Oh it will.

Firstly, neither OpenAI nor Anthropic is profitable, by a wide margin — investors are going to get impatient at some point.

Secondly, people that aren't enthusiastic about this whole thing are already experiencing something of an AI fatigue with all the AI features violently shoved into them by most software products they use. Being involuntarily subjected to slop in various online spaces can't be good either.

Thirdly, remember NFTs? So many people swore they were The Future™... until they weren't. But at least in that case it was much more obvious how stupid the whole idea is. The scale of the hype was also several orders of magnitude less.

Google can run the models profitable, thanks to their custom tpus. The rest? No idea

Google and Meta can both subsidize their AI divisions with their ad money.

That too, but I think someone said that Google AI cost is true costs. Unlike others.

My PC can give decent code recommendations locally, not relying on anything in the cloud.

Even if all major provides close down, it doesn't remove what's already out there. Glm / minimax / deepseek / gpt-oss may not be at the same level as current frontier, but you can download them and they're still very capable.

Crazy stupid ideas like cars with only touchscreens have still taken a decade to come in and then to get considered ill-advised even though anyone driving a car could tell how bad of an interface it is. We are still not fully out through the other side.

So while OpenAI or Anthropic are maybe not profitable today, they've got at least 5 years to figure it out. And there is already talk of inserting ads into the "chat", but hopefully that does not work!

But really, LLMs are useful (yes, sometimes only in appearance, but sometimes for real), and with that, there will continue to be investment into them until they are made profitable.

new != good

therefore

new != better