At the moment very hyped industry but it still seems under valued. When ChatGPT released their first great LLM model, people thought that the tech was cool and helped for example in writing but they didn’t see how current type of agentic workflow with even better models could distrupt many industries. I believe we still see small part of the whole picture and there is so much more we can get out of these technologies.
On high level I believe AI to take a lot of important roles in companies. My bet is that in a few years companies will produce more with AI than with human workforce which mean they might pay more for AI than humans. This obviously is the biggest market in the world.
The biggest risk I see in AI right now is that the companies aren’t able to capture all the value but countries realize that they lose tax payers and need to cover it with something. They start to regulate and want their share of the profits. The first companies are the ones who provide these AI services. The frontier labs.
The other risk with frontier labs is that they compete with copmanies who provide nearly as good open source models. These risks make me less interested investing in them and more interested in investing the underlying hardware.
No matter if company buys AI service from 3rd party or runs their own models, they need compute. At the moment memory seems the biggest bottleneck. No one company has moat over it so the production will increase at some point. The demand is just so high that the first movers are going to stay ahead.
ASML feels like the obvious final boss of AI. No one has been able to build machines as good as theirs, they have so much money coming in that they can invest research and stay ahead.
My vision of AI in 10 years is that LLM models have been engineered to work in many other domains and we will have new type of algorithms (e.g. Reinforcement Learning) applied to many more problems like robots. Compute is going to be so important that the current prices seem so under valued.