MSP_Logic
The first is that our current power grid isn't sufficient for the projected needs of AI. Forget water- most of the water used by the AI is used by the power plant, so really, it's just a proxy for electricity. There's lots of ways this worry can be addressed- newer training methods that use less power, new chips that use less power, long-term building new power generation (hopefully low or no carbon). But the people with the money need to actually acknowledge the problem and look to ways to solve it.
The second is the impact on jobs. This is personal, as while I'm pretty sure my own job is safe, lots of people around me aren't. And just because I'm pretty sure, doesn't mean I'm 100% sure I'm safe either. Sure, technology has always caused jobs to change- I'm not different from a buggy whip manufacturer when the automobile came out. I think a lot of people feel this way, from artists to programmers to taxi/uber drivers.
Sure, new technology does this all the time, but the sheer scale and speed of what AI can do means we're impacting more jobs than ever, on a much shorter time scale. A lot of people are scared.
I don't think we need to stop all AI work or anything like that- but we need to understand and address the various second and third order effects that AI is going to have, and we need thoughtful plans to address it.
And for the record- if any AI companies fails, I hope it's OpenAI. Anthropic is pretty mild in comparison, in my world view. :-)