25 stocks vs the S&P 500.
25 high-moat names chosen for an AI-driven world.
IM25 vs S&P 500
IM25
-0.35%
NAV 99.65
VOO
+0.53%
Index 100.53
-0.88% pp vs VOO
| Date | IM25 | VOO |
|---|---|---|
| 2026-08-21 | 100.00 | 100.00 |
| 2026-08-24 | 99.73 | 99.73 |
| 2026-08-25 | 100.48 | 100.04 |
| 2026-08-26 | 100.62 | 100.07 |
| 2026-08-27 | 102.35 | 100.72 |
| 2026-08-28 | 101.56 | 100.50 |
| 2026-08-31 | 100.68 | 100.17 |
| 2026-09-01 | 98.29 | 99.51 |
| 2026-09-02 | 98.05 | 99.96 |
| 2026-09-03 | 99.93 | 101.00 |
| 2026-09-04 | 99.65 | 100.61 |
| 2026-09-08 | 98.50 | 100.05 |
| 2026-09-09 | 97.29 | 99.60 |
| 2026-09-10 | 96.46 | 99.00 |
| 2026-09-11 | 97.18 | 99.84 |
| 2026-09-14 | 97.15 | 99.37 |
| 2026-09-15 | 96.01 | 98.93 |
| 2026-09-16 | 96.13 | 98.51 |
| 2026-09-17 | 96.66 | 99.62 |
| 2026-09-18 | 96.65 | 99.73 |
| 2026-09-21 | 98.60 | 101.29 |
| 2026-09-22 | 98.89 | 101.29 |
| 2026-09-23 | 98.59 | 100.55 |
| 2026-09-24 | 98.72 | 100.47 |
| 2026-09-25 | 98.76 | 101.01 |
| 2026-09-28 | 98.15 | 99.99 |
| 2026-09-29 | 98.49 | 99.82 |
| 2026-09-30 | 98.44 | 99.59 |
| 2026-10-01 | 98.90 | 99.81 |
| 2026-10-02 | 99.54 | 100.54 |
| 2026-10-05 | 101.26 | 101.22 |
| 2026-10-06 | 102.91 | 101.77 |
| 2026-10-07 | 101.99 | 101.51 |
| 2026-10-08 | 101.31 | 101.08 |
| 2026-10-09 | 102.82 | 101.69 |
| 2026-10-12 | 102.42 | 101.62 |
| 2026-10-13 | 101.89 | 101.63 |
| 2026-10-14 | 99.65 | 100.53 |
The Thesis
Why systematic moat investing?
01
Economic Moats Compound
The best businesses become harder to compete with over time. Pricing power, switching costs, and network effects strengthen as the business scales, delivering above-market returns on capital for decades.
02
AI Rewrites the Playbook
Most competitive advantages are AI-vulnerable. Network effects, proprietary data, and regulatory lock-in carry 80% of the moat score — the advantages AI cannot replicate or destroy. Bundling and encoded workflow still count, at 20%.
03
Concentration Beats Diversification
Owning 500 companies means funding mediocrity at scale. 25 high-conviction positions, each earning its place with composite ≥80 and moat ≥70, concentrate capital where it compounds fastest.
The Scoring Framework
How every stock earns its score
12 weighted moat types. Resilient group 80%, vulnerable 20%. Strength bonus up to +3 for extra strong resilient pillars.
Blended 3-5 year revenue CAGR with named adjustments:
Price vs. scenario targets with live price feeds:
The bands rate an asset on its own merits; portfolio inclusion is separate and stricter. Each pillar is standardised against the coverage universe before weighting, so a 1-sd move in any pillar shifts the composite in proportion to its stated weight. Eligibility requires composite ≥ 80 and moat ≥ 70. The 25 is a committed book.
The Moat Framework
Not all moats survive the AI era
Every business is scored across 12 competitive advantages. Seven are AI-resilient (80% of the score) because AI cannot replicate them; five are AI-vulnerable (20%) because intelligent agents can increasingly substitute for them — they modify the durable score, they do not match it. Each moat is rated strong (100) · intact (65) · weakened (35) · destroyed (0) · na (excluded). Individual moats can be reclassified per company where AI strengthens rather than erodes them — NVIDIA's CUDA, Palantir's ontology.
Value compounds with every new participant
Following Metcalfe's Law, value scales with the square of participants. Every new user makes the network more valuable for all existing users — an enormous structural disadvantage for any challenger.
Private, compounding data flywheels
Data that accumulates privately over time and cannot be purchased or replicated. The longer the company operates, the harder it becomes to catch up. Think HealthKit biometrics, Palantir's classified datasets, or Visa's transaction graph.
The authoritative source of truth
The company's data store is the canonical reference all downstream systems defer to. Replacing it requires migrating years of history and retraining every workflow built on top — so customers never voluntarily leave.
Government licences, certifications & mandates
Advantages granted by law: FDA approvals, financial licences, index inclusion, spectrum rights. These cannot be automated away; the certification process itself is the moat.
Sitting inside the payment layer
The business is embedded directly in the financial flow of every transaction. Removing it requires rebuilding critical infrastructure — not just switching a preference.
A unit-cost lead rivals cannot spend past
Volume, route density or a low-cost resource base lets the company make or deliver each unit cheaper than anyone else, and the saving funds the lower price that keeps the volume. Software rarely qualifies; physical networks and the best orebodies do.
A premium the name alone sustains
Customers pay more, or choose it by default, because of the name — and the premium holds through downturns without discounting. Only credited where it is a separate source of durability, not the face of a network already rated.
Embedded operational workflows
The software encodes years of accumulated business rules that employees rely on daily. While this creates significant switching costs today, AI can increasingly model and reproduce business logic, gradually eroding the cost of migration.
Value from combining complementary products
Multiple products packaged together create convenience that point solutions can't match. AI-driven software commoditises features rapidly, making it easier for focused challengers to replicate any single element.
Fluency built through years of UI habit
Users invest time mastering a specific interface — keyboard shortcuts, mental models, workflows. AI agents increasingly abstract away the interface layer, letting users command outcomes without learning a specific UI.
Rare human expertise as competitive advantage
The business depends on a small pool of specialists whose skills are hard to find. AI augments and in some domains replaces highly skilled human work, compressing the scarcity premium over time.
Privileged access to publicly available information
The company has a head-start aggregating data that is technically public but expensive to compile. AI web-crawlers and LLMs rapidly close this gap by training on the same underlying sources.
Each moat is rated strong (100) · intact (65) · weakened (35) · destroyed (0) · na (excluded). Inapplicable moats use status na and drop out of their group. The two groups blend 80 / 20 (resilient / vulnerable). A thin resilient book is blended toward intact, and a strength bonus of +1 to +3 rewards additional strong resilient moats beyond the second. That produces the Moat Score (40% of composite).
Common Questions
Questions worth asking
Ready to get started?
View the IM25
Explore the allocation, then drill into coverage, sectors, or the cross-cutting research — with moat scores, scenarios, and live valuations throughout.