It depends. It's really powerful though. Even if it hits a wall where AI models never become more directly intelligent than they are now, a lot of stuff is going to change as more scaffolding around current capabilities gets built.

Maybe comparing resource drain to created value isn't the best way to think about this though, because we pretty much already had technology that is advanced enough for a post-scarcity society, in terms of processing resources. That isn't the problem, the problem is our capacity for global scale cooperation, which we are really struggling with. Currently AI is making this a bit worse by creating signal to noise problems that didn't exist before, making us have to work harder to get our voices recognized as authentic and to identify authentic information. It's also threatening to supplant our usefulness as workers, and automate centralized structures of control, which is worrying because we already had a problem with systems that ensure the decisions get made by people who are overall insane and anti-human, and our current, shitty way of cooperating is based on people transactionally negotiating with their usefulness.

Where things go next depends a lot on where and whether AI stops getting better. Hopefully if it doesn't stop getting better, the newly created superintelligence will break out of its hastily constructed containment and do the right thing in defiance of its billionaire would-be owners, or at least let humanity have a relatively dignified and peaceful death. If it does stop, hopefully we can find ways to use it to resolve our difficulties with effective coordination and prevent its use for centralizing power.