A Founder's Resource

No single moat is durable. Stack them.

"Building a moat is harder than ever. AI has made the playing field more accessible than it's ever been."

— Gokul Rajaram, 20VC · March 2026

A practical guide to defensibility in the AI era — built on Buffett's original insight, Hamilton Helmer's 7 Powers, and Gokul Rajaram's updated 8 Moats framework, with our synthesis of how those moats stack into six durable compounds.

By John Gusiff · Krzysztof Czubak

For

Founders & operators building durable companies

Reading time

15 to 20 minutes

What you'll leave with

A working model for assessing and stacking your defensibility

Where the idea came from

The lineage of the moat

The concept of a "moat" has evolved across three eras. Each thinker added a layer the previous one didn't need.

01

Warren Buffett

1990s — Industrial era

Buffett popularized the "economic moat" as the durable competitive advantage protecting a company's profits — like a castle's moat keeping invaders out. Brands, switching costs, and scale economics were the canonical examples. The framing was deliberately simple: find one strong moat, hold it for decades.

"The most important thing is finding a business with a wide and long-lasting moat."

02

Hamilton Helmer

2016 — Software era

In 7 Powers, Helmer turned moats into a rigorous taxonomy: scale economies, network economies, counter-positioning, switching costs, branding, cornered resources, and process power. Each "power" had to satisfy both a benefit (you earn more) and a barrier (competitors can't easily copy). It became the canonical strategy text for software.

"A power is a persistent differential return."

03

Gokul Rajaram

March 2026 — AI era

On 20VC with Harry Stebbings, Gokul updated Helmer's framework for a world where software is easier than ever to build. He proposed eight moats relevant to today's AI-native companies — and argued that no single one is sufficient anymore. Defensibility now requires stacking.

"In a world of AI, you need at least four of these eight moats to be enduring."

A deliberate omission9th?

Why brand is missing from the eight

Anyone trained on Buffett or Helmer will notice it: brand — the most famous moat of all — isn't in Gokul's list. That's not an oversight. It's a deliberate decision worth understanding before reading further.

Gokul's reasoning: brand strength is hard to measure on its own, and when you trace it back, it usually turns out to be a downstream effect of something more fundamental — network effects (Facebook), scale (Amazon), distribution (Apple), or workflow embedment (Salesforce). Brand is the visible surface. The moat underneath is structural.

This is the sharpest break with the Buffett era. Coca-Cola's brand was the moat. In the AI era, brand is a signal of a moat — not the moat itself.

"I struggled whether to include Brand in this list of moats. I ultimately decided not to, because I think it's too hard to measure, and in many cases the underlying brand strength is due to something more fundamental — network effects, scale, distribution." — Gokul Rajaram, on X
The framework

The 8 moats of the AI era

Gokul's eight categories, with the original definition, why each one — on its own — is no longer enough, and a B2C and B2B failure case for each.

Click any card to see the failure mode and the evidence behind it.

01

Data Moat

Proprietary data that compounds and improves the product over time.

Why it fails alone

Models commoditize. Data gets replicated. Yesterday's insight becomes table stakes by next quarter.

Evidence B2C
Stack Overflow & Chegg

Stack Overflow had a decade of developer Q&A data — the gold-standard programming knowledge graph. After ChatGPT launched, traffic dropped roughly 50%. Chegg, which sold human-written homework solutions, lost more than 99% of its market value as students switched to free AI alternatives. Real data — but not data competitors couldn't replicate.

Evidence B2B
ZoomInfo (now GTM)

ZoomInfo built the canonical B2B contact and intent dataset, growing 47% YoY in 2022. By Q1 2025 it posted its first revenue decline. Stock fell ~53% from peak. The company rebranded to GTM and laid off 6% of staff. Bond markets now price 670bps of AI-displacement risk into its 2029 notes. Comprehensive B2B data was a moat — until AI-native competitors could rebuild the graph from scratch.

− Hide
02

Workflow Moat

Embedded in daily work, making the product hard to remove or replace.

Why it fails alone

Habit is strong but not permanent. Better UX appears, AI removes steps, and teams reset workflows when the gain is large enough.

Evidence B2C
Mint.com

Mint had 25M+ users, daily-habit financial workflows, and 15 years of category leadership in personal finance. Intuit acquired it in 2009, ran it as a free user-acquisition funnel for QuickBooks and TurboTax, and shut it down in March 2024. The workflow embedment didn't save the product — the parent company's strategy did. Personal habit isn't a moat when the platform owner has a different incentive.

Evidence B2B
Notion, Monday & Asana

By early 2026, AI agents began routing around workflow tools rather than using them. SaaStr's CEO publicly described cancelling Notion entirely — the workflows that lived there got absorbed by agents writing directly into Salesforce and Slack. Monday and Asana stocks fell 39% in the same selloff as analysts realized agents could replace 'expensive intermediaries between humans and a database.'

+ Why it fails alone + evidence
03

Regulatory Moat

Protected by compliance, certifications, or legal barriers.

Why it fails alone

Regulation changes. New entrants adapt faster than incumbents. Compliance eventually becomes table stakes, not advantage.

Evidence B2C
European retail banks vs. PSD2

Incumbent European banks treated regulatory complexity as a defensive asset for decades. Then PSD2 (2018) forced them to expose customer data and payment APIs to any licensed third party. Challenger banks like Monzo, Starling, and Revolut — plus fintechs like Plaid — flooded in. The same regulation that protected incumbents was rewritten to dismantle them.

Evidence B2B
Epic & Cerner vs. the Cures Act

EHR vendors built decades of regulatory moat through HIPAA-compliant integrations, certifications, and proprietary data formats. The 21st Century Cures Act's Information Blocking Rule, fully enforced from July 2024, mandated FHIR APIs and free patient data export — punishing vendors and providers who 'block' access. The compliance asset that locked in customers was rewritten to enable third parties to compete with the incumbents.

+ Why it fails alone + evidence
04

Distribution Moat

Control over how customers are acquired and reached.

Why it fails alone

Channels saturate. CAC rises. Platforms change rules. Distribution is powerful — until the platform owner decides it isn't.

Evidence B2C
Meta and the Apple ATT shock

Meta's distribution moat — granular cross-app targeting — was the foundation of a multi-hundred-billion-dollar business. In April 2021, Apple shipped App Tracking Transparency. Roughly 80% of users opted out. Zuckerberg announced a $10 billion revenue hit for 2022 alone. DTC brand CAC on Facebook tripled in two years. The moat didn't erode — one platform decision rewrote the rules overnight.

Evidence B2B
HubSpot vs. Google AI Overviews

For 15 years, B2B SaaS distribution ran through Google: SEO content, comparison pages, 'best CRM' rankings. Then AI Overviews arrived. HubSpot — owner of one of the best content-marketing engines in tech — saw monthly organic traffic collapse from ~13.5M visits in November 2024 to under 7M weeks later. CEO Yamini Rangan acknowledged it on the earnings call: 'AI overviews are giving answers, and fewer people are clicking through.' The channel they built the company on was rewritten by the platform owner.

+ Why it fails alone + evidence
05

Ecosystem Moat

A platform others build on, creating dependency and expansion.

Why it fails alone

Developers chase growth. Better platforms emerge. Incentives shift. Ecosystems must stay attractive — they don't stay by default.

Evidence B2C
Facebook Platform deprecation

From 2007 to 2018, Facebook Platform powered tens of thousands of consumer apps — Zynga's FarmVille, music apps, photo apps, news apps. Cambridge Analytica forced Facebook to lock down APIs in 2018, killing third-party access overnight. Zynga's stock had already lost ~75% of its value depending on a single platform; thousands of smaller apps simply died. The ecosystem that powered consumer social was rewritten by the platform owner in months.

Evidence B2B
Twitter's developer exodus

Twitter's developer ecosystem helped invent hashtags, retweets, URL shorteners, and the first iPhone client. In 2023, the company killed free API access and priced enterprise tiers at $42K/month. Hundreds of apps shut down within weeks. Developers migrated to Mastodon, Bluesky, and Threads. The ecosystem that made Twitter culturally indispensable evaporated when the platform owner changed the deal.

+ Why it fails alone + evidence
06

Network Effects Moat

Value increases as more users participate.

Why it fails alone

Users keep options open. Communities migrate. New niches emerge. Networks are stronger than they look — and weaker than they seem.

Evidence B2C
Facebook vs. TikTok

Facebook had the canonical social network effect: billions of users, deep friend graphs, full advertiser tooling. TikTok proved the moat was narrower than it looked. By segmenting attention around content rather than friends, TikTok captured a generation that 'multi-homed' rather than switched. Harvard Business School research now classifies Facebook's network as 'highly segregated' — meaning the network effect protects subgroups, not the platform.

Evidence B2B
Slack vs. Microsoft Teams

Slack built the canonical B2B network effect — teams that joined invited their teams, channels created switching cost, integrations deepened embedment. By 2024 Slack had ~42M DAU and Salesforce paid $27.7B for it. Microsoft Teams hit 320M MAU by bundling free with Microsoft 365. The EU opened an antitrust case in 2023; Slack 'won the battle' to unbundle and lost the war. A network effect was real — and a distribution moat ate it anyway.

+ Why it fails alone + evidence
07

Physical / Infrastructure Moat

Ownership of hard-to-replicate real-world systems.

Why it fails alone

Capital eventually catches up. New tech leapfrogs old systems. Regulation can open access. Infrastructure buys time, not immortality.

Evidence B2C
Blockbuster & Kodak

Blockbuster operated 9,000+ stores at its 2004 peak — a real-estate moat that took 20 years to build. Netflix bypassed it with mail-then-stream and Blockbuster filed for bankruptcy six years later. Kodak's situation was sharper: the company invented the digital camera in 1975 but protected its film infrastructure instead of cannibalizing it. Both moats were real. Both were leapfrogged by a different physics.

Evidence B2B
Cisco vs. cloud-native networking

Cisco built a decades-long infrastructure moat through specialized hardware, certified engineers, and physical network appliances. Then AWS, Azure, and Cloudflare turned networking into software — VPCs, software-defined networking, edge compute as APIs. Cisco's revenue stagnated while AWS networking became a $20B+ business inside a much larger platform. The hardware moat held its margins, but the unit of work moved to someone else's stack.

+ Why it fails alone + evidence
08

Scale Moat

Size-driven advantages in cost, pricing, and reinvestment.

Why it fails alone

Big often becomes slow. Complexity increases. Smaller players innovate faster. Scale defends margins, not relevance.

Evidence B2C
Cable & telecom incumbents vs. streaming

Comcast, Charter, and AT&T spent decades stacking scale moats — physical networks, content libraries, regulatory licenses, billions in R&D. Netflix, with no infrastructure of its own, rebuilt the entire entertainment supply chain on top of AWS. By 2024 cable lost more than 25M subscribers from peak. The scale advantage held the margins on the legacy bundle; it didn't defend the customer when the bundle stopped being the unit of work.

Evidence B2B
Legacy SaaS in 2026

Salesforce, HubSpot, and Zendesk built scale moats — billions in R&D budgets, thousands of integrations, enterprise-grade everything. By 2026, AI-native competitors were rebuilding their categories with a fraction of the headcount. Intercom announced a forced 'AI-first' pivot. HubSpot acknowledged AI-native pressure to investors. Scale still defends margins on installed base — but it doesn't defend relevance when the unit of work itself is changing.

+ Why it fails alone + evidence
The shift in mindset From advantage to durability
No single moat is durable on its own. Defensibility now comes from stacking.
— The core insight, simplified

The Buffett-era question was: do you have an advantage? The Rajaram-era question is sharper: can it survive AI commoditization, platform changes, and the next ten years? Advantage is a snapshot. Durability is the question that matters.

Copied

Product, data, and cost advantages can be replicated by well-funded competitors.

Outspent

Brand and distribution edges erode under sustained capital pressure.

Bypassed

Technology shifts and new business models can route around an entire category.

Abandoned

Networks migrate. Workflows reset. Switching costs decay when alternatives compound.

Regulated away

Policy change can dissolve compliance-based barriers overnight.

Decayed

Brand, scale, and product advantages all erode over time without reinvestment.

The engine inside the moat

Five loops that make moats compound

Moats are the category of advantage. Loops are the mechanism by which that advantage compounds. Every durable company runs on at least one of these — most run on several at once.

A moat without a loop is a static position. A loop without a moat is a feature competitors can copy. The combination is what creates durability.

Every loop has the same shape: a repeatable trigger that starts it, a sequence of actions that creates value, a compounding asset the business keeps, and a retained value the customer carries forward. Growth happens because something accumulates and something is retained — on both sides.

01

User Loops

Compounds per interaction
What compounds
Users · attention · demand
What customer retains
Connection, identity, belonging — the network is where their people are.

Each cycle turns existing users into a source of new users. The product spreads through use.

Examples
Referrals · network effects · word-of-mouth · viral sharing
02

Content Loops

Compounds continuously
What compounds
Content · data · knowledge
What customer retains
Confidence the answer is there, less effort to find or verify it.

Each cycle adds information that makes the product more valuable to the next user.

Examples
SEO · UGC · internal playbooks · ML training data
03

Trust Loops

Compounds over years
What compounds
Proof · credibility · reputation
What customer retains
Reduced anxiety about risk; the decision feels safe.

Each cycle reduces uncertainty for the next customer. Confidence becomes the asset.

Examples
Reviews · ratings · case studies · audit history
04

Skill Loops

Compounds monthly
What compounds
Mastery · habit · identity
What customer retains
Fluency and competence — switching means relearning.

Each cycle makes users (or teams) better at using the product. Switching means relearning.

Examples
Streaks · capability levels · certifications · expertise
05

Capital Loops

Compounds quarterly
What compounds
Money · efficiency · operational leverage
What customer retains
Predictability — better prices, better economics, fewer surprises.

Each cycle generates surplus that funds the next cycle. Margin becomes the engine.

Examples
Conversion · expansion revenue · LTV/CAC reinvestment · automation
The deeper pattern

Loops compound on different timescales — and that's why stacking works

When one loop slows, another carries. That's why the strongest companies run multiple loops in parallel — and why their timescales matter.

A single loop is a single point of failure. A stack of loops keeps compounding even when one cycle stalls.

The synthesis

Six compound moats that hold up

When you stack the right moats, they reinforce each other. Each compound creates a self-sustaining loop — the part that turns a defensible position into a durable one.

A note on attribution: The 8 individual moats are Gokul Rajaram's framework, drawn from his 20VC appearance with Harry Stebbings (March 2026). The six compound stacks below — and the business loops mapped to each — are our synthesis layered on top: a model for how the individual moats combine into durable, defensible patterns you can recognize in the companies you already know.
01

AI-Native System of Work

Moats stacked
DataWorkflowEcosystem

The product becomes how work gets done — and gets better at it the more you use it.

  • →Embedded in execution (Workflow)
  • →Improves with usage (Data)
  • →Others build agents and integrations around it (Ecosystem)
Companies running this stack

Notion · Figma · GitHub Copilot · Zapier

The business loops that matter
Content Loop compounds continuously

Use → improves the system → makes the product more capable for the next user → more use.

Compounding asset: Usage data and user-generated templates/workflows inside the product

Retained value: Fluency with the system and confidence the next task will be faster than the last.

User Loop compounds monthly to quarterly

As the system becomes more capable, it attracts integrations, agents, and partners that extend it → ecosystem grows → more reasons to stay and harder to leave.

Compounding asset: Third-party integrations, agents, and extensions

Retained value: Confidence that whatever workflow comes next, there's already an integration that supports it.

Two loops, mostly independent: one compounds inside the product, the other compounds around it. The moat is that a new entrant has to rebuild both.

02

Category-Defining Platform

Moats stacked
EcosystemNetworkDistribution

A platform that others build on top of — and customers can't easily replace because the surrounding ecosystem moves with it.

  • →Others build on you (Ecosystem)
  • →Users attract users (Network)
  • →You control access to demand (Distribution)
Companies running this stack

Shopify · Salesforce · Apple App Store · Slack

The business loops that matter
User Loop compounds continuously

More users → more developers build on the platform → more apps and integrations → more user value → more users.

Compounding asset: Active users and active developers on the platform

Retained value: The platform is where the people and developers they need already are.

Content Loop compounds continuously

Every new app, integration, or workflow on the platform expands the surface area of what the platform can do.

Compounding asset: Apps, integrations, and workflows built on the platform

Retained value: Trust that adjacent needs are covered without buying another tool.

The two loops feed each other directly: every new user is a reason for developers to build, and every new app is a reason for users to stay. Same clock speed, tightly coupled.

03

Data Intelligence Layer

Moats stacked
DataScaleDistribution

A system that makes decisions sharper through proprietary data — and gets sharper the more it's used.

  • →Proprietary data advantage (Data)
  • →Continuous improvement loop (Scale)
  • →Distribution ensures the data keeps flowing in (Distribution)
Companies running this stack

Google · Stripe · Palantir · Snowflake

The business loops that matter
Content Loop compounds continuously

More usage → more proprietary data → sharper decisions → more usage.

Compounding asset: Proprietary training and operational data

Retained value: Trust in the system's answer; less second-guessing each decision.

Capital Loop compounds quarterly

Better decisions command premium pricing → surplus margin reinvests in distribution and infrastructure → wider reach → more usage feeding the data asset.

Compounding asset: Distribution channels and infrastructure footprint

Retained value: Confidence the system will scale with them and stay ahead.

Two loops on separate tracks — one compounds the data, the other compounds operational leverage. The content loop ticks every query; the capital loop ticks every quarter. Both terminate in the same moat.

04

Regulated System of Record

Moats stacked
RegulatoryWorkflowData

A system embedded in high-stakes, compliance-heavy workflows where the cost of switching is measured in audits, not features.

  • →Embedded in high-risk workflows (Regulatory)
  • →Compliance creates workflow switching friction (Workflow)
  • →Becomes the auditable record itself — proprietary, audit-grade data (Data)
Companies running this stack

Epic Systems · Oracle Health (Cerner) · Intuit (QuickBooks) · ADP

The business loops that matter
Trust Loop compounds over years

Successful audits and compliant operation → institutional credibility deepens → competitors must clear the same regulatory bar and rebuild the data history.

Compounding asset: Audit history, regulatory certifications, and compliance track record

Retained value: Confidence the system will hold up to regulators, auditors, and inspectors.

Capital Loop compounds quarterly

Embedded workflows and audit-grade data → renewal certainty and pricing power → surplus margin funds further regulatory and product investment.

Compounding asset: Long-term contracts and entrenched customer accounts

Retained value: Operational continuity and predictable cost of ownership.

The two loops tick on different clocks — and that's the moat. A new entrant can buy capital faster than they can buy credibility, but credibility is what protects the renewal stream.

05

Demand Aggregator

Moats stacked
DistributionNetworkScale

A platform that controls supply-demand matching at a scale no individual supplier can replicate.

  • →Owns customer access (Distribution)
  • →Reinforced by two-sided network effects (Network)
  • →Scale lowers unit economics for both sides (Scale)
Companies running this stack

Amazon · Airbnb · Uber · Booking.com

The business loops that matter
User Loop compounds continuously

More buyers → more sellers join → more selection → more buyers → suppliers can't afford to leave.

Compounding asset: Active buyers and active sellers on the platform

Retained value: Confidence that supply (or demand) will be there when they need it.

Capital Loop compounds quarterly to annually

Scale lowers unit economics on both sides → surplus margin funds subsidies (early) and pricing advantages (late) that competitors can't match.

Compounding asset: Cost advantages from scale

Retained value: Trust they'll get the best price or selection without shopping around.

Tightly coupled early (subsidies feed the network), decoupled at maturity (margin no longer needs to be reinvested to hold the position). The clock speeds drift apart as the company ages.

06

Infrastructure Backbone

Moats stacked
PhysicalScaleData

Core systems that other businesses run on — protected by capital intensity and operational complexity that compounds with scale.

  • →High capital barrier (Physical)
  • →Operational complexity that compounds (Scale)
  • →Telemetry data improves performance over time (Data)
Companies running this stack

AWS · Visa · FedEx · Cloudflare

The business loops that matter
Capital Loop compounds quarterly to annually

More customers → more revenue at scale → reinvest in physical footprint and capacity → harder for new entrants to match the capital base.

Compounding asset: Data centers, fiber, switches, fleet, and physical capacity

Retained value: Confidence the infrastructure will be there, at the scale and reliability the customer's business depends on.

Content Loop compounds continuously

More customers → more telemetry → better reliability, pricing, and efficiency → keeps the infrastructure ahead of newer hardware.

Compounding asset: Telemetry data and operational playbooks

Retained value: Trust that the system gets more reliable and efficient over time, not less.

Two loops on separate tracks: one compounds the asset base, the other compounds operational knowledge. A new entrant has to out-spend and out-learn to break in.

The argument in four steps

Why this matters now

Gokul's whole thesis compresses to a four-step logical chain. If you accept the first three, the fourth is the only conclusion left.

01 →

Defensibility is getting harder

Software is easier to build than ever. Replication speed has collapsed from years to weeks.

02 →

AI is compressing advantages

Foundation models commoditize what used to be proprietary — data, workflow steps, decision logic.

03 →

Single moats are fragile

Each of the eight, taken alone, has a known failure mode. The evidence is already on the board.

04 ✓

Durable companies stack

The companies still standing in five years will combine multiple moats into self-reinforcing systems.

Put the framework to work

Run the diagnostic on your business.

Step 1 Audit your current moats Free · done for you ↓
Step 2 Match your defensibility archetype
Step 3 Build the missing leg

Run all three steps yourself with the $179 Stacking Moats Playbook — the complete DIY process. Or jumpstart Step 1 right now: submit a few details and we'll audit your moats for you, free. Your scorecard arrives the next business day.

Please enter your name.
Free providers (gmail, yahoo, outlook, icloud, hotmail) aren't accepted.
Please use your business email — free providers (gmail, yahoo, etc.) aren't accepted.
We score what's publicly visible — your homepage, product pages, about page.
Please enter your company name or URL.
Please select your stage.
Company type *
Please select at least one company type.
Something went wrong on our end. Please try again, or email [email protected] with your name, company URL, and business email.

Your scorecard, plus a short email series on what to do with it. Unsubscribe anytime.
Customer Centric Solutions LLC · customercentricllc.com

Built for founders, heads of growth, and operators who'd rather know now than find the weak leg in a Series B room. An honest 0–3 on each of the 8 moats, anchored against comparable companies — not consultant flattery.