Vertical Integration: Owning the Step That Decides the Product
Buying from specialists is usually cheaper, and sometimes it means you cannot be better than anyone else. The right boundary moves as an industry matures — and rarely gets revisited.
Buying from specialists is usually cheaper, and sometimes it means you cannot be better than anyone else. The right boundary moves as an industry matures — and rarely gets revisited.
Ten times the users gives roughly a hundred times the possible connections. The exact formula overstates things — the underlying point about compounding leads does not.
Species with identical needs cannot coexist — a small consistent advantage compounds until one is gone. Which means anything living side by side must differ, and finding how is informative.
Chip design is spread across hundreds of firms. Making the advanced ones happens in very few places — and that single asymmetry explains the shortages, the politics, and the fragility.
Profitable companies spend years building things and then hand them out for free. There's a clean piece of economics behind it — and it tells you where you sit in someone else's plan.
Retailers won't join without shoppers; shoppers won't come without retailers. Both are right to wait — and how you break that deadlock defines every marketplace business ever built.
Two shops on one street do everything right; one closes. Quality and effort are what you need to compete at all — a moat is what stops a rival matching you even when they genuinely try.
A newspaper needed presses and trucks to control its market. An aggregator controls a bigger one with neither — because the internet moved the choke point from distribution to attention.
Incumbents rarely lose because they were careless. They lose because listening to their best customers, protecting margins, and ceding the unprofitable low end are each individually correct — and collectively fatal.
None of these companies owns most of what it sells. They own the place people begin — and that turned out to be the most valuable position in the economy.
The interesting number in AI is not what a model can do. It is what a unit of machine reasoning costs — because when that falls tenfold, whole categories of product become possible.
Before the cloud you bought servers for your busiest hour and let them idle the rest of the year. The change was financial rather than technical — and that is why it mattered so much.
The graveyard is full of competent products people opened once. Downloads measure marketing; the second visit measures whether anything was actually solved.
How the money arrives is the least interesting part. Every model has a lever for growing revenue and a cost attached to pulling it — and where that cost falls predicts behaviour under pressure.
Two firms selling something interchangeable can only compete on price, and every improvement gets matched. The useful question is not how hard to compete but what you are competing on.
A rule needs a harm that is noticed, documented, understood, and politically salient. All four take years, while products change in months — and that gap explains most of the field.
A century of competition law detects monopoly by watching consumer prices. When the service is free, the test finds nothing — and the power sits somewhere it was never designed to look.