Annotations (12)
“When Cloudflare designed their motherboards for all their chips, they left an empty slot open because they didn't know what it would be for, what the use case would be, but they needed it just in case something came up. AI inference was that something. They could go to all their boxes and simply plug in a GPU, and then AI inference became available across 330 cities worldwide. 15 years of infrastructure planning paid off in one product cycle.”— Sam Eden
Strategy & Decision Making · Technology & Engineering · Operations & Execution
DUR_ENDURING
Empty slot = future optionality
“Cloudflare provides speed and security for your website through a single reverse proxy that intercepts all traffic. Unlike legacy providers like Akamai who required customers to split their network and decide what content to route where, needing sales engineers for complex implementation, Cloudflare said: redirect everything to us. No split rules. No multiple vendors. One global network handles CDN, DDoS protection, firewall, everything.”— Sam Eden
Strategy & Decision Making · Technology & Engineering · Business & Entrepreneurship
DUR_ENDURING
Unified interception vs. split routing
“Cloudflare's real innovation was the reinforcing loop: low bandwidth costs and easy product-led growth served the long tail of customers, which brought more traffic through their servers, which collected more signals and data, making them better at blocking attacks and optimizing speed, which attracted paying enterprise customers, which brought even more traffic. More traffic let them negotiate better peering deals with ISPs to reduce bandwidth costs further.”— Sam Eden
Strategy & Decision Making · Economics & Markets · Operations & Execution
DUR_ENDURING
7-step reinforcing loop builds moat
“ISPs were in a lose-lose situation before Cloudflare: they had to pay transfer fees to connect networks across the world, and they delivered slow internet to customers who blamed the ISP for bad performance. Cloudflare aggregated the long tail of small websites the ISPs were serving anyway. Then they said: I see you're transferring a lot of bandwidth to this region. Why don't I just put my server next to yours? We'll have a peering relationship.”— Sam Eden
Strategy & Decision Making · Economics & Markets · Business & Entrepreneurship
DUR_ENDURING
Aggregate long tail for leverage
“Legacy companies like Akamai didn't offer simple unified interception because large enterprises didn't want it. This is a classic innovator's dilemma. The short-term revenue wasn't there from serving the long tail, so Cloudflare could build for that market on their own. By the time their capacity and services reached a tipping point where they were better than legacy providers, they could win enterprise customers.”— Sam Eden
Strategy & Decision Making · Business & Entrepreneurship
DUR_ENDURING
Incumbents ignore long tail = opening
“Cloudflare uses commodity hardware, not dedicated specialized hardware like competitors. They created a software-defined network inspired by Google's approach. This was a strategic decision: cheaper hardware meant they could scale their network faster and spread costs across multiple product lines. Because all their servers can run all their products, the return on each incremental CapEx spend is diversified across all revenue streams. You're not building a separate network for each product.”— Sam Eden
Technology & Engineering · Strategy & Decision Making · Operations & Execution
DUR_ENDURING
Commodity hardware = flexible scaling
“Cloudflare introduced pool of funds in 2024: a bundling method where large customers commit to a multiple-year contract and can draw down from any product. This solved a friction problem. Companies using their web security, corporate security, and developer products often had 3 different buyers not talking to each other. Pool of funds encourages experimentation.”— Sam Eden
Business & Entrepreneurship · Strategy & Decision Making
DUR_ENDURING
Flexible budget drives adoption
“Looking for product simplicity in complicated industries is a powerful signal. Cloudflare serves the whole long tail of the internet despite sophisticated technical infrastructure. Snowflake has a powerful engine under the hood that's hard to replicate, but the product is a simple query interface, very easy to use and adopt. Datadog is very easy to get started and set up, but powerful and flexible enough to work with the world's largest and smallest companies.”— Sam Eden
Business & Entrepreneurship · Technology & Engineering
DUR_ENDURING
Simple product, complex engine
“Project Honeypot was Cloudflare's origin: a way to let hackers scrape email addresses from websites, but the addresses were trackers. If someone tried to spam that email, they got added to a list of bad actors. Hundreds of thousands of people installed Honeypot to help build the list because they wanted protection from the list. The more users, the better the service, because the list got bigger.”— Sam Eden
Business & Entrepreneurship · Technology & Engineering
DUR_ENDURING
Free users build the moat
“Cloudflare's CapEx strategy: they invest behind the demand curve. They see where traffic is and what demand exists before they build. They're not building for no reason. All servers can run all products and provide all services, so the CapEx and ROI is split across all product lines. You're not building a separate network for each product. CapEx is 11-14% of revenue, which is high for software, but the ROI is diversified and higher because of the unified infrastructure.”— Sam Eden
Strategy & Decision Making · Business & Entrepreneurship · Operations & Execution
DUR_ENDURING
CapEx OK if ROI diversified
“Great companies have multiple levers of growth. They find ways to solve more problems for more customers over time. Cloudflare expanded product lines, expanded customer archetypes within each market, and did so in markets that all have tailwinds for AI. Looking for businesses with stacked growth drivers, not just one trick, is a durable investing framework. Each lever compounds on the others when the infrastructure is unified.”— Sam Eden
Strategy & Decision Making · Business & Entrepreneurship
DUR_ENDURING
Multiple growth levers compound
“When the outage happened, it wasn't a security breach or attack. It was a process error. Their bot management software had machine learning features updated every 5 minutes. An upstream error doubled the feature size, and servers didn't have memory for them. Every 5 minutes, a corrupted file pushed out and broke services. Not dissimilar to CrowdStrike's outage: a process error in the weeds, not a fundamental vulnerability. What customers appreciated was transparency.”— Sam Eden
Leadership & Management · Operations & Execution
DUR_CONTEXTUAL
Transparency during crisis builds trust
Frameworks (2)
Network Effect Flywheel for Infrastructure Businesses
Seven-stage reinforcing loop for building defensible data-driven networks
A repeatable framework for building compounding network effects in infrastructure businesses. The pattern starts with low-cost product-led growth serving the long tail, which generates data and signals that improve the core product, which attracts paying customers, which generates more traffic, which creates negotiating leverage with partners to further reduce costs, which enables reinvestment in the network. Each cycle strengthens the moat and makes the system harder to replicate. Cloudflare ran this loop for 15 years to capture 20% of global web traffic.
Components
- Low-cost infrastructure + easy product-led growth
- Aggregate long-tail traffic and data
- Use data to improve core product
- Better product attracts paying customers
- Enterprise customers bring more traffic
- Negotiate better terms with partners
- Reinvest savings and revenue into network expansion
Prerequisites
- Commodity hardware or low-cost infrastructure strategy
- Product-led growth capability
- Data collection and analysis infrastructure
- Willingness to serve unprofitable customers initially
Success Indicators
- Increasing traffic volume from free tier
- Measurable product improvement correlated with data collection
- Enterprise customer adoption accelerating
- Unit costs declining as scale increases
Failure Modes
- Charging for volume too early, limiting data collection
- Failing to translate data into product improvement
- Moving upmarket before product quality justifies it
- Building separate systems for different customer tiers
Unified Interface Strategy for Multi-Vendor Markets
Simplifying complex technical stacks through single-point interception
A framework for disrupting markets where customers must manage multiple specialized vendors. Instead of asking customers to split their traffic or workflow across different systems, build a unified layer that intercepts everything and provides all services from a single interface. This eliminates integration complexity, reduces vendor management overhead, and creates lock-in through simplicity. The pattern works when incumbents are locked into separate product lines and can't easily unify without cannibalizing revenue. Cloudflare used this to disrupt the CDN, DDoS, and web security markets simultaneously.
Components
- Identify fragmented vendor landscape with integration pain
- Build unified interception layer
- Layer services on unified infrastructure
- Serve long tail first with simple self-service
Prerequisites
- Technical capability to handle all traffic/workflow types
- Product-led growth distribution capability
- Willingness to serve long tail at low margins initially
Success Indicators
- Customers adding multiple services without new integration work
- Long-tail customer adoption accelerating
- Enterprise customers consolidating vendors to your platform
Failure Modes
- Building routing instead of interception (customers still make decisions)
- Requiring sales assistance for setup (blocks long tail)
- Services still requiring separate integration despite unified layer
Mental Models (10)
Innovator's Dilemma
Strategic ThinkingIncumbents ignore new markets or customer segments because serving them would cannibalize higher-mar
In Practice: Explicitly referenced by Sam Eden when explaining why legacy competitors didn't build what Cloudflar
Demonstrated by Leg-se-001
Reinforcing Feedback Loop
Systems ThinkingA cycle where outputs feed back as inputs, creating exponential growth or declin
In Practice: Core explanation of Cloudflare's moat structure
Demonstrated by Leg-se-001
Optionality
Decision MakingPreserving future choices by making decisions that keep multiple paths open rather than committing to a single path. Cloudflare demonstrated this by leaving empty motherboard slots in their hardware design without knowing what they would be used for. When AI inference became strategically important, they could simply plug in GPUs across their global network. 15 years of infrastructure planning paid off in one product cycle because they designed for unknown future optionality.
In Practice: Example of long-term strategic thinking enabling rapid adaptation
Demonstrated by Leg-se-001
Aggregation Theory
EconomicsPlatform businesses aggregate fragmented supply or demand to gain bargaining power. By serving the long tail of small websites, Cloudflare aggregated demand that ISPs were serving anyway.
In Practice: Explanation of ISP peering strategy
Demonstrated by Leg-se-001
Shared Fixed Costs
EconomicsWhen fixed costs can serve multiple products or customers, the incremental cost of each additional unit approaches zero and ROI is amplified.
In Practice: Cloudflare CapEx strategy serves all products simultaneously
Demonstrated by Leg-se-001
Invest Behind the Demand Curve
EconomicsBuild capacity in response to proven demand rather than speculative future demand. Cloudflare sees where traffic is before building infrastructure.
In Practice: Description of CapEx discipline
Demonstrated by Leg-se-001
Network Effects
MathematicsA product or service becomes more valuable as more people use it. Project Honeyp
In Practice: Explanation of Project Honeypot origins
Demonstrated by Leg-se-001
Disruptive Innovation
Strategic ThinkingNew entrants serve overlooked customer segments with simpler, cheaper products. Over time, the new e
In Practice: Explanation of competitive positioning vs. Akamai
Demonstrated by Leg-se-001
Removing Friction
Strategic ThinkingIdentifying and eliminating obstacles that prevent desired behavior. Cloudflare's pool of funds stra
In Practice: Explanation of pool of funds strategy and NRR impact
Demonstrated by Leg-se-001
Stacked Growth Vectors
Strategic ThinkingGreat companies have multiple independent sources of growth that can compound on each other. Cloudfl
In Practice: General investing lesson from Sam Eden
Demonstrated by Leg-se-001
Connective Tissue (1)
Postal service sorting factory as analogy for internet traffic management
Cloudflare's internet services are analogized to a postal sorting factory that intercepts all mail before it reaches a warehouse (website). The factory blocks junk mail (spam attacks), scans packages for malicious content (security), sets up local warehouses to reduce international shipping (CDN edge servers), and creates dedicated fast freeways for postal service (optimized routing). The analogy makes abstract internet infrastructure concepts concrete by mapping them to familiar physical logistics. The parallel works because both systems face the same fundamental problems: filtering unwanted requests, securing contents, optimizing delivery routes, and managing distributed infrastructure. The key insight is that internet traffic management is fundamentally a logistics and filtering problem that can be understood through physical-world parallels.
Sam Eden used this extended analogy to explain Cloudflare's services to a non-technical audience. The analogy was developed iteratively throughout the conversation to map each technical service to a postal/logistics equivalent.
Key Figures (4)
Matthew Prince
8 mentionsCo-founder and CEO, Cloudflare
Co-founded Cloudflare in 2009.
- Designed motherboards with empty slots for unknown future use cases
Mark Anderson
4 mentionsPresident of Revenue, Cloudflare
Michelle Zatlin
3 mentionsCo-founder and COO, Cloudflare
Lee Holloway
3 mentionsCo-founder and Technical Architect, Cloudflare
Glossary (5)
peering
DOMAIN_JARGONDirect interconnection between two networks without intermediary fees
“They can negotiate better peering deals with ISPs to reduce bandwidth costs further.”
serverless
DOMAIN_JARGONComputing model where infrastructure is managed by provider, not user
“Cloudflare Workers, a serverless function service.”
propagated
DOMAIN_JARGONDistributed or spread across multiple locations in a network
“You deploy a function immediately and it gets propagated around the global network.”
egress
DOMAIN_JARGONData leaving a network or system, often charged by cloud providers
“Their storage is 10 gigabytes per month with zero egress fees.”
inference
DOMAIN_JARGONRunning AI models to generate predictions or outputs from input data
“They've started to offer AI inference at the edge.”
Key People (1)
Sam Eden
Investor at Square Peg's Global Tech Fund
Concepts (7)
DDoS attack
CL_TECHNICALDistributed Denial of Service attack that overwhelms servers with coordinated traffic
CDN (Content Delivery Network)
CL_TECHNICALNetwork of distributed servers that deliver content from locations closest to users
reverse proxy
CL_TECHNICALServer that sits between clients and origin server, intercepting requests on behalf of the server
innovator dilemma
CL_STRATEGYIncumbents fail to adopt disruptive technologies because serving new markets would cannibalize existing business
peering relationship
CL_TECHNICALAgreement between networks to exchange traffic directly without paying transit fees
zero trust
CL_TECHNICALSecurity model requiring verification for every access attempt, regardless of network location
edge computing
CL_TECHNICALProcessing data near the source rather than in centralized data centers
Synthesis
Synthesis
Migrated from Scholia