Korva is a scroll of the internet’s best ideas.

Early stage of the curve

Anthropic has something like 14 or 15 million daily active users, and probably a small portion of them truly use AI the way you can. The first stage of an S-curve is the tinkerers, then come the early adopters, then the early mainstream. Having Claude on your computer, linking it in, and building skills comes first, then true AI bots, and big corporations will build much larger ones.

The projected path

The projected path runs from 10 basis points to one to two or three percent, then to five percent, then to 15 percent in the next four years. This year a light switch went off in the enterprise, and everybody realizes they need to do it now and do it fast. It resembles needing a website in 1998, only this is coming together fast.

Sold-out compute

Enterprise application AI is less than 1 percent penetrated, and the compute is already sold out. Anthropic has half of the compute it needs right now, before this takeup arrives.

Why can't compute expand quickly?

A new data center takes one to three years to build, and the power connection can take longer. Advanced chips also depend on a few specialized factory steps, such as the advanced packaging that joins processors to memory, where capacity is limited. Each of those links expands on its own schedule. Capacity ordered today arrives well after the demand it was meant to serve.

How fast did the web reach most people?

The Pew Research Center found that about 14 percent of American adults were online in 1995. By 2000 roughly half were. Companies felt pressure to have a website well before most of their customers used one. The late 1990s saw heavy business spending on web presence.

Why do technologies stall before the mainstream?

Geoffrey Moore's 1991 book Crossing the Chasm argued that a gap opens between early adopters and the early majority. Early adopters tolerate rough products and incomplete features, but the majority wants a complete solution that works. Companies often have to change how they sell to cross the gap. Moore built his idea on Everett Rogers's earlier model of adoption stages.

More from Invest Like the Best with Patrick O'Shaughnessy

Business & StartupsThe 30 to 40 percent wallWhen a technology gets to around 30 or 40 percent penetration, exponential growth stops, the sell side catches up, and there are no longer big beats.
Enjoyed it?