Meta surprised the Street last week by beating earnings estimates, and it also said its capex on AI could double this year. What may be driving the willingness to invest more on AI is Zuckerberg’s view that this year, “AI starts to dramatically change the way that we work”, but the question is who’s work practices are getting targeted.
“I want to make sure that as many of these very talented people as possible choose Meta as the place that they can make the greatest impact, to deliver personalized products to billions of people around the world,” is what he said, but what products, who’s targeted, and what does “personalized” mean?
It could mean that Meta is simply spending on empowering their own workforce, which is what the article I cite suggests. The problem with that is how a return on AI is generated, and the Street demands returns. Microsoft was punished on its own earnings, because the Street didn’t see enough payoff. So is Meta blindly following Microsoft down the same rat hole, or does it have something different in mind? And what about Amazon and its investment in OpenAI? To answer these questions we have to follow the proverbial money.
The goal of any company is to raise its stock price, and there are two ways to do that. One, the “fundamentals” path, is to show growth in earnings, meaning profits. The other is the “hype” path, which is simply jumping on one of the Street’s many bubble bandwagons. The great majority of the AI investment made to date has fallen into the latter category, but the problem is that bubbles burst when investors start to worry about whether the hype will ever be realized. The emperor of the bubble is buck naked in the end.
That leaves fundamentals. If we look at the major AI players in terms of investment, we see that all of them have bet on a hosted form of AI, relying on enormous farms of GPU-equipped servers supporting an information base that’s often drawn from the whole of the Internet. This approach almost demands some broad mission, a “populist” model of AI usage. The question is how to actually realize profits from such a mission, and that’s a problem for everyone in the space.
There are two ways to earn from some online service these days. One is to get the service user to pay for it, and the other is to somehow rely on ad sponsorship. In the first case, you can draw on businesses, meaning workers, to use the service, with their company paying. You can draw on consumers too, and of course this is the big-buck mass-market opportunity.
My data and personal experience suggests that you can probably get about 15% of consumers to pay as much as $120 per year, and roughly 10% to pay $200 per year, for AI capability. This group generally aligns with the income of the household involved, at roughly the same percentage. That means that roughly 15% and 10%, respectively based on payment, would be targets. That may not sound like a lot, but with about 135 million US households, you’re talking about almost $29 billion per year.
The business side can also be profitable. There are roughly 93 million managerial/administrative and professional workers in the US. Of these, my current model says that 35% could justify a “personal” AI tool investment of $250 per worker per year, and another 40% could justify an investment of $170 per worker per year. This runs out to roughly $14.5 billion, so the total would be $43.5 billion, a decent piece of change. The savings comes in the form of improved productivity, and the problem is that once you’ve empowered a worker you’ve taken that worker out of the population available to fund future AI investment.
The ad sponsorship opportunity is harder to assess. Global adspend is about a trillion dollars, with about three quarters of it, $750 billion, being online. My model says that AI can expect to yield only about 7% of that, or about $52 billion. The problem is that this total is likely to tail off over time, since the biggest impact would come when AI is emerging as an advertising factor and having it could help a company steal market share from those behind the curve.
The interplay of all of this, in my modeling, is that the ad-related opportunity develops first and also dissipates the fastest. Worker empowerment develops second and probably takes at least 4-5 years to mature, and consumer use of AI develops slowest but likely matures over ten years.
This helps explain the current situation. Google, Alphabet, has the easiest path to gain from AI in advertising because of search, YouTube, and its role in ad placement. Microsoft has the easiest path to worker empowerment. Meta has a good opportunity in ad targeting and even ad selling, and Amazon has an opportunity in using AI to facilitate retail execution. Google has been most successful in validating its AI investment so far, with the rest facing skepticism from the Street.
We have, then, the story already cited that Meta talking about AI changing the way we work, and that Zuckerberg says “We’re starting to see projects that used to require big teams now be accomplished by a single, very talented person.” Could Meta be looking at elite empowerment of workers, aiming at the top unit-value-of-labor tier that enterprises say can justify AI. If so, it could be smart for Meta insofar as that target is surely the low apple above the ad space. It would be also challenging given that Meta has no real position in enterprise productivity enhancement. Both Google and Amazon struggled for enterprise cloud traction because Microsoft had a more influential position with enterprise IT organizations. How would Meta fare?
There’s always the chance that they’re seeing AI coding, as the article suggests, as a big opportunity, but despite claims of companies like Meta, enterprises who’ve offered me comment are still uncertain that AI coding really saves much. It may be, as some enterprises have suggested, that it depends on what the code is for; it’s one thing to use AI to build a GUI and another to get it to do a digital twin or event handler.
Then, of course, there’s the metaverse. Meta took the name because it believed it could create a whole series of AR/VR universes for people and workers, and maybe even devices and processes, to inhabit. Have they given up on that? Or might this be a way to personalize and touch billions? AI could surely contribute to a metaverse, and metaverses are close to digital twins.
So, in the end, what is Meta thinking? I don’t think they have serious designs on empowering workers at this point. I don’t think they will resurrect the metaverse concept either. I think they’re pinning their hopes on ads, because you can’t get consumers to pay for a social experience, or many online experiences. If that’s true, then the Street’s relative infatuation with them may ebb quickly, and we can’t expect them to lead any AI charge into the future.
