And Then, There Was One

So this was an interesting LinkedIn post:

My AI vendor upgraded the terms of service.

For 12% more cost per month, I will get only 60% of the monthly credits for my use. Less for more. Sounds like a great upgrade... for them.

Time to drop my provider, I guess.

Are the costs finally reaching the consumers now? Have the AI vendors run out of free money they can use to keep costs artificially low?

The last sentence stood out for me in in the context of the third line; this person believes the time has come to drop their generative automation provider, because that provider can, apparently, no longer afford to keep compute prices artificially low. This is the genesis of the the process that Cory Doctorow terms "enshittification." As companies find themselves in the position of needing to show a profit, they are forced to raise their prices. The resulting flight of their customer base hobbles them, they enter a death spiral, and are either forced to exit the market or are acquired by another player. The number of providers shrinks until the overall market is highly consolidated, and then prices start rising to repay investors for the subsidized costs that have lured in customers all this time. And because all of the remaining players are raising their prices (or there are no other remaining players), customers have nowhere else to go, even if lock-in effects aren't present.

And since all of the market players are now very large, new entrants to the market are at a sizable disadvantage... in order to compete, they'd need to have unbeatably low prices, which would require a large amount of up-front investment... which would need to come from people and organizations that aren't bought in to the current market principals. So there's nowhere to keep moving to for investor-subsidized services forever. And. of course, investors realize this; the whole reason they're subsidizing costs for customers at this point is the expectation that once the market consolidates, they'll make all of it back, and more.

One commenter noted that State of the Art (SOTA) models based in China have very low costs, but without insight into their economics, there's no way of knowing whether those prices are artificially low, themselves. Nothing prevents the government of China from attempting to ensure the Last Provider Standing, and thus the one that benefits from being able to squeeze customers, is a Chinese company. But even if they aren't, if they can force other nations to take over the role of subsidy provider, that's still a benefit to them.

Either way, this mindset of "not artificially low is too high" is being penny-wise but pound-foolish. The unwillingness to pay what one believes that things actually cost out of a pervasive sense of one's own poverty has never ended well. And while this time might be the first, I don't see much investor money backing that position.