AI Expansion 2026: vs. Internet Cycle Explained

Many people ask me about view on where are we in AI, and shouldn't it be time to see Internet-like boom and bust due to limited monetization? Here are some food for thoughts -

Why AI can continue expand into 2026?

Most believe 4 major CSP plus Oracle and xAI, will be the major AI cloud service providers, while there could be 1-2 sovereign fund that also wants to develop domestic AI cloud, so aggregately max is 8 cloud providers. However, when we talk with component makers, key US component makers suggest there could be a dozen cloud providers, which dont even include Chinese CSP. There are more bigger whales than most of investors (including me) thought.

This explains why component makers see backlog continue expanding in 2Q25, despite we are already 2.5yrs into AI build since beginning of 2023.

How different is AI vs Internet cycle?

In AI cycle, CSP invest due to Greed + Fear, while more enterprises, especially digital-native and software companies are investing with Fear of cannibalization in very swift time frame. This is evidenced by rising Capex yet lowering OPEX. 

While in Internet cycle, most companies invest due to Greed at initial stage, as small emerging companies can possibly change the game with Internet. Big companies like Yahoo AOL wasn’t aware and wasn’t in Fear mode. This is evidenced by rising Capex with rising OPEX. 

The power of Greed + Fear is stronger than Greed alone. That’s why we will see a bigger cycle vs Internet. Internet investment was 3-3.5yrs, and AI likely would be longer.