AI victims vs winners — value (new) Opportunity ▲ — but the first outright exit says the disruption can arrive as margin concession, which a cheap multiple does not protect against

Jacobs / Hay (Mar 11): AI/LLM fear has indiscriminately crushed a long list of fine, non-AI-victim companies — the opportunity is to triage beaten-down names into AI-victim / AI-neutral / AI-winner and buy the mis-sorted neutrals & winners. Exemplar: sub-$1B auto-insurance lead-gen marketplaces (QNST/MAX/EVER) at ~3.7–6× cash flow that are actually AI beneficiaries (LLM price-comparisons become a new lead source; carriers' state-regulated pricing algorithms can't be disintermediated). Pair with Hay's call to pivot out of US into "the value stocks that are still value" (former cheap names WMT/LLY/CAT/DE now "priced for perfection") and his multi-year-breakout-from-tight-range chart screen (~280 names) for AI-neutral/winner survivors. App Economy (Jun 27): Accenture is the bookings-side casualty — new bookings −2% (first Y/Y decline since Q3 FY25), the stock −18% (worst day ever, ~−50% YTD) on the AI-displacement-of-IT-services thesis, now ~6x EV/FCF with a ~10% combined dividend+buyback yield. Carlson (Jun 23/26): with S&P breadth negative ex-AI/energy he is rotating fresh capital out of the stretched semis into orphaned quality compounders — initiating Uber and DoorDash (funded by his first-ever ASML trim) and defending Netflix and Meta as flow-driven drawdowns (the Netflix/Spotify charts are identical), not fundamental ones. Carlson (Jun 26 portfolio + Jun 29): with the whole ~8-9% YTD index gain coming from just AI + energy, everything else is in multiple compression, not earnings deterioration, so he keeps buying the orphaned quality compounders and holds Meta + Amazon as his favorite 2026 names. Harris (Jun 30): the cheapest AI winners are the opportunity — Meta ~17×, Microsoft ~22-25× (Azure +37%) and Alphabet trade below the market multiple while the "software gets ripped out" disruption fear is overdone given enterprise switching costs; in a churning, high-multiple market he expects a lot more two-way volatility ("a friend if you know the businesses"), and the real secular losers are the packaged-food value traps (Campbell, Conagra, Kraft Heinz) bleeding shelf-space power to healthier brands. Singh (Jul 5): Shopify (SHOP) is the mis-sorted AI winner — dumped with SaaS on an AI-threat narrative but actually gaining share via its Sidekick assistant and AI-driven search (2× the new buyers of legacy search) with ~35% GMV growth; ~60% upside to $190 from $119.50 as it grows into the compressed multiple. App Economy (Jul 3): Meta reportedly entering cloud (renting its excess AI compute, an "AWS move"; stock +10%) turned a big customer into a rival for the neoclouds — CoreWeave −13%, Nebius −15% — sharpening the overspend-vs-new-strategy question. Hay (Jul 6): the June-26 ServiceNow–Accenture joint agentic-AI security offering (NOW's platform + ACN's deployment muscle) publicly rebuts the AI-roadkill bear case on both at once — NOW (~51% off its high, ~20% subscription growth, 32% margins) is becoming the agentic-AI "control tower" as integrators standardize on it (IBM, Cognizant, Google Cloud–HCLTech), and ACN's "prove-it" quarter shows AI as a bookings tailwind (OpenAI's first "AI Transformation Partner of the Year," Anthropic's most-certified partner; ~13% FCF yield at ~12–13× fwd) — own the pair: the enterprise-software and IT-services "victims" are the toll booth and on-ramp of AI, not its roadkill. Carlson (Jul 6): formalizes the trade as his "9 best stocks to buy in July" — because the short-term market "is a reallocation game," non-semi quality is being sold indiscriminately to fund the semi trade, so buy the orphaned compounders (META 18× fwd, AMZN, MSFT at a 5-yr-low ~23× = the 2022 bottom, NFLX −40%, UBER ~$70, DASH, Copart −50%, Constellation Software, Mastercard 23× 2027) sized against each name's own valuation history. Paulo Macro (2025-NOV-24, "The Worm Has Turned", back-fill — the early winners-vs-losers sort): the AI/datacenter narrative decisively turned when Gemini 3 leapfrogged OpenAI without NVDA GPUs — long GOOG against an NVDA funding-short (the revenue-vs-receivables divergence as the tell), ORCL the datacenter-credit posterchild; Jevons' Paradox cuts both ways — cheaper AI can bankrupt the players who levered up to build capacity. Barron's Roundtable (Jul 10): the victims-vs-winners sort at panel scale — Giroux shorts the long-cycle AI derivatives (CAT/GEV/ETN/Dell/SMCI/HPE) while owning the mis-sorted winners (AMD, AVGO, TSM, plus MSFT/AMZN/GOOGL as eventual FCF machines), and the other nine panelists shop the left-behind sectors the AI stampede orphaned — the same buy-the-neglected-quality playbook Carlson and Jacobs run, now the consensus of a ten-investor panel. Carlson (Jul 13): his "motivation framework" sorts the AI model-owners by incentive — Meta/Google/Amazon/Microsoft want the model layer commoditized (they monetize via distribution + cloud) while OpenAI/Anthropic need it premium (the model IS the product); in that tug-of-war "the hyperscalers win" as models grow cheaper and less differentiated — the bull case for Meta's cheap-model push (Musepark 1.1; META +21% in 15 days) and a lean-negative read on the pure-model labs. Aitken (Jul 28): systematically buying the AI-derated "victims" in Canada. Thomson Reuters, down 60% on legal/tax AI-displacement fear, has gone from ~30× EV/EBITDA (where he refused it) to 12–13×: "we're buying Thomson at close to 50 cents on the dollar… outsized returns probably for the next decade," and if wrong the margin of safety still leaves a decent return. The same screen produced Stantec ("we will make a lot of money from here") and adds in TMX Group. His filter: the disruption must show up in reported numbers — "we're ultimately fundamental investors and we like to see it in the numbers" — and so far it hasn't. 2026-AUG-07/08 — App Economy: NYT states the victim mechanism outright — "declining search and referral traffic from big tech platforms is making acquisition harder" (net adds down to 280k, shares −15%); Zillow's uniques and visits both fell 2% while revenue grew 18%, and it now courts ChatGPT and Gemini as acquisition channels rather than treating them purely as threat. Winner-side counterpoint: Shopify's AI-attributed orders convert ~2× when agents read its structured Catalog vs scraped web data, with 75% of AI orders outside the top-100 categories — machine-readable inventory is the winner's asset. Carlson (Aug 10): the AV discount on UBER (24 PE, ~20% revenue growth, ~4% normalized FCF yield) fails a volume check — Waymo's 500,000 rides a week is 17 minutes of Uber volume, "Uber is 600 times bigger… the distance between the two is getting larger, not closer" — and because network density must be re-earned city by city while Uber already owns it globally, "Uber can be late to the game with AVs and still be in the lead," even by four years. Same shape at NFLX (−37% over the past year): the engagement-decline narrative is contradicted by disclosed metrics (season-1→2 drop-off improved, subscribers grew, total engagement +2% YoY) and the per-user decline is a mix effect from expanding into low-TV-consumption markets. 2026-AUG-13 (Polomny, quoting and endorsing Robotti's Q2 letter): The narrative/reality gap: AI's recognized beneficiaries are priced with "little room for disappointment" while "the physical economy that must be built to realize any version of the AI future remains largely ignored" — "one does not need to predict which model wins… to observe that the demand for machines, materials, energy, and infrastructure is rising, straining supply that cannot quickly respond." Polomny: "if you believe in investing in value or sectors that are out of favor, Bob Robotti is your guy." 2026-AUG-13 — Joseph Carlson (Ackman's Q2 letter): the zero test on AI-disruption de-ratings — S&P Global fell >25% peak-to-trough on Anthropic's Claude Co-work launch (25×→19×, a five-year low) and Moody's went with it, yet Capital IQ is <7% of revenue and less of profit, 40%+ of Market Intelligence revenue is proprietary data embedded in customer workflows, and growth is "almost exclusively determined by the benchmark business" — zeroing the entire contested line still leaves a 26 forward P/E. "The concerns of Claude are real, but they are heavily overstated to a dramatic degree." Chris Mayer (Value After Hours, 2026-AUG-25) gives the quality-compounder version of the same read: judge AI disruption case by case, business by business, because the pattern rhymes with the late-1990s internet — "there will definitely be casualties… businesses that are just zeroed out," the way Amazon hollowed out brick-and-mortar retail. But the base case is ubiquity, not advantage: "today nobody would claim they have a competitive advantage because they have a website… I imagine AI will be something like that. It'll be so ubiquitous that everybody will have it… but no one's going to be able to claim that we use AI and that's our competitive advantage." The investable implication is that AI is a cost/threat variable to underwrite name by name, not a moat to pay up for. 2026-APR-16: Slegers' first full portfolio disclosure shows the derating from the inside — of 12 top-conviction holdings, Kinsale is at 'its cheapest valuation level ever,' Constellation and Topicus at 'the cheapest valuation level the company has ever traded at' (4.8% / 5.1% forward FCFA2S yields), Brown & Brown 'the cheapest valuation level of the past 10 years… and it's not even close,' Visa and Zoetis at decade lows. His own aside: 'do you start noticing a trend here? A lot of our companies are trading at very cheap valuation levels.' (Slegers 2026-APR-16) Slegers 2026-JUN-28: "there are a lot of Quality Companies trading at decade low valuation levels today. Think about companies like Mastercard, S&P Global, Novo Nordisk" — and acts on it, adding $50,000 across Visa, Kinsale and Ameriprise at named limit prices. Slegers 2026-JUL-12 puts numbers on the whole de-rated cohort from inside a book: forward PE 17.1x, expected revenue growth +6.8%, ~15% cheaper than the S&P 500, with intrinsic value compounding "nearly 20% per year" — so a 10% annual return needs only owner's earnings growing at the revenue rate plus a re-rating to 20x. His tentative turn signal is Berkshire: behind the index over a year, ahead over the past month — "it looks like we might be starting to see a rotation back into quality." Slegers 2026-AUG-11: in a Compounding Quality interview, growth investor Kris Heyndrikx describes buying Alphabet at the point when “everyone thought AI would kill it” — the value stage of his growth-to-value framing, with the claimed edge being domain knowledge (following AI closely enough to judge the disruption narrative wrong) rather than valuation work. Two days later the house twenty-year list ranks Alphabet #2 on “an irreplaceable data advantage for AI.” The same structure runs through Compounding Quality's Paychex and Constellation Software cases: concede the AI narrative, then test whether the moat actually breaks. Slegers 2026-AUG-23: the cleanest institutional expression of the victims trade yet — Chris Hohn's TCI fully exited Microsoft, its third-largest holding at end-2025, and rotated into aggregates producers Martin Marietta and Vulcan Materials, because "AI could disrupt Office and Azure faster than the market thinks. AI will have a very hard time disrupting a gravel pit." Compounding Quality reports the move approvingly and still rates MSFT a Buy on the same sheet without answering the argument. The wider list makes the de-rating visible: Adobe at 8.5x forward against a 24.0x five-year average, EPAM at 8.8x vs 30.8x, CoStar at 25.2x vs 73.2x, Gartner at 11.4x vs 33.4x — the AI-victim discount is now 60-70% of the historical multiple across software and data services. Slegers 2026-MAY-07: the AI-fear discount is worked case by case rather than as a theme. MSCI — "creating an index is easy, but creating an index that every pension fund and bank in the world trusts and has integrated into their software for 30 years is nearly impossible" (Q1 2026 revenue $850.8m vs $745.8m). Adobe is the honest failure: top of the published list at a 20.8% expected return, 10.8x forward and 58.4% undervalued, "its cheapest valuation level ever" on a decade of 15% revenue growth — and then declined, "this is under the assumption you don't believe Artificial Intelligence will disrupt their business model. For me personally, Adobe is in the 'too hard' pile." The screen and the author's own money now point different ways. Slegers 2026-JUN-07: two opposite AI verdicts on data businesses in one issue. S&P Global (Best Buy #3) sold off on AI-disruption fear that is rejected: proprietary licensed data such as Platts commodity pricing "goes back over a century" and "Large Language Models actually need S&P's private, licensed data to be accurate" — AI as customer, not competitor. Fair Isaac is the mirror case: AI is the reason the moat is at risk, because it lowers the cost of building a rival credit score ("AI could make it easier for VantageScore to become successful", 2026-MAY-21). The distinction — does AI substitute for the product, or lower the barrier to competing with it — is the operative test. Slegers 2026-JUN-14: the archive's clearest statement of the split — Intuit sold by three quality managers because AI threatens the pricing power behind a 50-70x multiple, while Constellation Software is bought because its software "is minimal compared to the operating costs of most businesses" and therefore is not where an AI-driven cost cut starts. By 2026-JUL-09 the presumed victims are priced for decline: Gartner's reverse DCF implies −1.4% growth against 8.0% expected and Adobe's −1.2% against 12.3%. Pieter Slegers (Compounding Quality), 2026-FEB-05 puts a number on the opportunity the AI-disruption fear has created: "Today, 51 companies are undervalued on each valuation method in our watchlist. This number (51) has never been higher." Adobe tops two of the three screens simultaneously — 13.6x forward against a 30.9x five-year average and the highest expected return in the earnings-growth model at 20.9% — with the condition stated plainly: "This under the assumption you don't believe Artificial Intelligence will disrupt their business model." Also caught in the same crossfire and cheap on their own history: Paycom (61.6% under), Paylocity (54.1%), Gartner (47.6%), MarketAxess (46.6%), EPAM (46.4%). The 2026-FEB-24 issue adds the confirming signal — "Two great examples of companies where insiders are now buying shares? KKR and MSCI." Pieter Slegers (Compounding Quality), 2026-MAR-08 → 2026-MAR-19: the AI-disruption derating widening from software into data, payroll, travel and vertical-market software — Constellation Software at 16.6x forward against a 31.5x five-year average, Topicus at 26.4x against 49.2x ("Free Cash Flow keeps going up while the stock went down heavily"), Adobe at 10.8x against 30.9x, Booking at its lowest forward P/E since 2020, and Adyen at its cheapest in a decade. The mirror-image trade is flagged as the real risk: "Fear is driving investors into 'AI-proof' businesses. Walmart and Costco now trade at higher P/E multiples than NVIDIA" — Walmart 46.1x and Costco 53.7x against NVIDIA's 39.9x — with the verdict "overpaying for Walmart or Costco could be a big mistake." The defence he applies to the derated names is switching cost and encoded local knowledge rather than technology: "AI might know the law… but it doesn't know how a small-town court in the Netherlands actually runs day to day." 2026-SEP-08 (WSJ AI & Business, Asa Fitch): Shopify is the cleanest public read on whether all-in AI adoption pays — an internal memo making AI a “baseline expectation,” AI scored in performance reviews, AI-first prototyping, and a hiring gate (prove the job can't be done with AI before you may hire for it) — set against revenue of ~$7bn (2023) rising to a projected $15bn+ this year while headcount fell from ~11,600 (end-2022) to 7,600 (end-2025), plus AI-driven traffic and orders tripling y/y in Q2 (President Harley Finkelstein). The stock is nonetheless −15%+ YTD at ~57× forward earnings — the operating win is not yet the price win. 2026-SEP-08 (David Hay / Haymaker) — the first outright exit from an AI-victim value name. ACN, bought 2026-MAY-08 at ~$178 on the “every $1 of hyperscaler capex needs $1 of integration” case, is sold in full at $178.26 — “one of our rare full sells” — and sold against the valuation, not with it: “notwithstanding the hefty revival, it remains a very cheap stock… but our confidence level in it retaining that hallowed status is eroding.” The disconfirming evidence is pricing power, not demand: an FT report (which he cross-checked) that “defections seem manageable at this point, but… a large number apparently are demanding price concessions to stick with ACN.” The caution generalizes to the theme's shape — if disruption shows up as margin concession rather than revenue loss, a low multiple is not the protection it appears to be. 2026-SEP-08 (Joseph Carlson): the screen stated as a threshold rather than a discount — benchmark the multiple against commodity businesses, not the company's own history: LULU at an 11.9 forward PE on shrinking EPS is "well below commodity multiples, meaning that Lululemon is currently being priced for capital destruction," and ADBE at a sub-10 PE is priced like an automaker. Two second-order tests he attaches: the damage must sort into temporary (management/marketing) versus permanent (structural) — "the biggest factor… is deciding whether or not this problem in the company is temporary or permanent" — and a depressed multiple is itself an asset via the buyback (LULU already retiring >5% of shares a year on very low dilution, "upwards of above 10%" possible with $1.39B of cash). 2026-SEP-10 (JF Tardif, Timelo) — a bought AI-victim with rising estimates: Colliers is down 42% over a year while profits and consensus went up, purely on "the fear of AI" hitting brokerage, property management and engineering — a fear he is "personally skeptical" of, rebutted with the CAD precedent: computers and design software were supposed to end engineering firms and instead "engineering work just continued to grow and grow and grow." What is left is 12x earnings, "one of the cheapest it's ever been," an ~8% free-cash-flow yield reinvested into acquisitions, and a return-to-office recovery as an unpaid-for kicker. 2026-SEP-07 (Ryan Tunis, Cantor): insurance brokers are his top P&C sub-group despite the AI-disintermediation scare — clients treat Marsh/Aon as “their actual insurance company,” carriers gain little going direct at ~30% expense ratios, and a third of claims are initially denied (“An AI bot or Marsh?”). Organic growth has bottomed at 4–5% vs 3% last soft market, at last-cycle multiples. Barron's (Chanko, 2026-SEP-02): Amadeus as the supposed AI victim turned buyer — its €1.2B countercyclical Idemia biometrics deal came “when software companies were considered worthless because of AI” (Baillie Gifford); $94 ADR target, ~40% upside. Chris D'Agnes (Hamlin Capital, 2026-AUG-10): 'fallen angel' dividend growers sold 10-40% while earnings estimates barely move - impatience with anything not AI-related, plus AI-disruption fear in business services (Paychex, FactSet, Thomson Reuters 'hit really hard') - are 'an area of real interest right now' for an equity-income manager. 2026-SEP-17 (Pieter Slegers): a record 57 names in his quality universe now rated Buy ("never been higher"), many presumed AI victims — Gartner, Adobe (8.5x forward), EPAM, Paycom, Paylocity, FactSet, Kainos — 60–70% below their own five-year forward multiples. Price = EPS × multiple with EPS still growing, so the drop is a sentiment de-rating (Rollins 60x → 31x in a year on steady ~12% EPS growth). 2026-SEP-18 (Carlson): re-runs Citrini Research's February 2028 Global Intelligence Crisis against a working agent (Meta's Muse) and sorts his book on two axes: how much value comes from removing tedious steps (vulnerability) versus how much scarce execution remains (resilience). Most exposed: BKNG (sold; marketing ~30% of revenue, so the risk is a gatekeeper swap, not a volume collapse), INTU (sold), LZ, SLQT/GOCO/EVER, Z. Mixed: DASH, UBER, AMZN, DUOL, GOOGL. Resilient: MA, V, ASML, SPGI/MCO, NFLX, COST, TXRH. Beneficiaries: META, MSFT, SHOP. His read: Citrini was right about friction-sellers and wrong about networks and trust layers.