TL;DR: Customers face friction when leaving — financial, procedural, or relational — and that friction is the moat. Switching costs don’t compete on product quality; they compete on the gap between what you’d need to pay to switch and what you’d actually pay. Lock customers in before they fully understand the math, and you can quietly raise prices for years.

What it means

Switching costs are the barriers customers must overcome to adopt a competing solution. They’re not about whether your product is better. They’re about whether moving to a better product is worth the pain. Enterprise software is the temple of this Power: migrate the data, retrain the team, reconfigure workflows, rebuild integrations, get re-blessed by IT and security. The switching cost isn’t just money. It’s time, risk, and organizational inertia — and any one of those alone is enough to keep a customer who would otherwise leave.

The real Power lies in when you capture the switching cost. A startup that locks in customers before they fully understand the total cost of switching enjoys a long window where price increases won’t trigger defection. By the time customers have visibility into how much they’re paying versus alternatives, the cost of leaving has compounded past the point where switching is rational. Timing is the asymmetry.

The argument

Three types, three different leverage points. Financial switching costs are the obvious ones — multi-year contracts, prepayment penalties, capital reinvestment in adjacent infrastructure. Procedural costs are stickier and less visible: loss of familiar tools, retraining burden, integration rework, the muscle memory of 200 employees who finally figured out how to use the product. Relational costs are the deepest of all: relationships between individuals at the customer and the vendor, organizational trust earned over years, habits baked into the daily rhythm of work. The best switching-cost positions exploit all three at once. Most products only ever build one and then wonder why churn ticks up the moment a cheaper competitor appears.

The takeoff window is critical and unforgiving. Early adopters don’t have comparison data; they’re solving a problem for the first time. Lock them in with contracts, integrations, team familiarity, accumulated configuration. By the time the market matures and competitors offer comparable features at lower prices, your customers’ extraction costs have become genuinely too high to bear. Miss the window and you become a commodity competing on price alone — which is exactly the position Thiel warns against (power-progression).

Switching costs seed other Powers. A strong switching-cost position often converts into network-effects over time (your data, your integrations, your community of users who train each other) or branding (customers who stayed become advocates and trainers). But switching costs alone are not durable in the long run — they degrade as switching technology improves (cloud migrations are an order of magnitude easier than they were a decade ago) and as competitors learn to build better bridges (every modern SaaS company has a “migrate from [competitor]” tool now). The wise strategy is to use switching costs as a temporary moat that buys you time to build a more durable one (seven-powers, power-progression).

The AI-era version

ChatGPT is building a switching cost that didn’t exist before LLMs: per-user accumulated context. Every conversation deposits memory, preferences, and trust that competitors literally cannot replicate without time travel. Victor Stepanov calls these relationship effects (never-go-viral-ai). They behave like switching costs in every way that matters — leaving means starting over with a new model that doesn’t know you — but they don’t appear in any of the classic three categories above. They might be the most durable switching cost ever invented for consumer software, because the user themselves is the source of the lock-in and they cannot be unbundled from it.

What links here