Why would companies like Google, Meta, and Microsoft want to invest so much in AI infrastructure? Wall Street looks at the rise in capex as a threat to future profits. Many, myself included, wonder whether all their AI hype has proved is that if you give something away, people will take it. So, do these companies think that some actual willingness to pay will emerge for some form of AI? Are they just hoping, or is there a chance that all this will pay off?

One thing that seems clear is that AI results in searches will reduce the number of times users click links, either by giving them an answer without clicking or by reducing the number of links they review. This piece in TechCrunch gives some interesting data on that. It shows that while AI referrals to websites are up, which means AI is generating ad revenue for the search provider and leads for the site owners, they fall far short of normal search clicks. I suspect most of us will agree; I think I rely less on AI summaries than most, but even I recognize that AI may help me refine my interests so I click on fewer links. Why, then, would the search giants want to invest in AI?

One good reason is that if AI gets a lot of hype, which it surely has, consumers might switch away from a search engine that didn’t offer it. Losing all the click revenue is worse than losing some by adding an AI summary capability. But that doesn’t address the longer-term risk. If AI summaries reduce net clicks, then ad revenue for the big search vendors is impacted.

Might that be mitigated through targeting ads better? Could AI perhaps provide click-through links that are more likely to be clicked, thus creating more revenue per click? Could advertisers pay more for these AI clicks? That’s possible, but on the surface it’s hard to see how it would work without having AI summaries bias their content to favor paying advertisers, a form of sponsorship that borders on (or even crosses over to) lying for money. It seems like this is something that consumers would catch on to, and more than that, something that would be exposed by investigative sites or even submitted to regulators as a complaint.

What about Meta, though? I think they open a deeper question, which is the inherent issue with ad sponsorship as a revenue source. Advertising is a zero-sum game in a very real sense. Global adspend tends to rise only slightly faster than inflation, so for any player whose revenue is sponsored by it, you can only gain by having something else lose. Print media was a rich source of loss, and so to a degree is mass mailing, but in the long run, all these players have to realize that they’re fighting for a limited number of dollars, dollars that cannot possibly grow fast enough to keep their own revenue growth at the pace Wall Street will demand.

What could AI do here? OK, one obvious answer is “transition to a service people will pay for”. Some do pay for AI, but could we expect that base to expand, given that the Internet has tended to whet consumer appetites for “free” stuff, meaning of course stuff someone is sponsoring? There is willingness to pay for content, meaning video and music, but would it extend to AI? I think that’s an open question, so maybe we need to look further into what AI might do, and see if it could generate some incremental-to-advertising revenue, or at least change the consumer decision process enough to open new ad channels that could give some early mover an edge.

I’ve blogged in the past about the notion of combining AI with IoT and AR/VR to create a kind of digital twin of the real world, a virtual world of commerce in which we move in a way that might be synchronized with our real-world movements, or might not be. I’ll use an extreme example of this to illustrate, but keep in mind that I am not saying something like this could be realized easily, only that approaching it might generate interest and value.

Suppose we had a complete and accurate 3D model of every public place in the world, one that could be virtually traversed by someone and represented to them via a TV, a computer, phone, tablet, or AR/VR glasses. We could visit any national park, walk any street. Suppose also that we could elect to augment that virtual model with avatars representing people, real, historical, or fictional. Suppose we could converse with them, or that our own avatars could. Suppose this model was augmented so that we could enter a “store”, see “products” available, and purchase them for delivery. We could live as much in the real and as much in the model as we liked, using AR to blend the two if we were out in the real world but also wanted to interact with some of these virtual capabilities. Now suppose we could have that virtual world push information to us, so that we could look down a real street and see the names of buildings, and perhaps see a flag on locations that had something we were looking for. Any enthusiast will tell you that the technology to do this exists now, but the ROI on creating it is suspect.

I think it’s easy to see that we could live and even work in this mixed-world environment. I also think it’s easy to see how creating and hosting it would offer a benefit. To Meta, it’s a metaverse we actually live in, or one we’d like to. To Google and Microsoft, it’s a new way to search, to shop. To enterprises, it could be a way to empower the 40% of workers who roam around in the real world rather than sit at a desk or in an office. There are a lot of things that this could help along, and that’s bad in a way because it creates what looks like a boil-the-ocean deal to VCs, who hate that sort of thing. The good part is that any piece of the whole can build toward it, and all the possible applications create a lot of pieces.

Can this justify AI? I think the whole pie I described could redeem any potential AI investment of any size, but that’s not the way it would have to develop. You can’t lift that sort of transformation into place as a single piece, because there are simply too many players who’d have to be involved in its creation. Can a piece of it do at least a credible first step, for both the concept overall and for AI? That’s likely going to depend on what the giants of the ad-sponsored world do next.