‘Complete hogwash’: Ed Zitron on AI promises from Silicon Valley

Ed Zitron has made a name for himself in recent years as an outspoken critic of the tech industry’s push on AI—specifically LLMs.

The CEO of EZ Primary Research, Zitron writes the “Where’s Your Ed At” newsletter and hosts the Better Offline podcast. In part 1, he discussed AI bubbles, government subsidies, and other topics; in this part, he talks about how Silicon Valley is approaching this moment, and why he’s skeptical of the potential of the technology to continue to drive the sector.

This interview has been edited and condensed.

One thing I have heard from people in the industry is that the ROI that they are expecting from all of this technology is not going to come until like 2030.

[Zitron laughs.]

The idea is that they’ll build out all of this infrastructure, lay all this groundwork, and then in four years the profits are going to come. I’m going to guess that you don’t agree that that’s realistic

Yeah, that’s complete hogwash. It is based on nothing. They don’t actually have any foundation to that argument, beyond fairy dust and dreams and vibes.

They desperately want people to think that this is like the dot-com bubble, and that this is useful infrastructure, AI GPUs, the things that are in those data centers. Those things have absolutely no use outside of generative AI, they are pretty much a one-use thing, just poor generative AI services. Well, there are other things you could do, some 3D modeling and all that. The problem with that is that’s not an industry that will substantiate all these data centers.

Now, if they claim the profits are going to come in 2030, well, that’ll be in four years, and they’re going to need revenue for those data centers a lot quicker than that. These data centers are going to come on in the next six-to-12 months, and they’re going to need AI companies that can pay them. The year after that, even more data centers are going to come online.

I estimate there’s probably, like, at least $200 billion a year in annual revenue for these AI compute companies to actually make sense. There needs to be.

If the profits don’t come for four years—and just to be clear, this is an extremely preferential timeline they’re choosing, that’s just far enough off that it’s going to be impossible to check, but close enough that it feels almost realistic…classic trick—I think that they are ignoring the very basic economics of this. They’re trying to make people not think too hard, they want you to stop thinking rationally and think, “Well, at some indeterminate point in the future something good will happen, so we can’t be mad at this. This is actually good, because in the future it will be good, it’s bad now, but it will be good in the future at some point, at some time that no one can say, but it will be so good.”

Then you should be excited for the thing that hasn’t happened yet, and that’s the AI industry, really. It’s just a bunch of hope and cope wrapped up in what is not a real business model.

One of the arguments against AI being a bubble is if this was a bubble, would we be building all of these data centers? Would we be spending this much money on digital infrastructure?

Yes. Absolutely, yes.

Why?

We’ve done this so many other times. We did it with the dot-com bubble. We did it with the great financial crisis to build a bunch of housing that people couldn’t afford, and we kept building the housing based on the mistaken belief that anyone could afford a house—because we stopped thinking about whether we should actually check whether people could afford houses. We have done overbuilds before.

This one’s so disastrous because there really is no future for this tech. If anything, we have more evidence that companies will waste money than we have that they’ll be thoughtful about it. If their argument is, “Well, we wouldn’t be so stupid to spend too much money on something that doesn’t make sense.”’ Yeah, that’s the history of the tech industry, except this is the dumbest version of it they’ve ever, ever done.

I think that this is so typical of the AI industry, but also typical of a tech industry that’s never really been regulated, and a tech industry that’s been allowed to do whatever it wants for as long as I can remember.

Anyone making this argument just doesn’t have a real one, they don’t actually have an answer to any of the problems. It’s a childish argument, honestly. And it’s frustrating, because arguing with a child is usually a losing proposition anyway, but fundamentally, these aren’t arguments in favor of AI, they’re arguments in favor of a dream that someone has for the future, which is very different, and also not the business model.

And it’s frustrating because regular people, if more of these companies get dumped onto the public markets, regular people are going to be the ones that suffer. They always are.

What do you see for the future of AI?

If this was going to turn into something, it would have turned into something by now. We are long past the point where it is reasonable to say, “Oh, it’s the early days.” They’ve spent over a trillion dollars on this stuff, they’ve got all of the regulatory capture they could ask for, they’ve got all of the manufactured consent, they have all the money, all the data, all of everything, and they still have not got a business model.

That is the sign that this industry is not capable of finding one, and there is using the excuse of, “Oh, we’ll work out in the future.”

That is an excuse that you wouldn’t accept from a teenager. Why are we accepting it from some of the largest—theoretically, at least, largest—companies in the world? It doesn’t make any sense, unless of course we have a captured business and tech media, and a government that’s sacrificing itself at the foot of capital.

It’s frustrating, because this will all end poorly for everyone involved.