· Enterprise AI · 2 min read
The Quiet Winners of the AI Boom
The real winners are not the most obvious names, but the ones deeper in the stack where the moats are unbreakable.
Two things I’m certain about AI: (1) In the long term, AI will have a profound impact on the whole world economy. (2) It takes time for AI to spread to every corner of the world economy.
This means currently there is an exuberance hyping up prominent players in the whole AI chain. Some names will not live up to short-term expectations. However, if you look at a longer horizon, 5-10 years, big winners will emerge.
The whole AI chain is large. Here is a very simplified line:
AI app (e.g., Cursor) <- AI labs (e.g., OpenAI) <- AI cloud (e.g., Azure) <- AI chip designer (e.g., Nvidia) <- Chip manufacturer (TSMC) <- Chip manufacturing equipment (ASML)
There are many adjacent businesses in that AI chain too, from Ethernet cable manufacturers to data center land providers.
The biggest winners will not be the obvious ones you see in headlines every day (OpenAI, Microsoft, Google…). Why so?
Say in 5-10 years, the AI boom will expand AI demand by 10 times (just a made-up number). This means that every node in the chain will expand roughly 10 times. There are some nodes with fierce competition, no clear moat, and low entry barriers (e.g., AI labs, cloud, or even chip designers). Incumbents in these nodes will have to fight each other and fight newcomers too.
There are other nodes where the barriers and the moats of incumbents are so high that an outsider entering is hopeless. One example outside of tech is whiskey makers. If you start a whiskey-making startup, you have to wait 12-18 years at least for the whiskey aging process before you have the first sale. That is why there is virtually no new entry in this market.
In the whole AI chain above, there are a couple of companies with such invincible moats; they are positioned to capture the whole 10x expansion of the AI boom if they execute right. Some of them are, for example, ASML, the Synopsys-Cadence duopoly, etc. There are potential challenges such as geopolitics, but that would affect the whole AI chain anyway.
I’ll revisit this post in 5 years to see if my prediction is true or not.
