Jun 25, 2026
On June 25, The Capitol Forum held a conference call with Jessica Wachter, Dr. Bruce I. Jacobs Professor of Quantitative Finance at the Wharton School and Research Associate at the National Bureau of Economic Research, to discuss her recent paper, co-authored with Jonathan Wachter, “What Investment Data Implies About the AI Transition.” The full transcript, which has been modified slightly for accuracy, can be found below.
TEDDY DOWNEY: Good afternoon, everyone. I’m Teddy Downey, Executive Editor here at The Capitol Forum. And today I’m pleased to be joined by Jessica Wachter, the Dr. Bruce I. Jacobs Professor of Quantitative Finance at the Wharton School and Research Associate at the National Bureau of Economic Research. We’ll be discussing Jessica’s recent paper, “What Investment Data Implies About the AI Transition,” which examines what current levels of AI investment may imply about future productivity, growth, economic performance, and financial markets. Jessica, thank you so much for doing this today.
JESSICA WACHTER: Thanks, Teddy. It’s a pleasure to be here.
TEDDY DOWNEY: So, I would love to start off asking you how you thought about writing this paper and what your goal looking into this topic was.
JESSICA WACHTER: Well, so it came out of a lot of conversations with my co-author, Jonathan D. Wachter, who’s my brother. And the two of us are kind of nerdy. And we’ve been talking about AI for years now. And he eventually got me to write this paper.
What spurred it was this very large amount of capital investment. And the idea of, well, this is money, this is real investment in various senses of the word. So, it’s going to have some pretty major implications for economic growth. And so, to trace out what those implications were seemed to be something that wasn’t out there.
So, there’s various task-based estimates. But this angle was something that we saw as something that was missing. And so, we really wanted to know the answer. I mean, that was literally the impetus was what is the answer? What are the implications for economic growth?
TEDDY DOWNEY: And so, there seem to be some different perspectives in the paper. And I’m curious how you explored those different perspectives. One is economic growth. One is return on the investment. One is what happens to these companies if they go bankrupt? How do you match up the return on investment, which is kind of like specific to the firms in many respects versus just what happens to the economy overall?
JESSICA WACHTER: Right. So, there’s two pieces of it. One piece is having observed this investment data, we can come up with something that’s back of the envelope of what should be true about productivity within these firms.
Now, of course, people could make mistakes and there’s a lot of uncertainty. But these investments, a bunch of it’s already baked in. So, we kind of treat this as these firms are playing catch up for productivity boom that’s essentially already occurred.
So, one piece of it is estimating that productivity boom.
The second piece of it is, once you’ve done all this investment, well, that’s a fact. And now you’ve got these new productivity numbers to apply to the investment.
So, think of it as like some just large number, say two trillion, and then you’re multiplying that number and then that gives you GDP essentially. That’s how these production economies work.
So, one piece is what is the increase in productivity?
The other piece is given this productivity, what kind of earnings, what kind of GDP, are we likely to see come out of this?
TEDDY DOWNEY: And you explore different types of productivity booms. Has any economy historically experienced this level of productivity increase in terms of the magnitude before? And what historical comparisons did you find to be most relevant to this exercise?
JESSICA WACHTER: Yeah. So, pushing on the historical comparisons I think is valuable here. And I think that the paper was potentially a little bit misunderstood on X.
And so, here’s the point. The productivity boom that we estimate applies to the AI capital, not to the overall economy. So, we’ve come up with a number. It’s pretty massive. It’s like 2.7 times. And just to give you some perspective, a typical increase in productivity is 5 percent. So, that would be 1.05. So, this is like pretty large, but it’s only to a sector of our overall economy.
But, of course, it is a sector and it’s a rapidly growing sector. So, if you take what’s already baked in—what I consider already baked in based on the numbers that we’ve gathered—then what this looks like, for the overall economy, is on the order of the 90s IT boom in annual terms.
Now, that’s what’s already baked in. That doesn’t include what might happen in 2030. So, that’s basically if we have what we’ve got now, we’ve got the investment we’ve got now, and then it stops and we don’t get the kind of improvements that everybody’s been talking about.
So, I think that’s a pretty pessimistic view. We call that the moderate scenario in the paper. And that’s basically five additional percentage points of growth by 2030.
TEDDY DOWNEY: That’s if there aren’t additional booms.
JESSICA WACHTER: Correct.
TEDDY DOWNEY: And you sort of lay those out in the phases. Can you describe those different phases? Like one is the boom that we already have and what needs to happen to make that make sense? And then another is, well, what if there are additional booms and what are the things that could cause those additional booms?
JESSICA WACHTER: And by the way, with the additional booms, that’s when we get into numbers that are of more historic significance, like the Asian miracle or the Industrial Revolution. And those things are pretty much—to have something like that in a developed economy would be pretty unusual. Let’s put it that way. Perhaps unprecedented.
So, here’s how we think about it. Okay, there’s a couple of facts. Because there’s a lot of things we don’t know about AI.
So, fact number one is the investment dollars that we know going through 2027. So, that’s something that you can gather from public sources. And we’ve spent a lot of time doing that.
Number two is that nobody knows. And the best way to model nobody knows is a coin flip, because 50 percent is the most uncertain thing you could have. So, you’ve got that coin flip, and then we’ve got the possibility that the same boom that we estimate could occur again. Now, of course, it could be larger. It could be smaller. But in a sense, that’s all encapsulated in this coin flip. And I can talk about the technicalities of that if you want.
So, now, you can’t have that going on forever. Because frankly, the economy then explodes, and none of your present value formulas make any sense. So, we basically say, all right, we’ve got two years where that could happen, where we’ve got that elevated probability of 50 percent of this large boom. So, in two years, if you’re flipping the coin, you could get one, it could come early, or it could come late, or you get two. And the two is what we call the singularity, where you get into kind of the larger numbers that then start to match what we’ve heard Elon Musk say about putting one terawatt per year in space. Those sorts of numbers start to sort of match with that number.
And the reason why we chose two years and not more is because we do hear a lot of these forecasts going through the early 2030s. Whereas, after that, I think there’s a general sense that we’re then back in the ambient level of uncertainty. So, that was our choice. So, that’s how we think about it. We’ve seen a boom. Okay, that’s a benchmark for what we could see again, don’t know, could happen again, could happen twice.
TEDDY DOWNEY: And so, you have these three scenarios. And there’s such a big range of outcomes in these three scenarios, it would be great for you to put in layman’s terms how you see those scenarios, sort of what your research found in terms of the outcome in those different scenarios?
JESSICA WACHTER: Yeah, let me pull up the specifics here. All right. And this is all actually outlined, even in the introduction of the paper. Okay, so if, again, we talked about the moderate scenario where we get roughly five percentage points of growth into 2030. Then there’s the transformative scenario where it’s 20 percentage points of growth. And there’s the singularity scenario of over 50 percentage points of growth, all by the early 2030s.
Now, after that, we’re interested in things that could happen past 2030. There, we also see divergent scenarios. Because if you have that, not only do you have those, say, 50 percentage points of growth in the singularity, or 20 percentage points of growth in the transformative scenario, you then have a larger base on which future productivity booms could occur.
Now, of course, it doesn’t have to be through that. There could be spillovers. But that’s where then, just to be—this is how uncertainty works—looking past 2030, we see an extremely large range of scenarios, where by 2050, you could even see, it’s within the confidence intervals to get, say, 400 percentage points, that would be kind of the very upper bound that we estimate.
So, going onto some of the historical comparisons that I know that people find interesting, the thing to keep in mind is that normally our economy grows roughly, we think, at about two percentage points a year. Now you could have various attitudes about that two percentage points. We use 2.5 percent in the paper, but so that over many years could actually have a pretty significant effect. But if we look at long run scenarios, at the long horizon, you can get numbers that are quite a bit larger.
So, if it’s a singularity, it’s a number that’s an order of magnitude larger than the Asian growth miracles. So, there the GDP numbers range about 10 times GDP. So, at a 30 year horizon, we’re looking at seven times GDP. And at a 50 year horizon, we’re looking at 37 times GDP. So, those are sort of those long horizon numbers that we computed for the singularity scenario.
TEDDY DOWNEY: Well, let’s take it from a different perspective. Where is the threshold? Let’s say you’re an investor, and you’re betting on these companies delivering on the promise of this technology. What needs to happen for them? You say in the paper, look. They could go belly up. Or this could just turn into something that’s just unprecedented productivity booms that justify this kind of investment effectively.
JESSICA WACHTER: Yeah, the simple answer is their earnings need to go up by roughly the amount of their productivity. So, the earnings here—this productivity is reflected in these firm earnings. And essentially, there’s various measures of productivity. Whether our 2.7 maps exactly into them is a sort of question in terms of what literal accounting number you’re looking at, but it’s going to have to be pretty substantial. So, the return on equity for these companies is going to have to go up a lot to justify this investment.
TEDDY DOWNEY: And I guess, from a, historic standpoint, how unprecedented would that be, right? According to this paper, I mean, you’re getting—and maybe that’s kind of like where the historical narratives are. It’s like, yeah. Where compared to the, I mean, obviously, the 90s is like a bust really.
JESSICA WACHTER: The interesting thing is that I actually don’t think that’s all that unprecedented. So, I think we are in an unusual period where large companies—these are all large companies—have had or may have very significant productivity increases. But to have a company have productivity increased by three times, a publicly traded company is not unusual in the U.S. economy. To have that be a lot of large companies, yes, that is pretty unusual. But I don’t consider this to be impossible.
TEDDY DOWNEY: So, when you’re thinking about not impossible, it still seems like low to a normal person.
JESSICA WACHTER: I hear that. I guess here’s what I would say. In separate research, I’ve documented the skewness that is present in U.S. stocks. And it’s like the most skewed thing in the world is individual stock returns at a monthly basis. So, my tendency is to believe that these numbers are for real. That’s where I am with this. Actually, we are going to see this productivity growth. Maybe it’s not going to be by 2030. Maybe it’s going to be by 2034. Or maybe it’s going to be by 2028. Or maybe we’re going to see something even better.
We might see something worse. Lots of bad things could happen that would hurt the economy and the aggregate. So, I’ll just throw that out there too. So, there’s all of that, things that could, that we are not, that still could occur. But I do not view this productivity as something that is outside the realm of what we’ve seen in companies in the United States, this kind of increase.
TEDDY DOWNEY: In terms of trying to understand, are there other ways that you can talk about this as being different from what happened in the dot.com era? Where there was a lot of investment and there was just too much hype, right? Like it ended up being a good thing, right? You got a lot of productivity gains out of the internet. I mean, I argue you could even have potentially gotten a lot more had it not been so swiftly monopolized. And it’s unclear what it would be like if there were lots and lots and lots of companies benefiting from the internet instead of seven. But clearly there was kind of a big productivity boom from that investment.
Now you point out in the paper, it didn’t really fundamentally change growth rates over the overall economy that much though. So, how can we think about the historic nature of this growth in comparison to that 90s? Like what are the different ways that it could work where it differs from what happened in the 90s and the dot.com bubble?
JESSICA WACHTER: Sure. And I hope you can hear me because you were cutting out a little bit.
TEDDY DOWNEY: Oh, sorry. Was there anything that wasn’t clear that I need to repeat?
JESSICA WACHTER: No, no, no. I can hear it. So, I got it. I just hope you can understand me. All right. So, first of all, the 90s, the boom was very real in the 90s and it did a lot of great things for people. It held down inflation. We had low inflation. We had high employment. We were able to pay down some of our debts. So, generally this was actually a good period. Now, it ended badly, right? So, there was a dot.com bust. The Fed had to come to the rescue. There was also an overinvestment in fiber optics. So, those are not exactly the same thing.
So, in terms of the actual dollars going into building stuff, this is more like the 90s fiber optic cables than it is the dot.com increase in the stock market. The dot.com increase in the stock market, through various measures, that was not rational. And many people pointed that out at the time as it got pretty high. You could look at things like failures of put-call parity. There was negative stub values.
So, for example, the famous example of Palm being—we don’t remember this today, but Palm was valued at a higher value than 3Com. That was not the only example. So, things like that. So, we do see that now. Like we see companies taking on AI superficially and seeing their stock price increase. But we don’t see the kind of out-of-whack overall price earnings ratios that we saw then. So, I don’t think it’s quite the same thing.
So, just to kind of reiterate, the fiber optic boom was a boom in investment like what we see. The dot.com bubble, there was actually not a lot of real investment. There was just a kind of overall optimism turned into mania—and probably coupled with a lack of ability to short sell—that elevated those stocks and then ultimately led to the bust. In terms of the fiber optics, there was overinvestment, probably misallocation of capital, at least ex post. But all of those fiber optic cables are humming right now. They ultimately were useful.
Maybe this is an overinvestment along the same lines. That’s possible. But I also want people to be aware of overlearning the lessons of the fiber optic cable episode. Because there’s also such a thing as investing too little.
TEDDY DOWNEY: You mentioned some of the mania and the hype. I have a hard time not seeing that right now when you just use the technology. And it seems like there’s actually a big disconnect between what people say it can be used for and then you use it and then it makes errors or what have you. It just falls far short of the claims. I lived through the 90s and the dot.com era. To me, it feels very apt. A lot of what you’re saying—in the paper, you treat Musk talk as sort of like this is one option, right?
JESSICA WACHTER: That’s right.
TEDDY DOWNEY: Sort of treat it straight faced.
JESSICA WACHTER: Yes.
TEDDY DOWNEY: Whereas I read it and I think it’s utter lunacy. That’s my bias, obviously, that I’m seeing the parallels. You’re obviously doing a scenario analysis. But when you’re thinking about what is the most likely outcome or where is the most likely place that we end up being, you end up just saying, hey, the best way to do it is these three options. And it’s sort of like a 50-50 that we’re going to get one of them.
How do you end up there as opposed to saying, well, just given some of this what we’re seeing in the real world, not feel inclined to move that needle a little bit. Maybe it’s 30 percent that they’re really going to get a second boom or what have you. I would just personally feel all the needs like be moving the dial all the time. [screen froze]
JESSICA WACHTER: So, I got most of that, which is what’s leading me to treat this as scenarios as opposed to what seems most likely. So, one piece of it is just my memory of what seems to be an extremely different episode, which was 2007, 2008. And I started really considering the impact of large-scale uncertainty, actually a little bit earlier than 2007. And I was just sort of fascinated by distributions. That’s a long way of saying that in 2007, people were not thinking about what would happen in 2008. It’s not that they were mapping out scenarios and it was low probability. It was not even one of the scenarios.
And that then got me thinking about rare events. I actually wrote my rare event paper about rare disasters, which I think is still actually quite an interesting topic, based mostly in 2007 before the 08 crash happened. I think there’s a lot of value in acknowledging how little we actually know.
In terms of Musk’s predictions, those are at the outer realm of what we are considering. But the other thing is that SpaceX is now a publicly traded company. I know it doesn’t have a lot of float. I know it’s got a non-traditional governance structure. But it’s got an S-1 and it’s subject to the Securities Act.
So, I take what’s written in that S-1 pretty seriously, warts and all. And it actually kind of matches up with Musk’s earlier statements—at least some of them—in terms of how this would pan out. So, I think that if he’s wrong, he’s going to kind of lose more money than anybody else would be one way to think about this. But I think that the reality is we really just don’t know.
TEDDY DOWNEY: Now, having lived through 2007, 2008, one of my customers back then was Steve Eisman. He was going around saying this is going to be a catastrophe. I was just in Florida. There are strippers there with fourth homes. It wasn’t that unpredictable at one level. There were a lot of people who saw something pretty bad happening. I know you’re saying, well, nobody was predicting it. There were some people predicting it.
But I still see a lot of value in saying we don’t know, right? I’m not saying we don’t see a lot of value and we don’t know. But do you worry at all that there are—or not worry at all—but how should we think about the value of just saying, hey, we don’t know. But if we want to if we want to take your model and then tweak it ourselves, what’s the best way to look at the different scenarios and kind of think through the model that you’ve created of like where I think things will end up?
JESSICA WACHTER: Great. And thank you for that corrective. Yes, people did foresee it. A small number of people who then made a lot of money. It was hard to see it in terms of the forecasts, say Fed forecasts or other forecasts, people were saying in terms of economic growth. Well, so we could talk about 2008 as a separate topic. In terms of acknowledging the uncertainty—so, having said that, I somewhat—could you restate the rest of your question?
TEDDY DOWNEY: I really want to think about your model.
JESSICA WACHTER: Oh, the model, yes, yes, yes.
TEDDY DOWNEY: I really want to think about your model is a scenario analysis. And then we can say, hey, I think it’s—because I’m maybe a little bit more skeptical of Musk or think he’s a pathological liar or what have you. I’m going to go to somewhere between scenario one and scenario two to see if they’re really going to make it, to see where I feel like this is going to end up.
I’m looking at these companies. I’ve seen people slow their spend, lose a little bit of enthusiasm about the opportunity here to really fire a lot of people or replace a lot of people, whatever. There’s obviously a lot of indicia in the economy to see is this productivity really happening and being realized? But what I love about your paper is these three scenarios with a lot of outcomes, how can we look at the model and interact with the model with our own views?
JESSICA WACHTER: Absolutely. Great. Thanks for that question. We have a spreadsheet. It’s on my website. Also, there’s a link on the paper itself and you can type in all kinds of possible numbers in the spreadsheet and it will show you exactly what will happen. So, the model is pretty transparent. And I would definitely encourage you to download the spreadsheet and play around with it.
TEDDY DOWNEY: I love it. We’ve got a question here. If you have questions, please put them in the Q&A. We’ll get to them. Or if you can put them in the chat, that also works.
First question. Are your back of envelope GDP estimates an underestimate since they assume none of the productivity increases in the AI sector propagate as cheaper inputs to the non-AI sector?
JESSICA WACHTER: Yes. So, we actually don’t model any kind of spillover. So, the economic effects just come from this sector and these companies getting bigger. And essentially, in this model, they grab the profits and we all just pay them money. Which is something that I hope doesn’t happen in a sense, that it’s going to be a little bit more evenly spread out. But yes, in that sense, it absolutely is an underestimate. We don’t we don’t incorporate that.
TEDDY DOWNEY: There just seem like a lot of variables that would complicate that, just the cost of lots of different things, increasing and decreasing certain commodities like memory chips and things like that just seems would be pretty tricky.
Another question. Is the possibility of these companies getting bailed out – i.e., too big to fail moral hazard—endogenous to the model? In other words, maybe the required increase in productivity is an overestimate since they can rely on the government to step in if the productivity fails to increase.
TEDDY DOWNEY: This is a phenomenal question. I spent a lot of time being very curious about this notion that the companies want the government to make an investment. Which to me, obviously, is indicia that they’re not going to have the productivity gains that they expect. Otherwise, why would they want the government to take an investment?
JESSICA WACHTER: Yeah. So, no, that is not in the model. I really hope—I’ll just say – that that does not happen. I think that if they fail, they should be allowed to fail. I mean, I think a bailout here would be ridiculous.
TEDDY DOWNEY: Actually, I’d like to hear why. Why do you think it would be ridiculous?
JESSICA WACHTER: I think that it just messes with incentives. And I don’t think there – there are times and places for that. I think that, for example, having FDIC insurance is very important for banks so you don’t have a bank run. I don’t see a bank run issue here. Like there’s no kind of theoretical reason why you need something like some kind of insurance mechanism.
I mean, I think that if these companies fail—I’m not saying it would be good for the economy. But I think the first order of fact is that a lot of wealthy people will lose money. I mean, that really can’t be something that the government should be expected to do something about.
TEDDY DOWNEY: I couldn’t agree with you more. Although, I did watch what happened with the Silicon Valley Bank, and almost exactly what you had seemed to unfold when it came to that. Obviously, they’re worried about contagion. It was a bank, technically. But a lot of rich people got bailed out.
Now, I would love to stay on this line of thinking for a second. I had another question I saw earlier. I don’t know where it went. You worked at the SEC. I was impressed that you still have faith in what people put in S-1s. I have less confidence than you in businesspeople telling the truth or accurately filing statements with regulators.
Tell me about your experience at the SEC, how you think about the integrity of the rule of law, the good faith or not good faith nature of what you’re seeing going on, when it comes to SEC enforcement. Because I would love to be reassured that things are still going well there.
You left on great terms. You think it’s still operating well. I have my skepticism of that. But you’re coming out of the SEC saying—what you’re saying actually gives me some faith that we do have good institutions here, but I would be remiss if I didn’t ask about it.
JESSICA WACHTER: Yeah, I was amazed about how little I knew about the SEC before going there. It’s an amazing group of people who really consider the integrity of capital markets to be their life’s work and religion. So, I don’t think that is something that changes so easily with the ins and outs of the political cycle.
It’s the philosophy of market regulation that you don’t mislead the public, and in return, you get to raise money, but the SEC is not a content regulator. The system of market regulation that we have is full disclosure. Don’t be misleading. Don’t omit material facts that would change how people make their decisions. That’s the definition of materiality, more or less. Then you can raise money. And even if it sounds kind of crazy, well, go for it. I think some of this is in that spirit.
Through the ins and outs of time is what’s allowed our country to prosper. I have a lot of confidence that in particular is going to continue. So, I think it’s a good setup that we created in the 30s. It’s stood the test of time. It’s co-evolved with all the changes that we’ve had. I think it can continue to do that.
TEDDY DOWNEY: One question here from the audience, and we have a couple more. Whether investment trends indicate antitrust concerns.
JESSICA WACHTER: I am not an antitrust economist. There are some excellent experts on that. To me, I see companies here that are viciously competing, and the model assumes that they compete the economic rents down to zero.
To some extent, you could interpret this increase in investment as really not monopolistic, because what a monopolist does is basically limit supply in order to charge more. With the proviso that I’m not an antitrust or competition economist, I don’t see any clear worries. Just a lot of companies getting bigger. That is true.
TEDDY DOWNEY: Wouldn’t you need to include some potential monopolistic conduct? The companies that are competing, one of them is literally a monopolist. Google literally lost two monopoly cases already. Isn’t there some risk that that would continue? Yes, maybe they’re spending now, but to monopolize later, effectively. That’s what they all talk about. That’s what they all want.
JESSICA WACHTER: Yeah, they may be assuming that they’re going to be the ones that win and everybody else is going to be dead. Actually, the paper contemplates that possibility. But the numbers in the paper do assume that they’re not going to earn monopoly rents. I definitely acknowledge your point that these companies have behaved like monopolists. And if they have the opportunity in the future, they may continue to do so. But it may be that this is a situation where it will be hard to have a monopoly, which in a way is maybe not great for the stock price. That is the direction the investment numbers point to.
TEDDY DOWNEY: I see. So, the two ways that this doesn’t work out if you’re a stock picker—I guess it might work out if you invested in all of them and one of them became a monopolist and you still had that money in that one. But the two-ways that this may not work out for a stock picker are they don’t reach the productivity gains and they go bankrupt, or it remains competitive and you just don’t get the kind of rents that you were. So, the benefit that you see in the paper redounds to the whole economy, effectively.
JESSICA WACHTER: Well, this is where things do get a little bit tricky interpretation-wise. In the paper, the productivity gains redound to these companies. So, that’s literally the interpretation. They are not monopolists in the sense that in economic terms, if you discount their present value. They’re investing to the point where this is a zero NPV, zero net present value investment. So, they don’t shrink their investments so that they can basically earn more money.
Now, the question of who is going to benefit from this, that’s a deeper question, which I’m just going to be honest, the paper actually doesn’t really contemplate. Because we don’t have a general equilibrium model that would require answering that question.
Now, as you can tell from this conversation, I’m an AI optimist. I would put myself in that camp, I guess, at least relative to many economists. And I do tend to think that this will be a societal benefit. But the paper is structured to have a calculation about GDP. It doesn’t answer that social welfare question.
TEDDY DOWNEY: But you are assuming that it’s competitive.
JESSICA WACHTER: Yeah.
TEDDY DOWNEY: And so, the benefits of a competitive economy typically, redound to a broader public than a monopolized economy.
JESSICA WACHTER: I think that is totally fair. Yes, that’s correct.
TEDDY DOWNEY: Now, I guess my last question, a lot of—well, we have another question, but you’ve kind of answered this already. But I’m just going to ask it anyway. And yes, again, if you have questions, please put them in the Q&A or in the chat and we will get to them. Under which conditions is current AI CapEx justified? I mean, it’s sort of what the whole paper is about.
JESSICA WACHTER: 2.7 is the number. 2.7 times increase in productivity. That’s it for these firms. That is when this becomes justified. And that should happen by 2030.
TEDDY DOWNEY: So, by 2030, their productivity has increased by 2.7.
JESSICA WACHTER: 2.7 times.
TEDDY DOWNEY: 2.7 times.
JESSICA WACHTER: Yes.
TEDDY DOWNEY: And at the risk of being redundant, that’s not really that hard. But it is historically pretty hard for such large companies.
JESSICA WACHTER: Yes, that’s true.
TEDDY DOWNEY: And so, when we’re talking about the size of companies that have done that—I mean, look, obviously, these are some of the biggest companies of all time. They’re such a huge percentage of the economy already. Some of them have had massive market cap increase in a relatively small period of time, which I think is more due to them being monopolies than other things. But can you give us a little bit of historical context over the types of companies that have been able to hit this kind of mark and how big they are? Just so that we can get, like, how unprecedented it would be for such a large firm?
JESSICA WACHTER: Yeah. So, I mean, some of this, honestly, I should still collect some of the data on. But I think if you look over the last 20 years, my sense is that these numbers are rare, but not unprecedented.
TEDDY DOWNEY: Okay. Okay, great. Oh, here we’ve got one other question. Is there good evidence or reason to think that investments in general do get the returns to validate them? Seems you’re assuming these investments are rational a priori.
JESSICA WACHTER: That is absolutely true. That is our a priori assumption that these firms are investing with some kind of optimality in mind, that they’re not burning money by putting—but maybe they are. Maybe they’re investing because they just want to invest. They’re not thinking about the future. If that’s the case, that would be a problem for our story.
What can I say? I think that, for all the hype out there, I do think that companies attempt to forecast the future and make investments accordingly. And part of the reason that we’ve done very well as an economy over however many years is that the largest allocators of capital actually do not do a terrible job.
Occasionally, they do a terrible job, and we tend to see that. When a large company makes some bad decisions, they actually do eventually go bankrupt or they shrink a lot and they lose market share. And I think we can all think of examples like that. And that’s important to allow that to happen.
But the fact is, our largest companies over the past 20 years or so have actually been quite successful in terms of how they’ve allocated their capital, which isn’t to say that they never make mistakes. But I would not be so cynical in terms of—or maybe cynicism is the wrong word here. I don’t want to put words in people’s mouths. But I think at least in a back of the envelope sense, they are hoping to make money from these dollars. I don’t think that’s crazy.
TEDDY DOWNEY: Can I just push back a little bit on this? Because I think I just 100 percent disagree. And I would make this argument, which is that these companies have remained big because they have acquired all of their competitors. They have acquired companies that let them continue to grow. They have not done very much R&D on their own—maybe with the exception of Google in some respects—but often it’s because they bought those companies.
Meta in particular has had a spectacular track record of failure with big spending on new initiatives in terms of the metaverse. Basically, anything they’ve ever tried to do outside of buying Instagram and WhatsApp has been almost a colossal failure.
Then you’ve got Elon Musk who—okay, Tesla and SpaceX, those are government. One was government subsidized. One is based on contracts from the government. The rest, the Boring Company, SolarCity, a lot of other things didn’t quite pan out that well. We could go on.
Even Apple’s had one product that’s done well for the past 30 years. We could spend a lot of time on this, but I don’t know if these companies really have done a ton of successful R&D driven spending.
And then we also haven’t gotten into all the real world limitations on the potential for this technology to be successful. They need to build data centers. There’s tremendous public pushback on that. They need electricity. The cost has already gone up. They need memory chips. That has also been monopolized by a small cartel of manufacturers that are now massively increasing the price and making them scarce.
There’s just like a lot of ways that this could go wrong, but you kind of don’t have all those in the model. Is that a concern to you that maybe it’s more likely than you’re saying that things could go south here?
JESSICA WACHTER: Well, so in a way I do, in the sense that the moderate scenario could be the end of this. And if the moderate scenario is the end of it, then I think we’re going to see a lot of pessimism. Because it’ll be disappointing in the sense that people have these other possibilities in mind. Some people actually believe that these things are not going to happen. And if we all converge on that belief, then what we’re going to see is a massive slowdown in investment. And so, what that will look like is when we reach 2030, essentially the growth slows down a lot. Now, we will still have had the increase in level, but we won’t see the growth that many are hoping for. And that is a contemplated scenario.
So, I think that the levels here of investment, I think that could be productive. And then if it turns out to be a disappointment, we’ll see. There’s going to be like a lot of news about that. Let’s put it that way. So, I think that the moderate scenario here looks like—in terms of how we’re going to talk about it—like things will have been disappointing.
So, in that sense, it’s contemplated. Could we have something worse than the moderate scenario? I think that’s unlikely at this point, but it is certainly possible. And I think the leading reason for that would be something happens outside of the model. I mean, outside of the model in the sense of outside of an AI type point. I guess I’ll say two things.
First of all, we are not focusing on there’s a lot of bad things that could happen that this, I think, is potentially going to prevent. Which I realize you and I may disagree on that. But, for example, the U.S. has borrowed a ton of money and we may not be able to pay it back. I think that AI has actually already helped us with that. Like, I think it’s actually helped hold down inflation and interest rates relative to what might be otherwise. I think that would be a more salient problem had we not had this boom. So, that’s point number one.
Point number two, I think for a number of people using this, there are specific tasks where this has just massively increased their productivity, right? So, for some, there are at least some tasks, some software tasks, for example, where the increase in productivity is not three times. It’s like a thousand times. I mean, there are certain tasks that I think most of us know that we would never have been able to do or would have taken us a month, and now we can do.
So, I think that having a sector of the economy increase productivity by three times is not outside the realm of the experience of at least people that I talk to. So, that’s essentially what’s being assumed right now. And so, I think that one can discount, can say, look, Musk, way over optimistic. Maybe we’ll get there 200 years from now, not five years from now, the way he’s talking about that. And still think, hey, this is going to be a pretty good thing over the next five years.
TEDDY DOWNEY: Last question, then I’ll let you go. You’ve been very generous. Is there anything that you think investors, policymakers, should be thinking about when it comes to the economics of the AI transition? You mentioned inflation. Obviously, financial policymakers need to be thinking about that. You’ve already mentioned don’t bail them out. That would. Be a bad idea. Any other things that we haven’t touched on, as a result of your paper, that people should be thinking about or watching to see how. Where this is going to land on the in the spectrum of the three scenarios.
JESSICA WACHTER: Yeah. So, as I said, don’t bail them out. But I also think don’t hold it back. I mean, I think we’ve got to let it happen. That’s my view. I mean, I think that as a society, we face a lot of challenges. I think this could help us with them. So, since you’ve given me the platform, I’m going to say I don’t think we should be regulating this stuff. I think we should let it loose, let it be what it wants to be.
TEDDY DOWNEY: Well, I just had an interview that ended on an AI philosophical question. So, we’ve gotten a little bit outside the scope here. So, I’m going to ask this final question to you. When you look at AI and the potential for growth and you have this singularity and you have these different outcomes here, what is the role—do you think about the role of what it is to be human? What is the point of GDP? What is the point of an economy? What is the point of a society at all when you’re looking at these things? Like an economy, in many ways, is just part of being in a human society. And so, if being human is enjoying art, is enjoying your job, is participating in a community, a lot of that AI can’t do, right? At a certain point, there’s a limitation on the scope of this technological advance.
I mean, I’m guessing you’re not really thinking about that in the model. But at some point, when you’re doing the paper, you’re looking at the model, and you’re like, is this going to be 50 percent—like such a huge percentage of the economy. At what point is this like meaningless to consider, right? Because like you’ve basically just completely upended humanity.
JESSICA WACHTER: Yeah. So, I think about that all the time. And I think one of the really interesting things about AI is how it’s led us to reflect on our own cognition and what might be similar and different. Because we only know about how we think about things. Now we’ve got this potential other model. I think it’s a pretty different model from how people think. I tend to think it’s a tool like other tools.
So, I don’t think it’s something that’s going to upend humanity at all. I think that there’s always going to be role for people always. The next 2,000 years, I would think there’s going to be role for people. And I think we’re just going to solve some problems that right now seem insoluble is my hope and just make life better for the next group of humans that are coming our way.
That’s my hope and my belief in this technology, that it’s mainly going to be used for that. Of course, there’s catastrophic outcomes as well. But such is the case for any technological tool, really.
TEDDY DOWNEY: I couldn’t think of a better way to end this conversation. Dr. Wachter, a true pleasure. The paper was very, very interesting. This call was super insightful and I can’t thank you enough for your time.
JESSICA WACHTER: Thank you very much. Thanks for your great questions.
TEDDY DOWNEY: And thanks to everyone for joining the call today. This concludes the call. Bye-bye.
