The AI investing market seems confusing with a recent stock market pullback in big tech names, over worries about AI spending, while private AI companies like Cursor raise at massive valuations. Which is it? Is there a bubble or not? Yes. And No. (View Highlight)
My #1 rule for thinking about AI, since it is “intelligence” is to related it to people. When we think about how we do things in AI, I always revert to “how do we do that thing with people?” as a starting point. What this means from an investment perspective is that unlike previous innovations which replaced a specific piece of human labor, AI is replacing many pieces. It’s closer to the invention of the steam engine than say, the loom or the tractor. It has the ability to replace many parts of the human labor force. (View Highlight)
The reason it feels jagged, is that some of those parts are easier and more readily replaced by AI than others. The confusion you are feeling isn’t whether AI will work or see a return or not - its what is automated first and what is automated later and what is automated last? (View Highlight)
When you ask the question about the ROI on AI, it depends what you are doing with it. Some tasks are about 100% replaceable, others are 100% replaceable but with some human supervision, others are 60% replaceable, others 20% replaceable, and so on. For a long time coming, AI’s path through the economy will be “jagged.” (View Highlight)
To answer the question about a bubble - on the tasks where AI is 100% ready or, will be 100% automated in the next 2-3 years, there is no bubble. On the tasks where it will take longer, yes there is a bubble as the near term and mid-term ROI isn’t there. As you look lower down the stack, this is why many investors are so excited about the component layers like chip manufacturing and networking. Even if the current version of AI is slowing down as LLM progress seems to stall, there are technologies coming behind it. Some of them will provide new capabilities. Some will lower the costs of the existing capabilities to expand their use cases and TAM. And edge AI is coming and will suck up much of the capacity of the chip industry if data center slows. (View Highlight)
There will be some small pullbacks and bumps along the way but, here is how I would think about playing them.
In public markets, your AI winners are going to compound at exceptional rates against everything else for many years. Stay in them and buy on pullbacks.
In private markets, the AI whole is going to be so big that even niches are massive. If a “niche” AI market is 4% of the whole market but the whole market is 2trilliondollars,evenifthehyperscalersownthemajorityofitthat’sa40B market for the niche. The niche market can support a few unicorns.
The infrastructure to run all of this has not been built. Even the current infrastructure winners in AI will be sometimes displaced and sometimes re-segmented by narrower competitors. Plus new capabilities are emerging that will need to be supported. (View Highlight)
