Skip to content

Comment on SK Hynix plunges after Nasdaq debut as memory chip euphoria coolsparent

Comments

I'm sure they've got a whole finance department devoted to studying it, so they know better than we, the general public do. $800 to $5000 is quite the spread. Frontier labs have competition with each other and China and third party hosting of open weight models. There is no doubt that we're at the $1 Uber ride point on the timeline, but Uber finally managed to be profitable. Advancements in hardware and optimizations in software, along with patient investors mean that when they do have to start making money, it's not guaranteed that they'll sink. It's not guaranteed that they'll win, either, but I'm not an investor.

The unspoken difficult to talk about detail is so what if it costs $5,000 a month? That's someone making roughly $23/hr. If businesses could get an AI employee that was as capable as someone making $23/hr that was never sick, didn't complain about working conditions, was sufficiently competent, would work 24/7 when asked, could be cloned on demand, could be fired the second the task was done; companies wouldn't have a problem paying $5,000/month, they'd probably pay twice that since they don't have to pay for other stuff like health care or holidays or HR or an office for that AI employee.

You can keep paying your employee $23 right up until they start agitating for a raise (many never do), and often for some time beyond that.

Your AI employees will all cost more each year and you will have very little leverage over them.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.