What value do the big model makers provide other than having a head start on gathering up humanity’s IP to train their proprietary models?

What’s their moat? Is it hoping for regulatory capture where scraping is made illegal the day after they finally finish scraping all human language?

It’s like OpenAI dammed the Colorado, and Anthropic dammed the Hudson, and now they’re both trying to sell us bottled water subscriptions at $100 a month. I don’t know how well the dam part of the analogy holds up, but the water part feels strong. Compiling models based on humanity’s written output feels like something no corporation should own.