A problem arising for AI companies?

I'm currently looking into AI developments and the usecases for it in companies I research. From the technology development at the moment I think we may see a hard competition and contraction in the AI space because models become more effective, auto improvement scales and we as consumers will be able to run a lot of the models on smaller and smaller hardware. Basically compute gets cheaper and cheaper and capabilities go up so you don't need a company managing your data in a data center for you, local becomes good enough. The models will become good enough for private person to use them reliably without a need for a data center there's no physical border I know of at the moment that limits this. There is a long way ahead on the road from my perspective and understanding. Here is a example for implemented self improvement from Minimax. I'm really curious how that plays out in the long run and if AI just becomes a commodity and benefit people and most of the AI companies will just vanish. From what I read it's going in multiple directions smaller and more capable models, auto improvement and optimization of hardware on the silicon level to optimize for specific models. I'd like your opinions and perspectives on this.
https://www.minimax.io/news/minimax-m27-en