What is Jevons Paradox and why does more AI mean more labor?

Jevons Paradox outlines that increased efficiency in resource usage often leads to higher — rather than lower — total consumption. So what does that mean for the big push into AI for industries outside of Big Tech?

Morning Brief Host Julie Hyman Barron's Investor Circle Newsletter editor Josh Schafer relate this economic theory to the AI Boom and what the pivot toward the "everything is AI now" philosophy has meant for AI consumption and labor demands.

I guess AI should take us to our next topic, um, which is jobs and AI. And I've been trying to figure out how to pronounce this, Jevons, I think it's Jevons paradox. Jevons paradox named after, um, an English economist who wrote about this in 1865 where he was writing about coal and the advent of the steam engine. And instead of decreasing demand for coal, it actually increased demand for coal. And over the years it's been used a lot, um, to apply to different energy, um, innovations specifically. But now it's being talked about for for AI. And all of a sudden, I feel like it's sort of more come into mass culture, this idea of Jevon's paradox. Uh, Torsten Slok over at Apollo has been writing about it a lot, Ezra Klein at the Times wrote about it. And the idea is that rather than replacing labor, somehow AI will increase demand for certain parts of labor. And that's supposed to make us all feel better.

It's so, BNP Paribas economist put it as it's not a zero sum game, which I think is really the takeaway here. If there's a lot of people that think that AI is just going to quote take our jobs and then you are left with nothing and there is no more demand in the economy. And the economist's argument is simply, well, demand is continuing to increase. If you look at the hyperscale CAPEX numbers from last week, Morgan Stanley now estimating it's going to hit 1 trillion in 2027. The money's going somewhere, right? And so the companies that are getting paid need more employees. I was looking at Sterling Infrastructure, which is another pick that we had uh since December, that stocks up almost 70%. So Sterling Infrastructure is sort of at the epicenter of building out the data centers. They used to build highways. They went from building highways to building data centers and I took a look at their employees. First quarter of 2025, 3,200 employees. First quarter of 2026, 4,400 employees. Why are they, why are they adding 30% more employment? Because demand for them is going up. Now, does that help the software

Does that help the software engineer now?

engineer? Now,

But those are, those are, those are picks and shovels though. That's not like the stuff that AI, that the end market AI is gonna affect, right? I mean, we you look at the Bitcoin miners also, a lot of them have pivoted to giving their capacity to over to AI.

But you mentioned Torsten, Torsten's also had charts that show technology jobs going up, right? Just IT sector jobs going up. I also think I think back to BlackRock gave this example at their 2026 outlook, the Blockbuster Netflix example. Blockbuster goes bust. Oh no, no one's going to rent DVDs anymore, the whole DVD business is gone. What happened?

Yeah.

The entire movie and entertainment industry 10x.

Yeah.

Netflix becomes a huge thing, streaming becomes a huge thing, YouTube is huge, you can now film yourself and be a creator and make money off of that. The amount of people that make money off entertainment in 2026 versus 2006 when we were renting DVDs is it's it's exponentially bigger. And so that's in the same industry too, right? It came for movies and it came for DVDs,

and those people pivoted.

But is the

But is the h- is Hollywood in terms of the number of people who work in Hollywood bigger? I mean, it feels like it's been shrinking again in the past couple of years just anecdotally.

But I'm expanding it to entertainment too.

Yeah.

Because if we're going to use Blockbuster and Netflix as sort of comps here, Netflix has documentaries, Netflix, Netflix has an endless amount of content is my point.

Yeah, I, I don't know if I'm entire. I mean, the other example that Torsten gave is call center workers in the Philippines that have increased since AI because they're using AI prompts and so they're they need I guess more of them. Lower cost per interaction does not mean fewer interactions, he says. It means more customers served, more channels open, and more markets worth reaching. But I feel like that is, um, in terms of call center employees, that's got to be an interim period. In other words, you only need the humans because the AI is not good enough yet. And once the AI gets good enough, what do you need the human for?

I don't know. So that feels like an interim

from

I I think it I think that sort of framing applies to most jobs though.

if there's going to be this transitional period.

If you think about what we're, what we would use AI for, use Slack as an example, an app that I know we both used. If that has a more advanced AI chatbot, you're going to need to ask more questions in the beginning process of it, right? Salesforce might actually need more employees to handle the transition. What happens in five years, I'm not entirely sure, but maybe that gives time for people to start to pivot and get into different industries and get around AI. I mean, the computer was going to take all of our jobs too and so was the internet, right? We played this out a lot over the last year and it just didn't happen.

Right.

The Internet took some people's jobs for sure, but the economy, the pie got bigger over the last 30 years. And I think you have to, you have to believe that that's how it's going to play out. Otherwise, you're just kind of an ultimate doomer. And then I get to the point that I always say, which is if we're going to be that doomy, what's your trade? Like, what does that do for me? If you tell me the robots are going to replace all of us, it isn't really that helpful.

Um, the other thing that Ezra Klein talked about is that it's actually tougher for society to, um, adapt if there is smaller job loss than if there is larger job loss. In other words, if you think about, um, NAFTA for example, and the, um, sort of slow drip, drip of manufacturing hollow out in the United States. It wasn't, it didn't affect a large enough number of people all at once for there to be some sort of policy action. Whereas COVID, all at once, a lot of people out of work, huge policy stimulus, you know, DC-led stimulus. So like, will it be harder if you have certain segments where you have layoffs little by little to sort of come up with a, um, a coherent and effective policy response? And I that sort of makes sense to me.

I think that definitely makes sense. And but I don't know which one is necessarily better. So do you, you want it to get really bad so you get a policy response? I I I don't know if that is net better for the overall US economy. Is we need things to get really bad so people take it more seriously is kind of a dangerous game to play. I think I'd rather have it be sector by sector and you slowly start to figure it out and sort of see where things go rather than, hi, we need help. Oh no, that doesn't sound great.

I mean, it does feel like at the beginning of the year when we were talking about the Citrini research piece and all of that, the whole question at that point was like, is it going to happen so fast that we can't deal with it or is it going to happen more slowly than these worst case scenarios suggest. And I think slowly is kind of turning out to be the answer so that we'll be able to adapt to adapt and deal with it.

I think we're too I think we're still too early to know. Yeah. We were, we're still talking about CAPEX being raised and people trying to get chips and there's not enough chips. Like I don't want to do the inning thing, but you are still kind of early innings when you look at where the money is flowing right now.

Is it is it 9:00 p.m. at the party and the party, Dan Ives' formulation, and the party goes to 5:00 a.m.?

If people still can't find enough chips, like we're not really talking about monetization yet. That conversation people have realized is more of a 27 28 question at this point.