Data owners — company one-pagers

• Derivatives pricing & trade data
• High-margin byproduct of the exchange

• Derivatives data flow grows with record volumes

• Sells valuable data, but it's not the thesis

• Sells market data conventionally

• Derivatives data into quant/agent stacks

• Premium, well-understood
• Owner-ish, but data isn't the re-rate

• Priced, data not the driver

• Volume cyclicality; FMX (BGC) attacking rates franchise

Not a data-AI story

• Volume cycles; data pricing

• Entity-linked financial data: fundamentals, estimates, ownership, transcripts
• 'Symbology' deep ticker-linking is the connective tissue agents need
• But much content is aggregated/licensed, not owned — caps the moat
• Workflow terminals for buy/sell-side

• Coverage expands steadily; much content aggregated, not originated

• Conversational FactSet Mercury shipped; 48/50 top clients on AI tools
• Clean, entity-linked data is ideal RAG fuel for finance copilots
• Up-ish the curve
• Aggregated data limits licensing leverage

• FactSet Mercury + transcript AI; aggregated content limits licensing

• Symbology/entity-linking as agent infrastructure

• ~4.3x EV/Sales on +5% growth — quality at a modest multiple
• Aggregated (non-owned) data caps the moat
• Modest favorable gap on metrics

• Quality franchise at a modest multiple — some re-rate optionality
• Aggregated content caps the upside
• Balanced

• Content-licensing input costs (incl. CUSIP) squeeze margins
• CEO transition; retention metrics softening

De-rated with the info-services group in Feb 2026

• Retention metrics; Mercury adoption

• Dominant US mortgage data (Black Knight/Ellie Mae) — origination/servicing graph
• Pricing & fixed-income reference data
• Hard-to-replicate corpus inside an 'exchange' wrapper

• Mortgage data flows with origination cycle; pricing data steady

• Steadily productizing pricing/reference data
• Mortgage data graph is AI-relevant
• Mid on the curve

• In-product mortgage-AI; data feeds sold conventionally

• Mortgage-graph grounding for housing/credit agents

• A real owner screens miss (files as an exchange)
• Mostly priced

• Quality priced
• Limited asymmetry

• Mortgage tech is deeply cyclical — bought at the top
• Black Knight deal debt still being digested

Read as an exchange, never as a data-AI play

• Mortgage cycle; IMB platform wins

• Credit ratings (MIS) + Moody's Analytics
• Orbis: largest private-company database (~500M entities)
• Default histories + ownership graph — decision-grade
• Essential grounding for credit agents, KYC, supply-chain AI

• Orbis entity graph keeps expanding (~500M+ entities)
• Ratings/transcript flow continuous; issuance cyclical

• Early OpenAI partnership; Research Assistant copilot
• MCP distribution into Claude/ChatGPT/Copilot
• Packaging data for agentic workflows — furthest on distribution
• High — arguably best-executed, hence richly priced

• Early OpenAI partnership; Research Assistant copilot
• MCP distribution into Claude/ChatGPT/Copilot
• No raw licensing — productized access only

• Agentic KYC/credit-memo workflows priced per seat
• Orbis private-company graph as agent grounding

• Best business + furthest-along AI
• ~11x sales / ~40x earnings to match
• Thinnest discount; DCFs flag it rich

• Best business, thinnest discount, DCF flags it rich
• Limited upside → low convexity

• Ratings revenue rides the debt-issuance cycle
• Duopoly position invites periodic antitrust/regulatory attention

Best-executed AI strategy is consensus; it's in the ~11x

• Agentic product attach rates
• Ratings issuance cycle
• Orbis monetization moves

• Fund/ETF data, star & analyst ratings; DBRS credit ratings
• PitchBook private-markets/VC dataset is the scarce crown jewel
• Fund data feeds advisor copilots

• PitchBook's private-company universe compounds with VC/PE activity

• Mo chatbot + PitchBook AI features
• Monetization mostly stays in-product
• Mid on the curve

• Mo assistant; PitchBook AI features; in-product only

• PitchBook private-market data licensing to AI deal tools

• ~$7.0B cap, ~3.4x EV/Sales on +8% growth
• Cheap for a PitchBook-owning franchise
• Market under-paying for the private-markets data

• PitchBook AI-deal-sourcing optionality, cheaply priced
• No hard catalyst
• Cheap enough to tilt positive

• Founder (Mansueto) voting control
• PitchBook decelerated with the VC downturn; DBRS is issuance-cyclical

PitchBook's AI value ~absent from the narrative

• PitchBook growth; advisor-AI launches

• Indices (World, EM) portfolios are built and measured against
• ESG/climate ratings, Barra factor/risk models, Burgiss private-asset data
• Benchmarks + factor models are chokepoints
• Index licensing is a recurring toll-road

• Index/factor data grows with markets; private-asset (Burgiss) expanding fast

• IndexAI connector; 'train clients' LLMs' roadmap
• Solid enterprise APIs
• Less aggressive than S&P/Moody's
• Mid/high — capable but measured

• IndexAI connector; 'train clients' LLMs' roadmap — no licensing $ disclosed

• Benchmark/factor licensing to agent platforms

• ~14x sales — one of the richest here
• Priced as the premium compounder it is
• No discount to the quality

• Richest multiple here
• Least convex — priced for the quality

• Client concentration in fee-pressured asset managers
• US political backlash against ESG products

AI seen as feature, not thesis

• Index flows; ESG/private-asset data attach

• 100+ proprietary market-data feeds
• Index & analytics data products
• A licensing toll-road like S&P benchmarks

• Market data grows with volumes; Verafin fraud signals compound

• Feeds quant/agent workflows
• Productized data
• Up the curve on data productization

• Verafin AI (fraud), market-data feeds; no LLM licensing line

• Surveillance/fraud agents; index licensing

• Premium valuation reflects the toll-road
• Quality owner, little discount

• Priced toll-road
• Limited asymmetry

• Adenza acquisition debt + integration
• Crypto-listings exposure adds volatility

AI in products, not in the multiple

• Fin-crime AI growth; data ARR

• Credit ratings, Capital IQ fundamentals/transcripts, Platts benchmarks
• S&P Dow Jones Indices + Mobility (CARFAX)
• Benchmarks are licensing toll-roads AI can't route around
• The grounding layer any financial LLM/agent needs

• Daily benchmark prints, transcripts, fundamentals — relentless flow
• CARFAX events + Mobility add new streams

• Kensho LLM-ready API live since Nov 2024; 300+ customers
• Anthropic MCP connector + Claude Cowork plugin (Feb 2026)
• Cohere North partnership (Jun 8, 2026) — sovereign/regulated AI
• Distribution into Claude, ChatGPT, Gemini, Copilot
• The most aggressive everywhere-the-agents-are strategy

• Kensho LLM-ready API (Nov 2024), 300+ customers (launch)
• Claude Cowork plugin + Anthropic MCP (Kensho)
• Cohere North partnership, Jun 8 2026 (PR)

• Per-seat / usage pricing for agentic data access
• Benchmark licensing to agent platforms (toll-road extension)
• Private-markets data into AI workflows

• ~8.8x EV/Sales on +8% growth — premium largely intact
• Top-tier AI execution already recognized in the multiple
• Quality fully priced; no metric discrepancy

• Quality + best-in-class AI execution already in the multiple
• Limited discrepancy on metrics
• Modest two-way payoff

• IHS Markit integration legacy; Mobility (CARFAX) is auto-cyclical
• Index fee compression a slow structural drag

AI execution is consensus among analysts; the multiple carries only a modest sector AI-threat discount

• AI-access revenue disclosure (none yet)
• More agent-platform embeds
• Ratings cycle + index flows

• 45+ yrs of proprietary syndicated IT/business research from ~2,000 analysts
• Magic Quadrants & Hype Cycles are de-facto standards CIOs buy on
• Price, salary & contract benchmarks from thousands of engagements
• Behind a hard paywall — not on the open web, not freely scrapeable
• >75% of contract value multi-year recurring, embedded in workflows

• Analyst output paced by headcount; inquiry/benchmark data grows with clients
• CV slowdown = the inflow risk: fewer clients → less peer data

• Two-sided: AI could commoditize 'advice' or make its data the grounding layer
• Rolling out AskGartner inside client licenses
• Has NOT licensed its corpus to labs — keeps it walled
• Contract-value growth slowed to ~1–5% — the market's disruption tell
• Early on the curve; data-as-grounding thesis unproven

• None — AskGartner ships inside existing client licenses
• AskGartner live across research portal (example)

• Corpus-grounded agent for enterprises (license upsell)
• Selective API access to benchmarks/peer data
• Price/SLA tiers for AI-assisted research

• ~2.1x EV/Sales for a 77%-gross-margin, mostly-recurring franchise
• The multiple embeds a full AI-disruption outcome; CV growth ~1–5% is the operational tell
• Cheapest quality owner on the board on metrics

• Profitable recurring base at ~2x sales bounds the downside
• Large upside if AI proves additive to the franchise
• Cheap quality + two-sided AI = positive convexity

• Conference/consulting segments are macro-cyclical
• EPS growth leans on buybacks; sales-force productivity in question

Narrative casts Gartner as an AI casualty; AskGartner and the paywalled corpus get little credit

• Contract-value growth stabilization (the single tell)
• AskGartner engagement disclosures
• Buyback pace

• Westlaw: case law, statutes, annotations built over a century
• Editorial headnotes/KeyCite are irreplicable human layers
• Practical Law, Checkpoint (tax), Reuters News
• Legal/tax = highest-value, lowest-hallucination-tolerance use cases

• Case law grows with the courts — slow, perpetual accretion
• Editorial annotations (headnotes/KeyCite) compound on top

• CoCounsel scaling fast — ~1M AI users
• AI-native Westlaw does grounded retrieval over its corpus
• Monetizes the data itself
• High — clear legal-AI leader

• No corpus licensing — deliberate walled strategy
• CoCounsel: 1M professionals, 107 countries (Feb 2026) (PR)
• Building proprietary LLM for regulated use cases

• Selective agent-platform access to Westlaw (MCP-style)
• CoCounsel 10x user target = the in-product unlock
• Tax/audit agentic suites later in 2026

• ~9x EV/Sales on +7% growth — a modest AI-threat discount against its quality
• CoCounsel at 1M users is distribution the multiple under-credits
• Premium franchise; the discount is partial, not deep

• Priced quality; AI leadership reflected
• Limited convexity

• Woodbridge (Thomson family) controls ~70% — governance is theirs
• Print/legacy declines largely done; tax season concentration

Market narrative treats agentic AI as a threat to legal-research seats; CoCounsel distribution under-credited

• CoCounsel next-gen GA + adoption metrics
• ACV growth reacceleration (the proof point)
• Competitive data vs Harvey/Legora/Claude Cowork

• The Work Number — unique employer-sourced income/employment records
• Verified income/employment ground-truth no LLM can infer
• Utility/telecom payment data extends the picture
• Gating data for lending, hiring, benefits
• Contributory — employers feed it (network effects)

• The Work Number records keep growing via payroll integrations
• Every paycheck is a new record — true flow asset

• EFX.AI built into new product models
• FCRA permissible-purpose rules cap AI exposure
• Monetization stays inside regulated rails
• Mid — gated by regulation, not capability
• Re-rate is cyclical more than AI-driven

• EFX.AI in-product; FCRA limits external exposure

• Verified-income rails for lending/hiring agents (permissioned)

• ~$20B cap, ~4.0x EV/Sales on +7% growth
• Cheap for the owner of The Work Number
• Re-rates on the lending/hiring cycle + verified-income AI demand

• Unique income/employment data at a low multiple
• FCRA caps direct licensing, but the asset is irreplaceable
• Cheap + cyclical-recovery optionality = convex

• 2017 breach legacy = elevated security/regulatory burden
• Mortgage + hiring volumes are the real earnings driver near-term
• CFPB / FCRA scrutiny is permanent

AI angle absent; mortgage cycle dominates the narrative

• Mortgage/hiring recovery; TWN records growth; any agent-rail pilots

• Third global credit bureau + marketing/identity/fraud data
• Best organic growth of the three bureaus
• Verified credit/identity data with network effects

• Same bureau flow; strongest organic data investment of the three

• AI products across credit & fraud
• FCRA-style rules cap ecosystem exposure
• Mid on the curve

• Ascend platform AI; in-product

• Same permissioned-rails option as EFX/TRU

• Reasonable bureau multiple
• Only friction is access (London listing)
• Quality peer to EFX/TRU

• Quality + reasonable price
• Regulation caps the convex upside
• Balanced

• UK listing discount; Brazil FX exposure

UK listing keeps it out of the AI conversation

• Cycle; NA mortgage volumes

• The FICO score — decisioning standard embedded in US credit
• More algorithm/standard than raw corpus
• But the score is a data product with monopoly economics

• Scores recompute on bureau flow; FICO originates little raw data

• Own FFM foundation model
• AI lending agents still need an accepted standard
• Mortgage-pricing change is a catalyst
• Not a corpus play

• FICO Foundation Model (FFM) announced; platform AI

• Score-as-API inside lending agents

• ~14x EV/Sales on +15% growth — still premium on metrics
• Moat contested (VantageScore push, AI underwriting)
• Two-sided

• De-rated standard with mortgage-pricing optionality
• But expensive on sales (~14x)
• Two-sided

• Pricing-power backlash: FHFA pushing VantageScore competition in mortgages
• Revenue concentrated in B2B scores; software segment unloved

Debate is pricing power, not AI

• Mortgage-score pricing; platform ARR

• Identity graph & data-collaboration network (25k+ publishers)
• Clean-room identity for the post-cookie/AI-data era

• Identity graph is refresh-maintenance, not accumulation

• Well-placed for AI-data era
• But the story is now M&A

• Identity/clean-room infra relevant to AI data flows

• —

• Being acquired ~$2.5B by Publicis
• Off the board as a standalone bet
• Signal: ad-holdcos paying up for identity data

• Taken out — payoff capped by the deal price

• Deal-close risk is the only variable left (~$38.50 cash)

Story is now the Publicis acquisition

• Deal close (~$38.50)

• Credit bureau + identity resolution (Neustar)
• Links offline identity to digital identifiers
• Identity graphs matter more as AI agents transact
• Contributory bureau data with network effects

• Credit + identity events flow with economic activity

• OneTru platform, TruIQ agents
• Identity products quietly AI-relevant
• FCRA-capped exposure like Equifax
• Mid on the curve

• OneTru platform, TruIQ agents; in-product

• Identity verification for AI-agent transactions

• Cheapest of the three bureaus
• Modest favorable gap
• Same regulatory ceiling

• Cheapest bureau + cycle/identity optionality
• FCRA caps the convex upside
• Balanced

• Neustar deal leverage; UK consumer business weak
• Same CFPB overhang

Same as EFX — cycle story, not AI story

• Cycle turn; Neustar identity products

• Decades of contributory claims, loss & property/peril data
• Nearly all US P&C insurers both feed and buy it back
• Catastrophe models built on the loss history
• Near-monopoly; no AI lab can rebuild it

• Contributory model: every insurer claim feeds it, by contract
• Cat-event data grows with each season

• Generative/agentic AI in underwriting/claims products
• Consortium-locked — not licensed to the open ecosystem
• Value unlock in-product, not via licensing
• Mid — deepest moat, deliberately walled

• Consortium AI in underwriting/claims products

• Walled option: claims-history grounding for insurance agents

• Deep moat, but ~12x sales / ~7% growth
• Fully paid for
• Quality High, value Low-ish

• Near-monopoly data, but premium & walled
• Bounded downside, limited upside
• Low asymmetry

• Consortium members push back on pricing; class actions over contributory data use

Quality priced; AI not separately valued

• Product attach; pricing renewals

• Pharmacy/dispensing & distribution data
• Optimizes thin-margin logistics

• Distribution data tracks volumes

• Logistics input, not sold

• Logistics AI internal

• —

• Fair defensive distributor
• Data-rich, not a data owner

• Defensive, data not a driver

• Drug-pricing policy; thin-margin model

Not an AI story

• Distribution volumes

• Healthcare commercial intel: providers, claims, affiliations, install-base
• 'The ZoomInfo of healthcare' — sells intelligence to life-sciences/med-tech
• A pure data owner, not a marketplace
• Continuously refreshed healthcare-entity graph

• Refresh continues but shrinking revenue funds less data collection

• Real owner, but AI is as much threat as tailwind
• Limited AI productization so far
• Mid/behind — business being repriced
• Erosion risk from AI-generated provider signal

• None disclosed

• Healthcare-commercial grounding data for pharma AI

• ~$0.1B cap, ~2x EV/Sales on declining revenue
• Distressed micro-cap; the data is better than the equity
• Cheap for existential reasons

• Distressed micro-cap → option on stabilization or M&A
• Declining revenue is the live left tail
• Cheap healthcare-commercial data if it survives

• PE overhang (Advent), serial goodwill writedowns, micro-cap liquidity

Micro-cap; no AI narrative attaches

• Revenue stabilization; strategic review odds

• Verified network of most US physicians
• The asset is the audience/engagement, not a corpus
• Workflow tools for doctors

• Most US physicians already on it — the graph is mature
• Engagement/newsfeed data still grows; the asset is breadth, not flow

• Strong AI tools (Doximity GPT), huge engagement
• But no AI revenue in guidance
• Data asset is the audience, not a corpus

• Doximity GPT free for physicians; ad AI in-product

• Clinician-verified channel for healthcare AI distribution

• ~$3.8B cap, ~5.6x EV/Sales on +13% growth
• Far from the ~18x I'd assumed — reasonable now
• Verified clinician graph; audience-not-corpus caps licensing

• Verified clinician graph + AI tools, now at a fair multiple
• Audience-not-corpus caps the data-licensing upside
• Balanced after the de-rate

• Pharma ad-budget concentration; engagement metrics are the whole story

Was priced for AI hopes; now reset to fair

• Ad market; AI tool engagement

• Claims/care-management data via Carelon
• Latent separable data asset
• Used to lower its own medical costs

• Claims flow with membership; flat membership = flat flow

• AI care-management lowers internal costs
• Closest operator to a separable data asset
• Still not pure-play

• Carelon internal AI

• Separable claims-data asset (never signaled)

• Cheap, but on insurer fundamentals
• Latent data optionality (Carelon)
• Cyclical

• De-rated insurer with latent data optionality
• Cyclical, not a data re-rate
• Mildly positive

• Medical-cost trend + Medicaid redeterminations; ACA subsidy politics

Insurer story

• Medical-cost trend; Carelon growth

• Rx-pricing & consumer prescription-behavior data
• Unique data, but an input to a discount platform
• Platform under structural pressure

• Pricing data flows; nothing accumulating in value

• Data feeds the platform; not licensed as a corpus
• Limited AI productization

• None disclosed

• Rx-pricing data into consumer-health agents

• Cheap, but pressured core
• Marginal owner with hard-to-monetize data

• Cheap with stabilization optionality
• But structural pressure on the core
• Binary-ish

• PBM dependence — a single partner change (Kroger '22) cratered it once

No AI narrative

• Platform stabilization

• Liquid-biopsy genomic + clinical-outcomes data in oncology
• Proprietary, scarce — a direct Tempus peer
• Longitudinal molecular profiles track tumor evolution
• Cannot be assembled from public sources

• Test volumes +25–35%/yr; each test extends longitudinal profiles

• Pharma data partnerships + co-development, earlier-stage
• Smart Platform multiomic insights
• Building the 'co-develop on our data' motion
• Mid — monetization layer still forming

• Pharma data partnerships (earlier-stage than Tempus); Smart Platform

• Tempus-style co-builds on liquid-biopsy data

• Scarce data, but ~12x sales and unprofitable
• Analyst upside exists
• Expensive growth, not cheap

• Scarce-data optionality, but ~12x sales + unprofitable cap it
• More a growth bet than an option
• Balanced, positive tilt

• Cash burn continues; patent litigation history with Natera
• Screening (Shield) economics still unproven at scale

Priced as diagnostics growth; data angle secondary

• MRD reimbursement; pharma deal announcements

• World's largest pharmacy-claims & prescription dataset (ex-IMS Health)
• Population-scale real-world evidence across global Rx
• Clinical-trial operational data as the largest CRO (ex-Quintiles)
• De-identified, compliance-grade — built under HIPAA/GDPR, unscrapeable
• Sold to virtually every major pharma

• Rx/claims flow is continuous and population-scale
• Trial operational data compounds with every study run

• IQVIA.ai unified agentic platform (Mar 2026): 150+ agents deployed
• NVIDIA partnership since Jan 2025 — custom foundation models on its data
• 19 of top 20 pharma already using IQVIA agents; 100+ AI patents
• Builds agents ON the data rather than licensing it out
• No longer latent — monetization architecture is live

• NVIDIA partnership (Jan 2025) → IQVIA.ai platform, Mar 2026 (PR)
• 150+ agents live; 19 of top-20 pharma using them (report)
• 100+ AI patents; agents built ON proprietary data, not licensed out

• Agent subscriptions as a separate revenue line
• RWE feeds for medical LLMs (compliance-wrapped)
• Trial-design agents priced on outcomes

• $16.3B FY25 rev, +5.9% (~7% TTM)
• Low-single-digit sales multiple for unique data
• ~$13B net debt is the caveat

• ~2.7x EV/Sales on +6% growth — cheap for the scarcest Rx data
• Locked in compliance contracts; low AI surface today
• Cheap + latent-unlock optionality = convex

• ~$13B net debt limits flexibility
• CRO bookings cyclical; pharma R&D budgets squeezed (IRA effects)

Cheapest scarce-data name; IQVIA.ai barely registers in the multiple yet

• Next earnings: ~late July 2026 (Q1 reported May 5 — beat; EPS guide raised)
• IQVIA.ai adoption: now 192 agents / 64 use cases; watch for monetization disclosure (Q1 call)
• R&DS backlog $32.7B (+5.3%); Q4 book-to-bill 1.18x — bookings reacceleration is the proof point
• $1.2B buyback remaining ($552M done in Q1)
• Duke obesity-trials collaboration (Feb 2026) — fastest-growing trial category
• De-leveraging from 3.62x / $13.9B net debt frees the multiple

• Genetic-testing / cfDNA data (MRD, prenatal, transplant)
• Large, fast-growing proprietary genomic dataset
• Outcome-linked longitudinal data is the durable asset
• Same scarce-data position as Guardant/Tempus

• Fastest test-volume growth in the group; outcome links accrue with time

• Owns the data; data-layer monetization still maturing
• Strong clinical-validation pipeline feeds the dataset
• Files as diagnostics, so screens miss it
• Mid on the curve

• Data feeds pharma trials; in-product AI

• Outcome-linked genomic licensing

• Irreplaceable data, ~12x sales on ~30% growth
• Quality High; multiple says priced, not discounted
• Volatile equity

• Data optionality vs a rich ~12x multiple
• Roughly balanced, slight positive tilt

• Reimbursement concentration (Medicare MRD decisions)
• Billing-practice scrutiny; GH litigation

Same — growth story, data unpriced

• MRD adoption; new indications

• Multimodal clinical + genomic data (~500-PB) pairing sequencing with clinical records
• Scarcest, most valuable category for biomedical AI — unscrapeable
• Built explicitly as an AI data company
• 140% net revenue retention on Insights/data
• Linked outcomes data is what makes it irreplaceable

• ~300PB and growing; every test adds linked clinical+genomic data (Q1 letter)
• Sequencing volumes growing ~25–30%/yr — the corpus is the byproduct of revenue

• $200M AstraZeneca/Pathos deal (Apr 2025): largest oncology foundation model
• Total remaining contract value >$1B; non-exclusive — can resell the motion
• Data customers: AZ, Novartis, Merck KGaA, Takeda, Boehringer, United Therap.
• Illumina collaboration trains genomic algorithms on its multimodal data
• Insights (data licensing) growing ~58%

• $200M AstraZeneca/Pathos data+model deal over 3 yrs (PR)
• Total remaining contract value >$1B (Q1 letter)
• Data customers: Novartis, Merck KGaA, Takeda, Boehringer, United Therap.
• Illumina algorithm-training collaboration

• Non-exclusive foundation-model co-builds with other pharma
• Expansion beyond oncology (cardio, neuro)
• Open-source pathology consortium as a funnel

• ~$8.5B cap, ~6.5x EV/Sales on +83% growth
• Strikingly cheap for the growth + scarcest biomedical data
• Priced like a normal growth co, not the data monopoly it's building

• Foundation-model + licensing optionality could make it the oncology-AI data layer
• Rich multiple + cash burn are the downside
• Large, real optionality = convex growth bet

• Founder super-voting control; Pathos is Lefkofsky-affiliated (related-party optics on the $200M deal)
• Convertible debt + only just adj-EBITDA positive
• Short-seller scrutiny history (data-quality claims)

AI is in the name and the multiple — but >$1B RCV arguably still under-modeled

• Next earnings: ~early Aug 2026 (Q1 reported May 5 — guidance raised) (Q1 8-K)
• 2026 guide raised to $1.59–1.60B revenue / ~$65M adj EBITDA — the leverage inflection
• MRD volume ~6,500 tests in Q1, +500% YoY — reimbursement decisions are the swing
• TCV >$1.1B; 70+ pharma data customers — watch new (non-exclusive) co-builds
• Insights (data licensing) +44% in Q1 — the annuity compounding

• Life-sciences CRM + proprietary OpenData/Link (HCP & reference data)
• Pharma depends on its reference data
• A separable corpus inside the SaaS

• OpenData/Link refreshed continuously; usage data grows with seats

• AI embedded in pharma workflows
• Up the curve
• Vertical-SaaS leader

• AI agents shipping across CRM/Vault (Dec 2025 wave)
• OpenData/Link reference data feeds its own AI

• Agent pricing on top of seats
• Link data into pharma AI pipelines

• Premium SaaS multiple
• Data is a real, under-discussed asset
• Equity priced for quality

• Premium SaaS; data underrated but equity priced
• Limited asymmetry

• Salesforce→own-platform CRM migration is a multi-year execution risk
• Core TAM maturing; growth depends on new apps

Read as a quality SaaS with AI features, not a data owner

• Agent adoption metrics
• Vault CRM migration completion

• Transactional used-car e-commerce & trade data
• Tunes its own pricing/inventory

• Transaction/pricing data grows with units; internal

• Input, not the product

• Internal pricing AI

• —

• Volatile, richly valued
• Weak fit for the screen

• High beta but valued on retail growth, not data

• Garcia family control + related-party history; leverage rebuilt the equity once already

Retail story

• Unit economics

• Verified CRE comps/property data, 35-yr research army
• LoopNet, Apartments.com, Homes.com
• Unscrapeable, walled inside terminals

• Research army keeps verifying; comps accumulate permanently
• Zonda adds a housing-data stream

• Walled, litigious; data locked in terminals — minimal AI surface
• Heavy Homes.com ad spend
• Strategic data, low AI surface area

• None — deliberately walled; litigious vs scrapers
• Zonda acquisition ($800M) extends housing data

• The big withheld option: licensed CRE grounding for real-estate AI
• Homes.com AI search features

• ~$14B cap, ~4.0x EV/Sales on +19% growth
• Much cheaper than I'd shown; Homes.com spend masks margins
• Unscrapeable CRE data at a reasonable price

• Unscrapeable CRE data, now cheap
• Low AI surface + heavy ad spend cap near-term
• Re-rate optionality as Homes.com spend rolls off

• Homes.com spend is an act of will (founder-CEO); activist pressure has surfaced
• Serial litigation posture cuts both ways

AI never part of the story; the data optionality is free at ~4x

• Homes.com spend roll-off (margin catalyst)
• Zonda integration
• Any posture change on data access

• One of the largest learning-interaction datasets (50M+ DAU)
• Granular data on how people learn, err & retain across 100+ courses
• Used in-product to tune pedagogy — not licensed
• Value captured as engagement, not a sellable corpus

• Learning interactions scale with DAUs (50M+, growing)
• Every exercise answered is new pedagogy data

• AI-first (Gen-AI 'Max', AI video calls)
• Shipped 148 courses in a year via generative AI
• Unlock shows up as engagement/ARPU, not a licensing line
• Mid — AI deepens the product moat

• None out; heavy OpenAI/GenAI consumer (Max, AI courses)
• 148 AI-generated courses shipped in a year

• Learning-data licensing (never signaled)
• AI-tutor pricing tiers

• ~4.0x EV/Sales on +39% growth — cheap on growth metrics
• AI-disruption fear embedded in the multiple
• A growth franchise at a non-growth multiple

• ~4x on +39% growth bounds the downside if growth holds
• Upside if AI features lift engagement/ARPU
• Positive convexity on metrics

• Founder control; monetization-vs-engagement tension
• Still expensive on earnings even after the crash

Narrative says ChatGPT kills language learning; the AI-first operating model is ignored

• DAU/booking growth stabilization
• Max attach rate
• Energy/engagement metrics

• LatAm marketplace purchase + fintech/credit data
• Powers its own ads/lending (input)

• Purchase + credit data compounds with GMV growth

• AI for marketplace/credit optimization
• Not a sold corpus

• Internal AI for ads/credit

• —

• Premium growth stock
• Data doesn't re-rate it

• High growth, but valued on the business, not the data

• LatAm FX/political risk; credit-book quality through cycles

Not a data play

• LatAm growth; fintech credit

• Viewing/interaction data across ~300M members
• Real moat for recs/greenlighting
• Strictly internal — never licensed

• Viewing data grows with engagement; internal-only

• Never licensed; AI = better curation only
• Internal-use data

• Internal only — never licensed

• —

• Premium mega-cap on subscriber economics
• n/a as a data play

• Priced mega-cap, data internal

• Content-spend discipline vs growth; live/sports costs

Recs AI assumed, not valued separately

• Sub growth; ads tier

• ~100k+ communities, two decades of upvote-ranked human conversation
• Largest archive of authentic opinion, troubleshooting, niche expertise
• Exactly what LLMs lack: recommendations, lived experience, long-tail Q&A
• Surfaces disproportionately in AI answers
• Classified as social media, not 'data services'

• DAU still growing; posts/comments compound the archive daily
• Two decades of vote-ranked history can't be replicated retroactively

• $203M aggregate contract value disclosed at IPO (Google + OpenAI)
• ~$130M/yr run-rate ≈ 10% of revenue; Google ~$60M/yr, OpenAI ~$70M/yr
• #1 most-cited source across AI models (~3x Wikipedia)
• Google renewal under negotiation — pushing usage-based pricing
• Litigates unlicensed scrapers (incl. Perplexity suit)

• $203M aggregate disclosed at IPO (TechCrunch)
• Google ~$60M/yr; OpenAI ~$70M/yr ≈ 10% of revenue (SEL)
• 2–3 yr terms struck Jan 2024 — now in renewal window

• Google renewal at usage-based rates (mgmt: 'open for business')
• Anthropic / Meta / xAI remain unlicensed
• Dynamic per-citation pricing models
• Int'l + vertical (commerce intent) licensing

• Best corpus + fastest unlock, but ~13x sales
• ~65% growth supports it — priced FOR growth
• Quality off the charts; valuation not a gap

• Big growth/licensing optionality = upside call
• But ~13x sales means real drawdown if growth slows
• Net mildly positive from the licensing option

• Community/moderator revolt risk is structural (2023 API blackout precedent)
• Altman's stake = governance optics
• Ad business still ~90% of revenue and competitive

The AI-data story IS the stock; renewal terms are the swing

• Google contract renewal & structure (report)
• Scraper litigation incl. Perplexity suit
• Meta forums app traction (the bear case)
• Data-licensing line in quarterly prints

• ~1B travel reviews; Viator experiences marketplace
• Widely scraped & substitutable
• Reviews feed AI trip-planning agents

• ~1B cumulative, but contributions follow visits — and AI answers divert visits
• The corpus ages if the flywheel slows

• Perplexity partnership (Jan 2025) now a measurable booking channel
• ChatGPT app launch partner (Oct 2025) for trip planning
• Distribution-into-AI strategy, not paid corpus licensing
• Viator + TheFork now >50% of revenue — the real value

• Perplexity partnership, Jan 2025 — hotels customer-acquisition channel (PR)
• ChatGPT app launch partner, Oct 2025 (report)

• Paid licensing of the review corpus (currently given for distribution)
• Viator inventory as the bookable layer inside AI agents

• ~$1.4B cap, ~0.7x EV/Sales on +3% growth
• Very cheap, but reviews are being disintermediated
• Value is Viator/TheFork, not the review corpus

• Cheap but melting
• Weak optionality

• Viator faces GetYourGuide/Klook competition; legacy meta declines
• Post-Liberty structure leaves strategic questions

AI read as existential threat; partnerships seen as defensive, not monetizing

• AI-channel booking disclosures
• Membership program launch
• Viator/TheFork growth (the real value)

• ~300M geocoded local-business reviews
• Structured local sentiment for 'best X near me'
• Classified as internet content, not data services

• 22M new reviews in 2025 (vs 21M in '24); corpus 330M, +7% YoY (FY25 PR)
• What's melting is consumption (app engagement), not contribution — yet
• Risk: contribution follows traffic with a lag

• Signed OpenAI agreement (disclosed Feb 2026)
• Perplexity has used Yelp local data since Mar 2024
• 'Other revenue' +17% on data licensing & transactions
• Expanding Yelp Assistant; Hatch acquisition (AI front-desk)
• Core local-ad business still the eroding center

• OpenAI agreement signed (Feb 2026, undisclosed) (FY25 PR)
• Perplexity has integrated Yelp local data since Mar 2024
• Data licensing inside 'Other revenue' (+17%)

• More assistant integrations (Gemini, Claude, Alexa-class)
• Usage-priced local-data API
• Transactional referrals from AI answers

• ~$1.3B cap, ~0.9x EV/Sales on +3% growth
• <1x sales — cheap, but the ad core is eroding
• AI-distribution optionality vs value-trap risk

• Cheap with AI-distribution optionality
• vs an eroding core
• Binary-ish

• Own antitrust fight with Google (plaintiff) — outcome cuts both ways
• SMB advertiser churn; restaurant/retail ads already shrinking

OpenAI deal is new and barely in the price; story still read as 'Google victim'

• OpenAI deal revenue contribution
• 'Other revenue' growth each quarter
• Services-ads resilience vs AI search

• Zestimate + listing data + largest US housing audience
• Much listing data is MLS-shared, not fully proprietary
• Consumer housing intent data

• Listings turn over rather than accumulate; Zestimate history compounds quietly

• Strong in-app AI; partial data moat
• Real-estate AI agents could use it
• Mid on the curve

• In-app AI (natural-language search); MLS data shared

• Housing-intent data for real-estate agents

• ~$8.6B cap, ~3.1x EV/Sales on +16% growth
• Cheaper than I'd shown; partial (MLS-shared) moat
• Housing-cycle leverage on top

• Housing-cycle optionality
• Partial moat caps the data upside
• Balanced

• NAR commission-settlement reshapes agent economics — its customers' wallets
• Housing-cycle beta; Showcase/mortgage execution

AI features noted, not a data thesis

• Housing cycle; Showcase attach

• Elsevier science (The Lancet, Cell, Scopus) — peer-reviewed at scale
• LexisNexis legal + LexisNexis Risk Solutions (identity/fraud)
• Three of the most defensible corpora on earth in one company
• Scientific literature is critical for frontier capability

• Global science output grows mid-single-digit %/yr; submissions rising
• Caveat: AI-generated paper flood is a quality-control burden

• Lexis+AI, Scopus AI, ClinicalKey AI, Protégé all live
• Embeds data in grounded retrieval vs raw training access
• Among the best-positioned grounded-AI owners
• High — productization mature & shipping

• No raw licensing; grounded products only
• Lexis+ AI, Scopus AI, ClinicalKey AI, Protégé all shipped

• Elsevier corpus licensing remains a withheld option (big if ever)
• Agent-access tiers to Scopus/Lexis
• Risk-data feeds into KYC agents

• ~9x EV/Sales on +7% growth — a small AI-threat discount embedded
• Grounded AI products shipping across all three corpora
• Durable compounder; thesis is durability, not deep value

• Fully-valued premium compounder
• Durable, but limited asymmetry either way

• Open-access mandates (Plan S) pressure Elsevier's model
• Exhibitions segment is cyclical

Same legal-AI threat narrative as TRI; grounded-product execution under-credited

• Lexis+ AI penetration disclosures
• Any Elsevier AI-licensing posture change
• FY guide post-crash

• Peer-reviewed STM journals/books; Cochrane co-publishing
• Vetted scientific text — what labs pay for to lift capability
• A 'smaller Elsevier' — quality corpus, narrower than RELX
• Editorial vetting + citation links add provenance
• Proprietary, not freely on the open web

• Submissions +25%, output +13% — the journal flow is accelerating (Q1 PR)
• Caveat: some of that surge is AI-assisted writing — vetting is the product

• $92M lifetime AI-licensing revenue; $29M in Q1 FY26 alone
• Anthropic strategic partnership (Sep 2025) + projects with 3 top tech cos
• Recurring inference pilots: pharma, chemical, space-exploration cos
• One of the only names with disclosed, recurring AI revenue
• Recurring AI line gives it proven monetization few peers can show

• $92M lifetime AI revenue; $29M in Q1 FY26 (PR)
• Anthropic strategic partnership (Sep 2025)
• Projects with 3 of the largest tech cos (unnamed)
• Recurring inference pilots: pharma, chemical, space

• Convert pilots → recurring corporate R&D subscriptions
• License on behalf of partner publishers (agency model)
• Agent-citation / RAG licensing beyond training

• ~1.9x EV/Sales with disclosed, recurring AI-licensing revenue ($92M lifetime)
• Flat underlying top line is the offset
• Cheap on metrics for the rare proven AI licensor

• Low multiple + proven licensing = bounded downside with optional upside
• Flat core growth caps the slope
• Asymmetry modest but positive

• Library budget pressure + consolidation of academic spend
• Post-divestiture portfolio still re-finding growth

AI-licensing story is prominent in coverage; expectations now elevated

• Next earnings: Tue June 16, 2026, pre-market — FY26 Q4 + FY27 guide (notice)
• AI recurring revenue <10% of AI revenue today; mgmt expects the proportion to triple next year (Q3 call)
• OpenEvidence partnership: 5-yr multimillion licensing + Wiley equity stake
• Nexus licensing service at 36 publishing partners — the agency model scaling
• Emerald Publishing acquisition (Jun 2, 2026) adds proprietary research corpus
• Q3 raised margin/EPS guidance to high end; ~4.5% dividend while you wait

• Legal, tax, health & regulatory information + workflow (CCH, UpToDate)
• UpToDate is a premier point-of-care clinical reference
• Authoritative corpora like RELX/Thomson Reuters
• Subscription, deeply embedded in workflows

• Regulatory/tax/clinical updates are a built-in perpetual flow

• AI workflow tools shipping across segments
• Same grounded-AI position as RELX/TRI
• Up the curve, productizing its corpus

• AI embedded in UpToDate/CCH; no corpus licensing

• Clinical-grounding deals for medical AI (UpToDate is the prize)

• Premium compounder
• AI quality understood & paid for
• Durability, not discount

• Durable but fully valued
• Limited asymmetry

• CEO transition (long-tenured McKinstry era ended)
• Health segment competition intensifying

Quality understood; AI optionality not separately priced

• UpToDate AI products; FY guide

• Web of Science — citation graph linking ~2B scientific citations
• Derwent (patents) + Cortellis (drug-pipeline intelligence)
• ProQuest academic content: dissertations, archives, ebooks
• Valuable for research/IP agents — 'a poor man's Elsevier'
• Data quality seen as better than the company's execution

• Citations/patents grow with global publishing — steady, not accelerating

• Signed access deals (Anthropic) + MCP exposure
• AI research assistants in pipeline, slow to ship
• Citation + patent networks useful for IP/research AI
• ~$4.5B net debt constrains reinvestment
• Behind on the curve — data ready before the company

• Anthropic access agreement + MCP exposure for Web of Science
• No disclosed $; debt limits investment

• Patent/citation grounding for research agents
• ProQuest licensing to labs

• ~2.3x EV/Sales (EV ~$5.7B, mostly debt) on $2.46B rev
• Equity (~$1.5B) is a small levered stub
• Cheap on sales, but the debt is the risk

• Small equity stub over ~$4.5B debt ≈ a call option on the enterprise
• Bounded loss, multi-bagger upside if it de-levers/monetizes
• Convex but a high-probability left tail — size accordingly

• ~$4.5B debt wall dominates everything
• PE overhang; serial restructurings and writedowns

Debt story drowns the data story entirely

• De-leveraging milestones
• Any AI-licensing disclosure
• Segment divestitures

• ~500M licensed, rights-cleared, caption-annotated images & video
• Exclusive editorial archives spanning a century
• iStock + Unsplash extend the catalog across tiers
• Rights-cleared image–text pairs = ideal multimodal training data
• Legal indemnification is the product AI builders need

• 160k+ events covered/yr; ~600k creators; thousands of assets ingested daily (Q2 PR)
• Editorial is a daily flow machine, not just a vault — FY25 grew both segments
• Risk is creative-side inflow: genAI erodes contributor economics

• Perplexity multi-yr display deal (Oct 2025)
• Generative tools with NVIDIA; licensed-data posture vs scrapers
• Shutterstock merger (UK-cleared May 2026) adds its lab licensing deals
• Litigation (Stability AI) continues to define the rights frontier
• Licensing not yet replacing what AI takes from stock demand

• Perplexity multi-yr display deal, Oct 2025 — undisclosed $ (PR)
• NVIDIA-powered licensed generative tools (Getty/iStock)
• Shutterstock brings lab deals (OpenAI, Meta, Apple, Amazon) post-merger

• Post-merger: consolidated licensed-visual-data vendor to every lab
• Display/attribution deals with other AI search products
• Indemnified training data as a product line

• ~$0.3B cap — a deep-distress equity stub over ~$1.3B+ debt
• ~1.5x EV/Sales on ~$0.9B revenue
• Cheap + levered = a lottery ticket on the data

• Distressed, levered equity on ideal data — near-binary
• Multiplies on a licensing/M&A catalyst, or drifts to zero
• Steeply convex, lowest-conviction high-convexity name

• ~$1.3B+ debt; controlled company (Getty family + Koch)
• Shutterstock integration risk; CMA found UK editorial concerns (remedies)

Every AI headline attaches to it; the balance sheet, not awareness, is the constraint

• Shutterstock merger close (UK-cleared May 2026)
• Combined AI-licensing revenue line
• Stability AI litigation outcomes
• Debt refinancing

• Education content, assessment & learning-outcome data
• Proprietary curriculum + testing content
• Education is an AI-disruption epicenter

• Assessment/courseware data flows with enrollment

• AI partnerships to license/embed content
• Two-sided disruption: tutoring threat + licensing optionality
• Mid on the curve

• AI partnerships announced 2025 with Microsoft, Google Cloud & AWS for learning products

• Curriculum licensing into AI tutors; assessment data moats

• Disruption discount
• Optionality + threat both real
• Owner with genuine two-sidedness

• Content-licensing/AI-tutoring optionality
• vs a real disruption threat
• Two-sided convexity

• Enrollment cliffs + OPM decline in higher ed
• Multi-year strategic rebuild under newer CEO

Two-sided: tutoring threat vs licensing option

• Enrollment AI products; partnership revenue

• High-frequency satellite imagery + Spectra AI geospatial intelligence
• Rapid-revisit imagery over own and third-party sensors
• Growing multi-year defense backlog
• Same theme as Planet, earlier in scaling

• Constellation growth (Gen-3) raises capture rate; archive accrues

• $100M+, 7-yr international defense contract (Jan 2025)
• $30M+ multi-year Gen-3 tactical ISR deal (Q3 2025)
• Backlog $323M, 91% international
• Spectra AI analytics layer over own + third-party sensors
• Same curve as Planet, earlier and cheaper-cap stage

• $100M+, 7-yr int'l defense contract, Jan 2025 (PR)
• $30M+ Gen-3 tactical ISR deal (Q3 25); backlog $323M, 91% int'l

• Gen-3 constellation upsells
• US budget normalization
• Spectra analytics licensing to allied gov'ts

• ~$1.2B cap, EV ~$1.25B on ~$107M revenue
• ~12x EV/Sales on ~flat revenue — richly valued, not cheap
• Defense backlog is the story; the price is not a discount

• Unique imagery + defense backlog = real optionality
• But ~12x sales on flat revenue means you pay up for it
• Not the bounded-downside cheap option it first looked like

• Dilution history; international customer concentration
• Gen-3 execution timeline risk

Defense-AI story increasingly recognized; ~12x sales already pays for it

• Gen-3 launch & tasking milestones
• US budget resolution
• New int'l capacity commitments

• Works on gov geospatial/intel data it doesn't own
• Palantir-type: analytics layer on others' data

• Doesn't own the data it works on

• AI analysis agents on others' data
• Services, not a data owner

• AI services on government data it doesn't own

• —

• Cheap services multiple
• Not a data-owner screen fit

• Services multiple, no data optionality

• Recompete cycles; budget continuing-resolution exposure

Services multiple, services story

• Award cycles

• Images the entire landmass daily (~3.5m), plus high-res SkySat/Pelican (~50cm)
• A unique multi-year temporal archive no competitor has
• Change-over-time is the moat — can't retroactively collect history
• Increasingly delivered as AI-ready analytics
• Defense & intelligence is the fastest-growing buyer

• Whole-Earth scan daily — the archive grows every 24h by construction
• New satellites add resolution/cadence; history can't be re-collected

• Anthropic partnership (Mar 2025): Claude applied to satellite imagery
• First prime win on NGA Luno ($12.8M, maritime AI analytics)
• MDA SHIELD IDIQ prime — eligible for Golden Dome task orders
• Backlog ~$900M (+79% YoY); Q4 revenue +41%
• AI analytics is the product; defense is the buyer

• Anthropic partnership (Mar 2025): Claude on satellite imagery (report)
• NGA Luno prime win $12.8M (SpaceNews)
• MDA SHIELD IDIQ prime (Golden Dome-eligible); backlog ~$900M

• Golden Dome task orders
• AI-analytics subscriptions over the archive (insurance, ag)
• More foundation-model partnerships on temporal imagery

• ~28x EV/Sales, pre-profit — the data and backlog are the appeal, not the multiple
• Backlog ~$900M anchors forward revenue
• Rich on every metric

• Unique archive + ramping $906M defense backlog = large 'if it scales' upside
• Pre-profit/capital intensity is the downside
• Strong positive convexity

• SPAC-era dilution legacy; Pelican capex cycle
• Government contract concentration & timing lumps

AI + defense premium fully in the ~28x; expectations are the risk

• Next earnings: ~early Sept 2026 (FQ1'27 reported Jun 4 — record print) (Q1 8-K)
• FY27 guide raised to $425–441M (+41% mid); Q2 guide $102–107M with adj-EBITDA breakeven-to-positive
• Backlog $906M (+72%), RPO $816M (+81%); ~40% of backlog converts within 12 months
• Pelican cadence: 3 launched in Q1 incl Sweden's first sovereign recon satellite
• $731M cash funds the capex cycle; NGA $22M extension; Golden Dome task orders the option

• Weather/maritime/RF data (Spire); hyperspectral imagery (Satellogic)
• Niche proprietary sensor data
• Early and capital-intensive

• Continuous sensor flow (weather/RF/hyperspectral); small base

• Real but early sensor datasets
• On the curve but small

• Niche gov/defense sensor contracts

• Weather/RF data into forecasting AI

• Speculative micro-caps
• Watchlist-only owners
• High risk, thin coverage

• Micro-cap sensor data — binary
• Large upside if a dataset scales, fat left tail
• High-variance convexity

• Cash runway and listing-compliance history — survival-grade risks

Below the radar entirely

• Contract wins; cash runway

• Exclusive official league-data rights (NFL, NCAA, EPL)
• Now the NCAA's official data provider
• The other half of the official-sports-data duopoly with Sportradar
• Growing media/ad data layer (post-Legend acquisition)
• Multi-year rights = a hard moat

• More leagues, deeper tracking (player-level optical) each season

• AI for fan engagement and betting integrity products
• Media/ad data layer monetizes the rights twice
• Growing ~25%; up the curve like Sportradar
• Owner actively monetizing, not just holding

• No corpus licensing; exclusive NFL/NCAA/EPL rights in-product
• BetVision + media/ad data layer (Legend acq.)

• Second monetization of rights via media/ads
• AI integrity & fan-engagement products

• ~$1.7B cap, ~3.6x EV/Sales on +31% growth
• Cheap for the growth + the official-data rights duopoly
• Media/ad layer monetizes the rights twice

• Rights moat + media-data optionality = asymmetric upside
• Growth-priced, so not deeply cheap
• Convex if the media layer scales

• NFL warrant dilution; rights renewals can reset economics
• Only recently profitable

Same as Sportradar — the duopoly's AI angle is unpriced

• Next earnings: ~early Aug 2026 (Q1 reported May 8)
• Legend closed May 1 → FY26 guide ~$990M–$1.01B rev / $270–280M EBITDA (~28% margin) (Q1 call)
• NFL rights locked through Super Bowl 2030; GeniusIQ to automate the full rights portfolio by end-2027
• Prediction markets: market makers onboarded in Q1 on low-latency feeds
• Targets: positive GAAP net income 2027; ≥60% uFCF conversion by 2028; ~$100M H2'26 cash flow

• Official, licensed sports-data rights — 900k+ events, 80+ sports
• Real-time play-by-play feeds, pre-match & live odds, streaming
• Multi-year exclusive league contracts = hard-to-replicate moat
• Half of a duopoly with Genius for official betting data
• The data backbone of the global betting industry

• Event coverage (900k+/yr) and in-play depth keep expanding

• AI for in-play personalization, risk/trading, content generation
• Higher-margin products (MTS, 4Sight) lift take-rates
• A genuine owner monetizing its corpus
• Up the curve; AI deepens products vs a new licensing line
• Recent Kalshi deal extends into prediction markets

• No corpus licensing — official-data rights monetized in-product
• Kalshi deal extends feeds into prediction markets

• AI in-play products lift take-rates (4Sight, MTS)
• Prediction-market data feeds scale

• ~3.0x EV/Sales on ~12% growth for a rights-duopoly owner
• Reasonable on metrics for the moat
• Fair-to-slightly-cheap

• ~3x sales on duopoly rights gives a floor
• Upside from take-rate growth on new products
• Balanced, slight positive tilt

• Rights-cost inflation: leagues extract more each renewal
• Founder (Koerl) control; bookmaker customer concentration

Priced as a betting vendor; data-rights duopoly rarely framed as AI

• Next earnings: ~early Aug 2026 (Q1 reported early May)
• NOW: FIFA World Cup (Jun–Jul 2026) — major in-play/MTS volume event (Q1 call)
• FY26 reaffirmed: 23–25% cc revenue growth / 34–37% EBITDA growth
• Prediction markets 'imminent, potentially material' — H2 ramp
• IMG Arena synergies above 25% target; >700k streamed matches in 2026; H2 restructuring for leverage
• Short-seller reports — CEO pushed back on call; monitor, don't ignore

• Ad-verification/fraud data (DV, healthier franchise)
• Cross-platform audience measurement (Comscore, distressed)
• Proprietary measurement data

• Verification events track media volumes

• Measurement owners; AI + walled gardens pressure the moat
• DV is the credible franchise; SCOR a broken business

• AI-content verification products (DV)

• Verification layer for AI-generated ad content

• ~$1.6B cap, ~2.0x EV/Sales on +14% growth
• Cheap for an ad-verification data owner (DV)
• DV the franchise; SCOR the distressed lottery leg

• ~2x sales for a profitable measurement owner
• AI + walled gardens pressure the moat
• Cheap enough to be convex

• Ad-budget cyclicality; IAS rivalry compresses pricing; SCOR is balance-sheet-fragile

De-rated with adtech; AI angle minor

• DV growth; SCOR restructuring

• Panel/clickstream traffic, keyword, conversion estimates for nearly every site
• The dataset everyone uses to track digital behavior — incl. AI-search traffic
• Broad coverage, but modeled/estimated, not a first-party record
• Continuously updated digital-intelligence feeds

• Clickstream flow is constant but panel-based — quality needs constant defense
• Privacy/cookie shifts are structural headwinds to collection

• Sells data feeds/APIs + MCP integrations into AI workflows
• Uniquely positioned to measure (and feed) the AI-search era
• Ahead for its size — high AI exposure per dollar
• Catch: modeled data less defensible than owned

• Sells AI/clickstream datasets + MCP integrations into AI workflows
• The standard source for tracking ChatGPT/Gemini traffic share

• AI-data ARR as a disclosed line
• Agent-platform data feeds
• Strategic acquirer interest (data fits many buyers)

• ~$0.36B cap, EV ~$0.21B on ~$283M revenue
• ~0.7x EV/Sales — strikingly cheap, even for modeled data
• Deep-value + AI-licensing optionality; small & illiquid

• <1x EV/Sales with AI-licensing pull = asymmetric
• Small, illiquid, modeled (non-owned) data = the risk
• Cheap enough that convexity tilts positive

• Nano-cap liquidity; SBC heavy; privacy rules threaten collection methods

Its datasets are quoted everywhere; the equity is ignored at ~0.7x EV/S

• Next earnings: ~mid-Aug 2026 (Q1 reported May 13)
• Second large LLM training contract expected 'over the coming quarters' (Q1 6-K)
• AI revenue trajectory: 11% of Q4 revenue, ~3x YoY — does it keep compounding?
• RPO $297.7M (+18%); multi-year ARR at 64% — contract-quality migration
• FY26 guide $307–315M; low end already raised once

• Ad-bidding/bidstream data + UID2 identity framework
• Powers its own bidding (demand-side platform)
• Vast behavioral data, but an input

• Bidstream data scales with ad volume; ephemeral by nature

• Stewards the UID2 identity standard
• Identity-data optionality, not a corpus sale
• De-rated; case on ad-platform fundamentals

• Kokai AI in-platform; UID2 stewardship

• UID2 as identity layer for agentic commerce

• ~3.0x EV/Sales on +18% growth — value territory for profitable adtech
• UID2 identity optionality on top
• Open-web AI fears embedded in the multiple

• Modest multiple + identity-standard optionality
• Data is an input, not a sold corpus
• Positive tilt on metrics

• Founder super-voting; Amazon DSP is the real competitive event
• SBC and the credibility hit from the '25 stumble

AI read as open-web risk; de-rate reflects it

• CTV share; UID2 adoption; growth re-accel

• B2B contact + company intelligence: emails, dials, org charts, technographics
• Buying-intent signals across millions of companies
• A live 'who's-who' graph of decision-makers
• Real, but increasingly replicable as AI shifts buyer behavior
• Renamed platform around 'GTM AI'

• B2B contact data decays ~25–30%/yr — must be rebuilt constantly
• Customer churn weakens the contributory refresh loop
• The clearest decaying-asset risk in the table

• GTM Context Graph native in OpenAI's Codex for Work — agent context layer
• AI is both distribution and disruptor
• Cut 2026 guidance + ~20% of staff on AI-driven shifts
• Ahead on plumbing, behind on the seat-based model
• Clearest live case of 'data doesn't protect the equity'

• GTM Context Graph natively in OpenAI's Codex for Work
• No disclosed licensing $; positioning as agent context layer

• Per-call context pricing for sales agents
• More agent-platform embeds (Claude, Gemini)
• Data-only tier decoupled from seats

• ~1.7x EV/Sales — lowest multiple on the board
• But revenue is declining; the cheapness reflects decay risk
• Statistically cheap; operationally a falling knife

• ~1.7x sales embeds heavy pessimism — small asymmetric base
• Re-rates hard if revenue stabilizes as the agent-context layer
• Declining revenue is the live left tail

• Debt on a shrinking base; SBC dilution; churn is the whole story

The market's AI-victim poster child; the Codex embed is ignored

• Next earnings: ~early-mid Aug 2026 (Q1 reported May 11)
• The trough test: FY26 guide cut to $1.185–1.205B (−4% mid); Q2 $300–303M — does it hold? (Q1 call)
• Agent embeds: Salesforce prospecting agent ships with ZoomInfo as first/primary external data source (150k+ customers); HubSpot native; ChatGPT/Claude/Copilot/Perplexity connectors live
• Pricing pivot: Copilot moving from seats to prepackaged credits/consumption
• Mgmt points to growth returning H2 2027; 35% AOI margin + cost cuts fund the wait

• Wholesale used-car condition & transaction data (ACV inspection corpus)
• Granular vehicle-condition/pricing data
• Still primarily marketplaces

• Inspection corpus grows with every vehicle listed (ACV)

• Feeds AI pricing
• ACV more data-distinctive
• Operators, not data-unlock plays

• ACV inspection-AI in-product

• Condition-data licensing to pricing AIs

• ACV the more data-distinctive
• Both operators
• Corpus enhances the platform

• ACV growth + condition-data optionality
• Valued on the marketplace
• Mildly positive

• ACV not yet sustainably profitable; OPENLANE balance sheet

Marketplace story

• GMV growth; take rates

• Auto listing, pricing & shopper-intent data
• Largely audience/marketplace
• Listings not fully proprietary

• Listings churn; intent data flows with traffic

• Useful intent data, Zillow-like
• Not a data-unlock play

• In-product pricing AI

• —

• Reasonable valuations
• Operators in the Zillow mold

• Operator, limited data asymmetry
• Balanced-to-low

• Dealer-count churn; marketing-spend treadmill

Marketplace story

• Dealer counts

• Salvage-auto auction & vehicle-history data (IntelliSeller)
• Decades of auction-outcome data
• Serves its dominant auction marketplace

• Salvage auction outcomes accumulate with volume

• AI tools in-product, not licensed
• Data deepens the moat, isn't the product

• Internal auction AI (IntelliSeller)

• —

• Premium, high-quality operator
• Data deepens the moat, isn't the product

• Premium, data not the re-rate driver

• Leadership transition from founder era; totals cycle depends on used-car values

Operator story

• Volume cycles

• Grocery-purchase + fast-growing retail-media ad data
• Rich first-party purchase data
• Powers its own high-margin ads (input)

• Purchase graph deepens with order history

• Strong data-driven ad engine
• AI-relevant, but feeds its ads, not sold
• Operator class

• Retail-media AI in-product

• Purchase-data into commerce agents (never signaled)

• Reasonable on ads + delivery
• Strong ad engine
• Data is an input

• Retail-media optionality
• Valued on the business, not the data
• Balanced

• DoorDash/Uber entering grocery; ad growth must outrun fee pressure

Grocery/ads story

• Ad revenue growth

• Merchant transaction flows & fraud signals (banking/payments processing)
• Real data, but serves its processing

• Transaction flow tracks processing volumes

• In-product fraud/upsell, not a corpus

• Fraud AI in-product

• —

• Cheap-ish fintech
• But not a data re-rate

• Value fintech, data not the driver

• Worldpay separation aftermath; bank IT spending cycles

Fintech story

• Banking IT spend

• Among the largest transaction datasets on earth
• Regulated, privacy-bound byproduct
• Not licensed as a corpus

• Payment volumes grow ~10%/yr — among the largest data flows on earth

• Increasingly productized
• But privacy-bound; not a corpus sale
• The ultimate data-advantaged operators

• Internal fraud/credit AI at vast scale; agentic-commerce pilots

• Agentic payments standards (who authorizes an AI's purchase?)

• Valued as payment giants
• n/a as a data re-rate

• Priced payment networks; data is internal

• Interchange regulation (CCCA) and DOJ debit suit (V)
• Stablecoin rails as long-term routing threat

Agentic commerce chatter rising; data never the thesis

• Agentic-payment standards; volume growth