Do you have any links to further reading on this. I haven't read anything like this and also couldn't find anything in their newsrooms. The only thing I read is that they want to have more people work on implementations.
Say OpenAI builds GPT-11 and they realize it can do 95% of their clients work. Why not hire away key people from the client or the client's competitors and start eating into their market share? OpenAI would realize if GPT-11 has this ability then Anthropic is probably able to do the same thing. The company that has the most money and credit available is going to be able to buy more compute. OpenAI and Anthropic cannot slow down unless mandated to by the government.
Can you give me an example where this would make sense, because I just don't see e.g. OpenAI starting a grocery store chain or Anthropic building a wind turbine manufacturing company.
What I can see is that some company's products become useless, e.g. translation tools, as it is something that you can just do within ChatGPT or Claude.
For sales I'm sorry, but I don't think sales companies exist? Sales is usually just a part of another company. Some of those employees might be made redundant as AI products complete part of their work, but that would also happen through OpenAI's or Anthropic's products.
So for both of those I don't see what kind of company they could spin off (excluding the example of a company that does AI implementations as I mentioned that one in my initial comment) that would compete with an existing company. For all of these things offering a product that allows existing companies to replace employees sounds like the more scalable, profitable and flexible strategy.
Also, in such a future businesses would be reorganized to use AI. That probably means less employees and reduced human complexity. The business processes are further abstracted from people making it easier for a big AI lab to move in.
If I was looking at businesses today where the big AI labs could step in I would look for businesses where they earn high margins based on intellectually complex tasks. Maybe the most pure example would be trading in the stock market since it can be completely done by a computer. Then you have areas like law which process large amounts of data but are also heavily based on relationships.