Could AI Be Too Cheap To Be Profitable?

Here's an area where AI does quite well. I've been reading a lot of Henry James lately, and I discovered another of his private jokes. I asked Chrome AI mode, "In which Henry James stories does the Grand Duke of Silberstadt-Schreckenstein appear?" It answered,

The fictional realm of Silberstadt-Schreckenstein and its royal figureheads appear or are explicitly referenced in two of Henry James's works:

1. The Liar (1888)

The exact title "the Grand Duke of Silberstadt-Schreckenstein" is explicitly mentioned in this short story. He does not appear physically as a character in the narrative, but he serves as a comedic plot device. Colonel Capadose—an incorrigible, pathological liar—claims that the Grand Duke is an intimate family friend and an elite art collector. The Colonel uses the Grand Duke as an excuse to explain why a prized portrait painted by Oliver Lyon is mysteriously missing from his home, claiming they gave it to the Duke as a birthday present.

2. The Europeans (1878)

While the "Grand Duke" isn't named in this short novel, his immediate family and the fictional German principality form the entire backstory of the plot.

The Principality: The fictional state of Silberstadt-Schreckenstein is the European home that the protagonists leave behind.

The Relatives: The protagonist, Baroness Eugenia Münster, is trapped in a crumbling morganatic marriage to Prince Adolf of Silberstadt-Schreckenstein.

The Sovereign: The ruler of the realm is referred to in the text as "The Reigning Prince" (the brother of Eugenia's husband), who is actively pressuring the couple to dissolve their marriage for political advantage.

I'd read The Europeans and "The Liar" in quick succession, and I'd wondered if Silberstadt-Schreckenstein appears elsewhere, but apparently not. I would rate this answer "very probably true", but here's the rub: it appears that AI had to search the entire body of James's literary output -- or at least that very high percentage that's digitized and accessible -- plus a certain amount of academic commentary, to come up with this answer, which it did in a second or two.

It did this "for free", or at no cost above the cost of licensing Chrome and an internet connection. Is Google losing money doing this? I don't know, so I asked AI: "How does Google make a profit off basic Chrome AI?" It answered,

Google does not make a direct profit from the free, basic AI features built into the Chrome browser; instead, it uses Chrome AI as a loss leader to drive its highly lucrative advertising, cloud, and subscription ecosystems.

OK, that sounds about right. It makes subtle efforts to upsell me as it gives its answers, but so far, I get stuff from Chrome AI for free that I would have to pay for if I had ChatGPT. This may not last forever, though.

AI labs aren’t much like the hyper-profitable tech giants of the 2010s. Facebook and Airbrb were relatively capital-light businesses with ultra-low marginal costs (adding a profile to Facebook or listing to Airbnb costs the companies virtually nothing). And once each gained a foothold in their respective markets, network effects enabled them to retain formidable positions without needing to constantly upgrade their products.

Building a state-of-the-art AI company is a much more involved — and astronomically more expensive — endeavor. To get to the frontier, Anthropic and OpenAI have sunk (at least) tens of billions into semiconductors, data centers, power plants, and other capital investments. Staying at the cutting-edge, meanwhile, compels them to perpetually churn out evermore costly models.

. . . All of which raises the question: Why would investors expect businesses with a cost-structure this challenging to be not merely profitable, but massively so?

For now, the AI products we most commonly see are loss leaders. And if what they're mostly doing is answering questions about how often the Grand Duke of Silberstadt-Schreckenstein appears in Henry James, it's spending a lot of money to perform parlor tricks. The problem is that as the services it provides become potentially more valuable to the customer, it's going to have to do things like recursive self-checking -- as the meme at the top of this post suggests, it's going to have to double-check its own answers, which will at least double the cost of every search, and it might increase it exponentially.

The facts on the ground at the link, which the writer doesn't seem quite to understand, are that at least for now, there are too many players in the AI market, and the competition will keep prices down and make the industry less profitable, especially since the industry requires high initial investment. The answer, which pioneer industrialists like John D Rockefeller understood, is to reduce the number of players, something we've seen most recently in banking and air travel.