Is AI performance inherently limited since more and more intelligent models are constrained by energy requirements?

TLDR: AI infrastructure requires tremendous amounts of energy. Smarter AI models to do human jobs will require even more energy, likely limitless energy powered by nuclear power plants. People will be (mostly) unwilling to set up more nuclear power plants dedicated to this. Thus, AI models aren't likely to take many higher level IT/ software engineering jobs because of constraints on the energy requirements of AI infrastructure.

The general thought process of the average person is that AI will get extremely intelligent and render most software jobs completely useless. The expansion of AI has already started encroaching on software jobs at basic levels, but it will eventually reach all levels and positions.

The issue is, increasing AI "intelligence" and effectiveness is inherently limited by energy requirements. The more "intelligent" the system, the more energy is required.

Energy optimization of the energy requirements of AI models has already basically peaked out. Models are incredibly efficient using our current levels of technology / the current levels of technology available. ​The only greater optimization using current energy would actually come from using nuclear power.

Nuclear power plants dedicated to AI infrastructure will have to become a reality with more "intelligent" AI models that can replicate human decision making patterns at a higher level. The real issue is that it's not likely this will become widespread because of the energy required and the lack of willingness to dedicate nuclear power plants for AI model usage.

What are everyone's thoughts on this?