast of future returns.
[1] 259 active public software companies, 474 total software companies, including those that have been delisted (through strategic acquisition or being taken private at some point in the past five years).
[2] Aggregate market cap/revenue = sum of market cap of all companies divided by sum of revenue of all companies.
[3] Quadrant valuation tables use companies that have growth, margin and a market cap for that period (297 in FY2021, 259 in TTM). Delisted companies are included in each historical period they were public for.
[4] Six companies in the High Growth/Hi Margin quadrant lack 2021 data and are excluded from the movement chart.
[5] Quadrant valuation levels use the latest available market cap (8 September 2026). The monthly series and the drawdown table use calendar month-ends, so they run through 31 August 2026.
[6] Adjusted EBITDA = EBITDA plus stock-based compensation.
Comments