It would tend to agree with the energy costs. The amount of processing power to generate an image of the type presently being used is minuscule compared to the constant power being drawn in for things like cryptocurrency. It also depends on what is being used to generate the power as to what byproducts could be affecting the environment. The abstraction level would make it kind of difficult to pinpoint what those costs are when you are generating a job that takes maybe a minute to make out of the existing load of thousands of other jobs the same computer is running and is made to do efficiently.
As for the art itself, AI’s can manipulate images with a precision to the pixel. A human artist with photoshop can do the same. The human artist may take longer to do it in but they will likely have less mistakes with the results. The human artist will likely do a better job and require less iterations of the art until it is done because the artist can self-correct along the way and has a better “understanding” of the intended results. But like others have said sometimes there is a lack of human artists that are willing to take the commissions. People often can only do so much in a period of time and they might be saturated with work or they are in a position to pick and choose what they “want” to make. The AI’s have no such issues. Their biggest limiter today is the lack of good training and frankly the chains people place on them. Most AI’s are purposefully hobbled so they cannot generate NSFW work easily, if at all. They can do portraits and body images, but they have to be still PG rated because other humans placed restrictions on them. Even if you can get one that shows nudity they are usually so badly trained as to make them almost pointless for results and require an artist to edit what it did do to finish the image. Maybe someone made one out there that is trained fully but so far all the ones I have tested came up very poorly.
Theoretically though AI’s could allow for scenes that may be outside physical limits but still appear as if it was real. This could allow for an author to imagine a scene and bring it to life. They might be able to find an artist to do the same but finding one with the style the author wants and is willing to do it would be a factor.