- Network effects no longer guarantee defensibility because switching costs have collapsed.
- Fragmentation, APIs, and behavioral saturation reduce the marginal value of additional users.
- This erosion creates the opening for memory-based moats to replace connection-based moats.
The Classic Network Effect Story (Now Outdated)
For two decades, the dominant platform logic looked like this:
- Each new user increases value for every other user.
- More users = more data = better experience.
- High switching costs produce winner-take-all outcomes.
- Early scale becomes an unassailable moat.
This model powered Facebook, LinkedIn, Google, and the early SaaS platforms.
But the underlying assumptions have changed — and the moat is weakening.
Force 1: Multi-Homing (Switching Costs Near Zero)
Users now maintain simultaneous identities across platforms:
The cost to join or leave is:
Social graphs used to be sticky; now they’re portable.
This destroys the exclusive advantage of “having everyone in one place.”
Force 2: Platform Fragmentation (Density Beats Size)
We’ve moved from general-purpose networks to niche, high-density communities:
- Creators cluster by domain
- Professionals cluster by vertical
- Communities self-organize into micro-networks
Users prefer relevance over scale.
Quality of interactions beats quantity of nodes.
A network with 10,000 highly aligned users outperforms a network of 100M generic users.
This flips the old law that “bigger is always better.”
Force 3: API-Mediated Access (Value Without Being On-Platform)
Users no longer need to live inside a platform to extract value from it.
Aggregators and APIs shift the relationship from:
- “I must be here to use this tool”
to - “This tool plugs into whatever workflow I already use.”
Examples:
- Calendars sync across apps
- Payments abstracted through Stripe
- Content pulled through embeds
- Agents using APIs instead of native interfaces
If the value can be accessed externally, the platform loses its moat internally.
Force 4: Declining Engagement Returns (Value Saturation)
The old curve assumed marginal value increased with each user.
Today, the opposite happens:
- the 10,000th connection isn’t more valuable
- the feed gets noisier, not better
- recommendation algorithms flatten
- engagement saturates
Beyond a certain point, scale adds friction instead of value.
Traditional network effects turn into diminishing returns.
The Result: A Weakening Network Effect Curve
Where the old curve showed:
- steep climb to critical mass
- runaway value creation
- a moat that strengthened over time
The new curve shows:
- early value
- flattening returns
- easier bypass via multi-homing and APIs
- fragmented demand pools
- lower switching costs
Traditional network effects still work, but they age quickly and are no longer sufficient to defend a platform.
The Moat Is Strong but Finite — And a New One Is Emerging
Connection networks (traditional platforms) are losing out to memory networks (AI-native platforms).
Memory networks create:
- compounding personalization
- reasoning patterns
- workflow entrenchment
- switching costs that rise every day
- non-transferable intelligence
- exponential defensibility
Unlike traditional networks, which rely on external participants, memory moats are:
- internal
- cumulative
- persistent
- personalized
- non-fragmentable
- impossible to multi-home across
This is the fundamental economic shift AI introduces.
Traditional networks defended access.
Memory networks defend improvement.
Traditional networks plateau.
Memory networks compound.
This is why the era of network effects as the primary moat is ending — and the era of memory effects is beginning.
More Resources
