Expedition I

For four years we had been describing artificial intelligence to our LPs as a data-hungry monster, and drawing the consequence as a loop with three beats. AI improves. Improvement creates demand for data the internet cannot supply. That demand puts sensors and connected machines into the world, which improves AI again.

We went to check the third beat. Not whether the machines existed — whether they were being deployed at the scale the argument required, at a cost somebody was actually paying, in industries that look like the ones our portfolio operates in.

We came back having added a fourth.

One of the stops was Unitree. At the time the quadrupeds were the business and the humanoids were research and entertainment, and our host was blunt about it: the machines were still following a path a human had drawn for them.

Since that visit, Unitree's humanoid revenue overtook its quadruped revenue for the first time. Its STAR Market IPO application was accepted in the same stretch. Our delegates had already walked the floor, met the team, and asked their questions.

That is what an expedition buys. Not information — information is free and late. Timing.