The last 2 weeks can predict the future of AI

Hey

We’re being told that AI is either inevitable, or that it's a bubble.

But what if both are true ?

A few years back, we were all getting claims that AI was going to hit a roadblock.

That would require more and more computers to make incrementally better LLMs.

But that hasn’t really happened. While more computers has been necessary, nothing on the scale of what people were predicting has been necessary. Instead, many smaller Chinese startups have shown talent is more important than the sheer size of computers and data.

In the last 2 weeks, various AI companies have launched models; Meta, moonshot, OpenAI, Deepseek, etc. Companies have also slashed prices, and reset limits for users.

This all shows 1 thing :

AI is world changing, but selling LLMs is not a great business model.



My main reason to believe this is that the cost of intelligence is trending to 0.

What does this mean? Well, basically it means AI models are getting cheaper and cheaper while improving in performance.

You might argue that this reduction in price is subsidized, and it is in many cases, but many recent open source models represent genuine advances.



So what does this mean for AI companies?

First, Meta, Amazon, Google, and Microsoft are pretty much guaranteed to be fine. I also think that while SpaceX is overvalued as a stock, as a company it's going to be fine.

Well, then that leaves 2 big groups, first the Chinese AI companies, and second, OpenAI and Anthropic.


Out of these, the company I think Anthropic and the others have a chance, but the one that’s pretty much guaranteed to fail is OpenAI.

Subscribe to learn why I think this in my next post.