Software / SaaS de-rating Multiples compress ▼ — but three September prints say the disruption fear ran ahead of the evidence

App Economy (Jun 13): the "SaaSpocalypse" — AI-disruption fear is compressing software multiples even on clean prints. Adobe is −40% YTD at just 9× forward EBITDA despite +13% revenue and AI-first ARR tripling past $500M; its answer is a freemium pivot for Firefly/Express ("the Acrobat Reader playbook applied to AI," creative freemium MAU 50M→90M), judged on paid conversion. DocuSign's credit-based IAM platform (12.6%→18% of ARR) is the same shift from per-seat pricing. The re-rate holds until the AI-native funnels prove they convert. App Economy (Jun 6): the disruption cuts both ways — Veeva’s Falcon brings agentic "headless" labor (agents that replace a job function, not assist) while GitLab pairs its AI pivot with a 14% layoff; (Jun 5) AI is also a genuine tailwind for cybersecurity (Palo Alto, CrowdStrike) — but the thesis is "already priced," both beat-and-raised yet fell. App Economy (May 29-30): the deeper split is consumption-vs-seat — Snowflake (+38%, usage grows as agents work harder) beats seat-based Salesforce (the "penalty box," where AI may erode seats); and the security/data names show "beat-but-guide-spooks" deceleration (Zscaler −21% on a 16-17% FY27 guide, Elastic −11%). Carlson (May 26): post-implosion (multiples 30s→teens), software bifurcates — distribution + proprietary-data + system-of-record names rerate UP (SPGI/MCO/FDS/MSCI, plus AI-insulated UBER/DASH/SHOP/ABNB and learned-behavior SPOT/DUOL), while commoditized single-feature SaaS rerates DOWN on bundling/AI-copy risk (DOCU e-sig, ZM vs Teams), with CRM/ADBE/INTU/ADSK "less predictable." Jacobs (Mar 11): the de-rating has bled into structurally different businesses — auto-insurance lead-gen marketplaces "got lumped in with the software and SaaS scare, but it's so different" — a source of mispricing (see "AI victims vs winners — value" below). Link (Halftime, Jun 29): with IGV −17% YTD vs SOXX +85%, the relative value is in software — but only 'mission-critical' names a customer can't rip out: IBM (her final trade, ~18x), Snowflake, Synopsys. Goldman Sachs (Borges, Jul 8): the de-rating is now mechanical in the sell-side marks — GS cut ServiceNow's target to $145 from $163 (27x from 30x) purely on lower software comps while keeping Buy and near-unchanged estimates; for AI-exposed SaaS the re-rating question is proving relevance in the enterprise-AI stack (stabilizing organic revenue + upward whole-company revisions the proof points). Singh (Jul 26): Niles — AI cost control comes "at the expense of non-security software companies, IT services and headcount" (the call's one new thematic short). IBM is the case study ("next quarter could miss as well… I just don't see a reason to really own it"). The exceptions: the platforms and the overlooked cheap — JPMorgan's CIO says the bank still relies on ServiceNow ("the AI organization engine"), and Roper beat-and-raised +10% "because nobody was looking at that name." Aitken (Jul 28), the other side of the trade: he owns CGI through a four-to-five-quarter derating and is "closer to being buyers than sellers" — "we're really not in that camp" on AI disruption, though he won't claim a five-to-ten-year crystal ball. The numbers are "still very much intact." Managements across the cohort say AI is a tailwind; he discounts that ("management teams will be the last to know") and expects truth "somewhere in between." Shopify: formidable moat, no disruption visible in results, but he won't argue it's cheap. Ives/Luria (Eisman Ep 70, Jul 27): the SaaSpocalypse is mostly private — PE-owned software was milked instead of reinvested (Medallia already failed) while listed software is net-cash: "software debt is private equity." The mechanism is CIO consolidation from ~100 packages to ~30 ("guess what I'm cutting? All the small ones"). Luria condemns CRM "regardless of AI" ("has not been adding value in years… it keeps charging more"); Ives names ADBE and INTU as franchises that "miscalculated what AI is going to do to the business model" (the 1995 typewriter company); IBM's pre-announce was "stunning"; ServiceNow explicitly excluded from the loser bucket. Eisman endorses the framework in his own voice (Jul 31). 2026-AUG-10 — Steve Eisman (Ep 72): the public-comp ruler — "ServiceNow is down over 50% from its peak. So, any software company owned by private equity has to be down something like 50% from… where it was bought"; the borrower's cash flow answers the wrong question ("I don't think it matters… the valuation of the company is half"). Worthington's boundary: absent recession, no widespread default "outside of say software and SaaS." 2026-AUG-08 — App Economy (PRO): the agentic pricing transition is now visible in both directions — HubSpot cut FY26 guidance because free agent trials and outcome-based pricing (qualified leads, resolved tickets) lengthened sales cycles; Figma's first full quarter of AI-credit monetization raised gross margin 2.5pp to 85%; Atlassian's Rovo adopters grow ARR at >2× non-adopters; Klaviyo's autonomous resolutions +80% in two months. Consumption/outcome pricing is displacing seats, with near-term revenue friction. 2026-AUG-06/07/10 — CNBC Halftime, the squeeze goes fundamental: Brown (Aug 6): the July blow-up fund "was not only long memory, it was short software, and those shorts went against them" — IGV had been 37% off its April highs and has round-tripped most of it; "a healthy bull market takes out its own trash." Talkington (Aug 7): the tell is Nadella saying "models need to be swappable" — if models are interchangeable no lab holds pricing power and value accrues to the application layer: "software is continuing to make higher lows… Microsoft is at the epicenter… where you make money as these commodity models get commoditized"; her line: "the market's like peak-OpenAI valuations." Aug 10: IGV +8.5% on the week (Palantir +30%, Atlassian +40%, ServiceNow +9.5%, Snowflake +7.5%), Krinsky "reports of software's death have been greatly exaggerated," Goldman "the beginning of a multi-quarter fundamental inflection point," Amoroso "slowly pivoting to software" (earnings growth ~22%) as the beneficiary when memory bottlenecks resolve — while Terranova recognises the laggard-with-new-momentum pattern and still won't buy: imminent Anthropic/Databricks/OpenAI IPOs mean "significant disintermediation"; he trusts only cyber and Zoom. 2026-AUG-14 (Eisman): "The SaaS apocalypse narrative seems to have taken a bit of a vacation as software stocks have all rallied off the bottom." Winners = AI-value-add newcomers plus incumbents who keep reinvesting; the data arrives with the PE-owned software refinancings landing "sometime next year" — "we need data and that data will emerge soon." AUG-16 (Jay Singh, SSR): a private-equity bid may have set the floor. Silver Lake, with over $100B of tech-focused assets, is reportedly weighing a take-private of Workday — "one of the biggest software LBOs of all time… likely putting a floor on the software sector" — and the stock rose 18% after being −20% on the year against IGV −6% and the S&P +20%. The second, mechanical tailwind is the unwind of the Situational Awareness long-semis/short-software book, which drove Atlassian +35% in a week. Against that, the supply overhang is real: PE "gorged on software companies since 2021… before the advent of ChatGPT when software companies were trading at all-time highs," and software is now "the most troubled part of the pipeline" (Thoma Bravo's Proofpoint amended-and-extended after nearly going bankrupt; Vista's Solera unsold for a decade). 2026-AUG-17 — Joseph Carlson (After Hours): the disintermediation thesis gets its highest-profile execution: Chris Hohn (TCI) exited Microsoft entirely, taking a 17% position to zero on the argument that Claude and AI plugins "could overtake many of the responsibilities and suppress the pricing power of Microsoft Office," with Azure falling behind it as a domino ("a lot of Azure is reliant on Microsoft Office"). The proceeds went to Google (+12%, ~8% combined) — the exact reverse of his 2018 trade. Notably this is now a moat call on a mega-cap incumbent, not just single-feature SaaS. Carlson takes the other side on magnitude while conceding direction: value is "certainly transitioning… to Claude's interface," but MSFT "still remains very wide moat… I become more bullish on Microsoft over time" — and Ackman added another 10%. Scoreboard so far favours the holders: MSFT has recovered into the green YTD, so Hohn sold at the lower valuation. 2026-AUG-17 — Jay Singh (David Lin Report): because AI longs dwarf individual software names, funds hedging them created overwhelming short interest and systematic selling — the Situational Awareness book was long CoreWeave/SK Hynix/SanDisk and short Adobe. He bought the dislocation from March at ~10× cash flow (DDOG, WIX, NOW, SNOW; cybersecurity the biggest overweight) and the forced cover after Citadel absorbed the book launched everything (PLTR +40% on earnings unowned, TEAM high-50s → 165, sold for the gain). His admitted error: far too early — see IGV's −35% / −15% / −25% sequence. 2026-AUG-30 (Jay Singh, SSR call): the reversal is now a fact, not a hope. The crowded trade was long semis / short software, "which is why the SOX index is up 62% year to date, and IGV is still up only 4%" — and since the June 22 momentum unwind "IGV has rallied 25%… while SOX is actually down 22%," with the Morgan Stanley momentum pair −36% and its concentrated TMT sleeve −54% from June 22 to August 28. The bullish twist: "AI agents will access software tools 10 to 100 times more often than humans." Three fundamental exhibits — Atlassian +35% (Aug 6, solid earnings), Workday +18% (Aug 13, on a reported Silver Lake take-private, since "private equity has higher bars to clear given their use of leverage and holding period than the typical investor"), and Salesforce +23% (Aug 26, on results plus an Anthropic alliance letting users "execute actions natively inside Claude without needing to open traditional software screens," and a shift to usage-based from per-user pricing). Plus Okta +19% and CrowdStrike on a record $333M of net new ARR (+51%). The counter-example that keeps it honest: Intuit −3.2% on FY27 EPS guided ~$5 below consensus — "the risk of AI disruption in tax seems to be a bigger issue" — with the general rule to de-risk point-solution vendors lacking embedded workflows, platform scale or a distribution alliance. 2026-AUG-31 — Jeff Keller (Other People's Money): the Q3 software rally (+20% software & services vs −7.5% semis) is substantially technical: SOX and IGV traded at a −1 correlation and ServiceNow/Salesforce popped several percent on bad AI news — the unwind of paired long-AI / short-software books amid ongoing degrossing. He reads the original application-software crash as a rational multiple recalibration (a decade of double-digit growth stepping down to 5–15%, where "it's really hard to dream the dream"), not the market pricing AI destroying the business. Cyber and usage-based infra bounced hardest because short-term certainty trades at a huge premium — but at 25–30× revenue that's "lofty territory… more so than AI, frankly," and the labs themselves intend to disrupt cyber. 2026-FEB-08 → 2026-APR-05 — Pieter Slegers (Compounding Quality), back-fill — the quality-investor's contrarian sort: with the Nasdaq 100 flirting with record highs while software trades at multi-year lows on "AI will replace traditional software," his February Best Buys goes all-software. He concedes AI kills basic task-completion software but rejects it for quality names on three moat tests AI doesn't defeat — proprietary data (ADP pays 1-in-6 US workers; S&P Global's century of data + Platts/CARFAX), high switching costs ("most CFOs won't risk their entire operation to save a few dollars"), and operating leverage from embedding AI in their own products (Firefly — AI credits +3x q/q, Kensho, ADP Assist). The exhibits: Constellation Software — up 20,000%+ since IPO and never previously down more than 25% — is >50% off its high (compounded by Mark Leonard's retirement; successor Mark Miller holds ~$700M and is buying on the open market), FICO fell from >$2,200 to ~$1,400 on the "lenders will vibe-code their own scores" narrative (he calls VantageScore, not AI, the real risk), and MSCI, SPGI, ADBE (buying back ~9%/yr) all sit near decade-low multiples with insiders buying at CSU/MSCI/KKR. By April the basket broadens to the derated toll-bridges and compounders (MELI −35%, KKR ~−50% partly on 7% software exposure, 3i at a NAV discount). 2026-APR-16: the serial-acquirer end of the de-rating quantified — CSU at a 4.8% forward FCFA2S yield and Topicus at 5.1%, both all-time cheap, with 'Mr. Market very depressive about Constellation Software right now — it even caused Mohnish Pabrai, a deep value investor, to buy' it; HgCapital Trust's unlisted European software book grew sales +17%/EBITDA +19% while the trust trades ~30% below a £5.62 NAV. (Slegers 2026-APR-16) Slegers 2026-AUG-23: the whole rated universe is published with forward multiples against each name's own five-year average, and the software/data cohort is the extreme: Adyen 24.9x vs 64.3x (61% under), Adobe 8.5x vs 24.0x, Qualys 12.1x vs 30.1x, Paycom 12.2x vs 43.8x, Paylocity 13.7x vs 42.7x, Constellation 15.8x vs 31.5x, Topicus 25.6x vs 49.2x. 62 companies are undervalued on all three valuation methods simultaneously. The Adyen case is argued in full: its first two acquisitions ever (Talon.One €750m, Orb €335m) are read by the market as a growth signal, against guidance already raised to >20% constant-currency growth and nine years of 35% EPS growth. Slegers 2026-APR-30: buys the de-rating in two places. HgCapital Trust — a private European software book — is added to at a 37.5% discount to a 560p NAV (share price 350p), up from ~30% a fortnight earlier on the same NAV: "I think Mr. Market is being way too negative on software companies like the ones HG Capital Trust invests in." And Topicus at a forward FCF-to-shareholders yield above 5% (FCFA2S +23% in 2025 vs Constellation's +14%), "one of the cheapest valuation levels the company ever traded at." A second, less common tailwind is named: "European sovereignty concerns and stricter data regulation act as moats for local software companies." Pieter Slegers (Compounding Quality), 2026-JAN-22 supplies the mechanism the de-rating debate had been missing, applied to vertical-market software: "AI can make it easier to code, and create software, but AI can't replace the industry-specific expertise and the human relationships that you need to sell and customize niche software." Three properties are named — low churn ("customers rarely switch because the software is deeply integrated into their workflows"), high pricing power ("a lack of viable alternatives and the critical nature of the product") and deep relationships. Evidence of how live the fear is: Constellation Software held a conference call specifically to address AI, is in "its biggest drawdown in its entire history as a public company," and trades at 17.7x forward against a 31.5x five-year average — while COO Mark Miller bought $3.6m of stock in the open market and Topicus spent more on acquisitions in 2025 than in the prior three years combined. The operating results and the multiple are moving in opposite directions. Slegers 2026-JUN-14: the de-rating is being bought by the managers Compounding Quality tracks — "software businesses went on sale in the first quarter of 2026. Why? The market is afraid of AI disruption" — with Mohnish Pabrai named buying Constellation Software in its largest drawdown ever. The mirror trade is the sell side: Fundsmith and AKO Capital exited Intuit entirely and Valley Forge cut ~15%, on a 50-70x multiple plus AI pricing-power risk. By 2026-JUL-09 the screen extreme is Adobe at a forward PE of 8.5 against a 30.9 five-year average (72.5% below), with an Adobe director (Eli Lilly CEO David Ricks) buying 10,000 shares at $194.5. Pieter Slegers (Compounding Quality), 2026-FEB-15 names the trigger and dates it: "Anthropic recently released Claude Cowork. It seems like the market thinks everything could be automated by a Claude plugin" — automating legal, sales and marketing work, "once the bread and butter of SaaS (Software as a Service)." One month's damage across the software/ratings/data complex: FactSet -31%, S&P Global -25%, Moody's -21%, Paycom Software -19%; the iShares Software Sector ETF is down more than 25% in six months (2026-FEB-12). His three sectors "Mr. Market is moody about right now" are software, credit agencies and data providers — "We love these companies as they have high margins and a lot of recurring revenue." The rebuttal is stated as switching cost rather than technology: "It is not the development costs, but mission-critical data, complex regulations, deep integration into work processes and high switching costs that form the moat" (Michael Gielkens, quoted 2026-FEB-01), plus Jensen Huang calling the idea that AI replaces software "illogical" because "companies are building AI that can operate the software we already have" (2026-FEB-12). Constellation Software at 16.2x cash flow against 35x in 2024 and a record 6.8% FCF yield (9.7% expected on 2027) is the archive's cleanest quantification of the de-rating. Pieter Slegers (Compounding Quality), 2026-MAR-15 → 2026-MAR-31: the vertical-market-software complex at record-low multiples on AI fear while the cash keeps compounding. Topicus FY25: revenue +20% to €1,552.3m (4% organic), cash from operations +19% to €412.7m, FCFA2S +23% to €218.7m — against a share price that has returned zero since summer 2021 and "one of its cheapest valuation levels ever." Constellation's cash return on incremental investment is put at 29.1% over three years ("for every $100 you invest, CSU makes $29.1 per year"), and he added $15,000 at CAD 2,600 on 15 March: "I truly think Mr. Market is in a depressive state for Constellation Software today… I don't agree with Mr. Market and the idea the company will be massively disrupted by AI." By 31 March, Constellation, Brown & Brown and Fair Isaac are shown as the three names "trading at one of their lowest valuation levels ever." 2026-SEP-02 (CNBC Halftime — Terranova / Weiss): the cyber complex de-rates on a record print. Palo Alto posts "record earnings, a quarter that they have never had before — it's historic" and falls 10½% (worst day in ~2½ years, still +82% ytd), taking CrowdStrike, Okta, Fortinet and Datadog with it. Terranova: "if you're relying purely on fundamentals, then this quarter would tell you that all the cybersecurity names should move higher… positioning is full and the expectations could not be exceeded no matter how well they actually performed" — and it is "not a one day event." Weiss adds the structural leg: "more announcements from OpenAI, from Alphabet, from Anthropic about their cyber tools… it's smelling a little bit like we saw with software… the competitive landscape's changed here as well," on valuations he says cannot be justified. (2026-SEP-03, Dan Niles, Excess Returns — challenging the de-rating) Three prints challenged the bear case: Atlassian, "supposed to be one of the names getting disrupted," rose 35% the next day; Silver Lake pursuing Workday at roughly $50B, which requires underwriting terminal value 7–10 years out with 30-year yields at 2007 highs ("I thought that marked the bottom in software to some degree"); and Salesforce up over 20% on its Anthropic deal. Relative value now works — software is up about 4% year to date against the semiconductor index up over 60%. His three "safe spaces" are security, systems of record and video games, though "security… has a pretty high valuation because now that's sort of become consensus." The durable change is billing: off the seat licence to "usage based per token as well as outcome-based." Sizing note — OpenAI plus Anthropic annual run-rate revenue went $29B to $105B in seven months, material against roughly $1T of software spend but invisible against $35–50T of knowledge-worker pay. CNBC Halftime — Scott Wapner’s roll call (2026-SEP-04): a third consecutive session of good news being sold across the group — the IGV down 4% on the week “as even some of the names that had pretty good earnings didn’t trade all that well. Zscaler… a beat, guidance was upbeat, stock down. Palo Alto down about 10% this week. It’s not like their report was bad. Fortinet down. Adobe is down” (−6% on naming an insider CEO). Palo Alto’s record quarter on 2026-SEP-02 therefore did not clear in a day, which is the positioning read rather than the fundamental one. Against this theme’s own “disruption fear ran ahead of the evidence” line, this is price evidence pointing the other way while the prints keep beating. 2026-SEP-07 (Jay Singh) — time to fade the squeeze: software has outperformed the Nasdaq five straight weeks — “the counter-trend rally and short covering… the best streak in the post-COVID era” — and is now only 12% from the highs. His conclusion is tactical: “probably time to take some risk off in software.” The mirror: the GS TMT momentum pair “just printed its worst two months on record,” down 50% from the June highs. 2026-SEP-07 (Excess Returns — Matt Zeigler): a bull-side mechanism for the software recovery off the lows: if the later adopters ("the person who's not the early adopter on the crossing-the-chasm math") are only now seeing agentic ROI, "they're not just going to do this once and then say done. There's another investment cycle that now follows with the software delivering of agentic stuff where people can see that ROI." Forehand calls Niles' phase framing "the most compelling bull case framing I've seen." 2026-SEP-07 (RiskReversal — Nathan): names Oracle's print as the week's live test — “probably the clearest connection between what SaaS and where it sits as far as AI build-out,” with Adobe alongside it on the calendar. 2026-SEP-08 (WSJ AI & Business, Asa Fitch): the January “AI coding tools replace corporate software at a fraction of the cost” panic that hit Salesforce and ServiceNow “hasn't happened, and isn't likely to happen” — AI writes code well but maintains and updates it poorly, and continuous maintenance is most of what enterprise software actually is. Two strong earnings reports since the panic have a recovery in software stocks afoot, with room to run as investors accept a multi-year rather than multi-quarter disruption timeline. 2026-SEP-08 (Joseph Carlson): ADBE as the cleanest expression of the de-rating overshooting — "in the center of the AI disruption debate" at a sub-10 forward PE, "basically being treated like a car company, like Ford… getting close to the airline category. And it's a high margin, faster growing software company," on a >10% FCF yield (8.5% after ~19%-of-FCF stock comp) while still growing 12.5% with every segment up. His read is that the fear is "likely fully priced in," and with 14 straight quarters without an EPS miss he predicts another 2–3% beat, where "any glimmer of hope… could cause renewed enthusiasm." App Economy Insights (2026-SEP-12), Adobe Q3: creative freemium MAU past 100M (+70%) and AI-first ARR >$650M (+150%), yet AI-first is only 2% of $27.5B ARR, total ARR still ~11%, and freemium slowed net new ARR and RPO growth (+8%). "The next phase of the story is monetization rather than adoption." Dan Niles (2026-SEP-15): OpenAI + Anthropic run-rate revenue went from $29B to ~$105B this year against ~$1T of software spend — "that spend has got to come from somewhere and some of that has got to be software." Safe buckets: security, systems of record/databases, video games. He passes on Snowflake despite accelerating revenue ("I don't want to fight the hard battles") and declines the host's CRM/NOW/TRI rebound rotation before the midterms. Mike Taylor & Keith McCullough (Hedgeye, 2026-SEP-15) are both long software: select software (ServiceNow, Palantir) becomes the middleman that aggregates compute for tens of thousands of embedded clients — 'free business' as compute pricing falls. Barron's (Chanko, 2026-SEP-02): Amadeus shows the fear ran ahead of the evidence — sold in January's “SaaSpocalypse,” then H1 2026 revenue +5% cc and adjusted EPS +7.3% with every segment up; recovered 23% yet still 16x forward (21x a year ago) with an 8.4% FCF yield. Holders: a zero-error system for perishable inventory “no airline can build… at a sensible price.” Joseph Carlson (Aug-27 / Sep-15, 2026): Salesforce's +22% print (~8–9% organic plus a ~15% debt-funded share shrink) "diminishes" the AI-eats-SaaS scare; most software is a messy both story — consumer tools (Intuit) look dicey, enterprise (ServiceNow, Salesforce) safer — and disruption calls keep missing (NICE: stock destroyed, revenue still growing). 2026-SEP-17 (Pieter Slegers / Compounding Quality, Sep Buy-Hold-Sell List): the 2026 AI-disruption sell-off — ServiceNow −33.3%, Salesforce −37.6%, Workday −39.6%, Accenture −50.4% — has already rebounded +31% to +57% off the lows within ~90 days with no change in earnings: "The fundamentals didn't change. The narrative did." Analyst price targets tracked price both ways; Gartner (11.3x vs a 33x five-year average) is his spotlight for an unwarranted de-rating — the multiple swinging, not the business breaking. 2026-SEP-18 (Carlson): splits software by what an agent removes. Consumer software that sells guided form-filling is most exposed (TurboTax is "just a big form"; LegalZoom templates), while enterprise platforms that hold permission context (who can see which record, who approves spending, how activity is audited) are insulated: Microsoft is "the agentic operating system" and "far more insulated from agentic threat than many other companies" despite lagging as a software stock. Rob Vinall (2026-SEP-19): in March "the question wasn't can I find a software company I want to own, it was which of the 20 or 30 which are down 60, 70%... should I be buying now" — he bought Constellation Software, a company "full of owners" without the egregious stock-based comp common in software. Hedgeye (2026-SEP-18): momentum is a basket, not a sector — software now leads 3-month momentum from "extinction-level" positioning while semis still top 12-minus-1-month lists but are sliding. RPK expects haves and have-nots in both: Snowflake and Twilio ("valuation's not that crazy") in, "Adobe still sucks"; Sam sees a broader software momentum basket in 2027.