If AI revenue is $15,000/year per job replaced, then it'd need to replace 40 million jobs.

I think this is the wrong way to think about it. Replacing people isn't what makes AI exciting. It's about enhancing human productivity. The real question is how much productive value can be gained. Can we increase the productivity of 40 million people by $15,000 a year? Absolutely.

Even if AI increases productivity dramatically, only a small portion of that is going to get captured by AI companies. For example, a lawyer could use ChatGPT for very little, and then maybe the value is gonna get split between the lawyer, their firm, and their client. Perhaps, it ripples down the value chain.

I believe the economic value from increased productivity will be massive, far beyond what we're currently estimating. Sure, some companies, especially those creating "me too" products or niche models that big players like OpenAI will eventually cover, are bound to fail. But the upside here is in the trillions, not just billions.