YouSaid · the spoken record

Steve Hou

lines on the record: 62
first: 2026-04-29
most recent: 2026-04-29
sittings or episodes: 1
sources: podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

“There were words that were used that gave people the wrong impression of what was meant. So I think hopefully better communication will also come along as a result of AI. So I'm actually quite excited. I can't pretend to say I can fully predict the comprehensive effects, but I do think we'll get a better economic analysis and policies.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Working with more powerful tools, right? So the analogy I like to give people is a bit like you want to be in a situation where you are asking a mission chef to make a fried rice because that's when they want to creativity is the greatest because you want to be mismatched with the problem where you are really comfortable with the key tools that you can be creative with it. You don't want to be in a position where you just about match with the tools that you're just about able to use them. You're not really able to, I think, use it quite as well. So I think economics will actually be very, very significant impact by AI. And of course, the writing part, that's the most boring part is the communication part of being able to explain to people policies, implications. And I think part of the reason why the last five years, the Fed has had such struggle. I do think was a communication problem. There was a lot of misunderstanding.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Fat the paper into AI and say, Hey, how can I adapt this model to this context? And it just does it like that, right? And you start playing with it. I feel much more confident about tools that I had learned, but didn't fully grasp during school, during grad school. We have all been there at the moment, right, where sort of you go through the course of the semester towards the last third of the semester. Exactly. You heard about this thing. You did the homework, but you didn't fully grasp how exactly to do it. And if you ask to do a point bank, you probably wouldn't feel comfortable, but not with AI. you feel quite comfortable like i if you ask to do anything like i can i can i can you know build up a common filter you know very comfortably and start interrogating the results and and so so i think it allows people to feel more comfortable”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“This equilibrium response, right? How would different parts of the economy system, financial system respond to it? And that is actually something that a genetic simulation can actually do a lot of work on. We have sort of genuinely realistic agents and you can do many, many times of such a simulation to get a sense of the different types of ways in which scenarios can emerge. And I think there have been different AI startups have actually created exactly such applications. I think that's one direction I think I find quite exciting. More generally, I think we're just going to get more realistic modeling, borrowing models from different disciplines. I was talking to my friend and potential co-author, Alex, and he was borrowing a paper from a completely different topic.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, and leaving breathing economy like physics studies stuff that have obey and follow inanimate laws that don't change when you put in policies or change the rules of the game, whereas the macroeconomy is constantly evolving even without policy changes and policy changes are always being introduced. So what I do think AI is going to impact is we're going to have much more modeling, a lot more very modeling by everybody, really. And we're going to have richer models. And I think we're going to have much better agentic simulations. So this idea that currently how do you evaluate the impact of a policy supposed to say we there's this discussion of the Fed is going to massively reduce the size of this balance sheet and in tandem cut interest rates, unlike any other discipline.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“I think he's got a tremendously impactful thing. Fundamentally, we shape the way we think about economic modeling. Economics is a very young profession. That's the first thing. But when people give macro grief, I want to remind people that physics is much, much older than economics. Macroeconomics especially was invented by John Maynard Keynes who first sort of did this rough decomposition of the abandoned composition of C++, I plus G plus investment, each of which was then procedure to model. I mean This idea that we do poorly on macro forecasting, I mean, no kidding. I mean, it is an incredibly”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“That ties into the last question I have for you is actually just around how you think about AI impacts actually your profession of economics in general, especially the world of econometrics. I mean, I find even myself as just a simple, humble macroprognosticator, it's become very useful for me to just be able to run a bunch of different regressions on all kinds of random different things. And I can just imagine the potential power for Connice wheel to leverage these tools. I'm just curious, yeah, how do you think about the future of that industry?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Absolutely, right? And we have a massive shortage of electricians. And now we're building out these data centers, and they are brand new. I think very quickly, we're going to have an issue where all these data centers are going to be maintained, right? And they need people who are not just sort of random electricians for sure with a typical toolkit and just sort of start unscrewing things, right? There's going to be some knowledge training and so on and so forth. But you just don't have that type of workforce for. I think Bomo's disease will actually rain hard in ways that we currently do not, I think. Yeah.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, I mean, one funny example of that is I don't know if you saw the Dwarkash podcast where Jensen was on there and he was saying that the biggest constraint they have right now is actually plumbers for their data center. That's where the bottleneck is. It's just like real-life plumbers. And that has a lot of real-world impact of suddenly the average family does not have as easy access to a plumber.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“House care or get your house fixed, get your basement plumbing done and education and so on and so forth. And that stuff. I think AI will help, right? But I think that journey is perhaps a little bit more drawn out than we currently have.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Nanny. I mean, I'm having, I know exactly. I'm having my roof repaired. I've got this 100 euro house with beautiful Spanish tiles that are also quite nettlesome when they break. And I've had it's been a struggle for me since last fall. I've been looking for contractors and it's quite difficult to find them, right? And when they are busy, they don't answer your calls. They've got terrible websites. When they show up, they don't speak your language. And then it's just a few of them. And when they are there, they're fast and they're slow. So this type of thing, I just don't see how AI is actually directly impacting it right away. And that's exactly the type of stuff that's really what's causing people anxiety, right? When it comes to inflation, it's not really, I think, sort of a bag of chips becoming 5%, 10% more expensive, right? It's the teenager problem how much it costs you to get childcare, right? How much it costs you to get”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, it feels like the other one too is just the. The assumption in there is almost just this scalability associated with it that to your point about Hasaf's book where he's the same that we can reach this cornucopia where everything's free and limitless. The main anchor point there is that all of this AI stuff, like we've been talking about, is anchored to real world components, whether that's energy inputs to power the data centers or just like these very acute, you know, you mentioned a few of them earlier, like just these very random things that nobody's really paid attention to for quite a while other than the core experts within those industries. And suddenly, Nanny, I mean, I'm having a...”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Which I think right now is quite far away from everything we've seen is all just language AI and LLM and so on. And I think for that impact to get here, it really needs to make a lot more, we need to make a lot more progress in the physical world, especially so deeply afflicted by this idea of concept of Bormo's disease, which for people who are not familiar is basically this idea that some sectors such as nannies or childcare or healthcare doesn't really progress with technological innovation as you would have with, say, manufacturing. So that is on the broad strokes how I think about this question.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Now, there is actually a problem there in that when you try to boost productivity by, say, investing in AI, you can advertisely raise R, right? Because you. Compete for capital, right? That's the timing issue there. We ignore that for now. There's also a longer term issue of if let's say the government has to create programs to support workers that have been affected by AI as a result of job displacement that could also also affect the second term. So I think the net impact of AI is a little bit unclear, at least in terms of timing, in terms of in the short run, right? In the long run, there's, I think, a loose hope that this thing is going to work magic, right? We're going to get this miraculous in the GDP ground.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“An investment and so on and so forth. And we've tried the idea of cutting raising taxes and cutting spending, each of which has proven politically pretty impossible, right? And this is the political economy issue, right? I think we've more or less given up on the idea of raising taxes. It's certainly not on the middle class if you want to have any shot of electoral success and cutting spending has also proven quite difficult. So the only option left is to grow the denominator, to grow the GDP. So if you decompose that sort of GDP is basically the sum of interest rate minus GDP growth times your current GDP, that tells you how much the thing is organically growing plus the primary deficit GDP. That's how much you're adding to it. We can try to affect G through productivity.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“And because the dead load itself, which peeped, last peeped, I think, over 100% of GDP during World War II, and then went down in the 70s and then now came back up. And now we're on the secular high again. You suddenly have that service interest cost itself being a big taker of your tax dollars. So how do you make the math work so that tax receipt is not hopelessly smaller than the sum of interest and your primary deficits that's just coming from spending on your discretion once governing government programs? So that is actually fundamentally the debt problem. How do you prevent this spiraling of this debt GDP ratio, which of course is going to have its own adverse impact on the economy in terms of suppressing productive use of capital?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, so if you look at the slide right here, right? I think this equation actually helps us make equation. It's just like, you know, and what is the problem with that, right? I think that should be resolved first for anything. It's not a moral problem. It's an arithmetic problem, right? The arithmetic problem is the following. If you plot tax receipt, federal tax receipt is a share of GDP over the course of the US history, modern history. It's always about 18 to 20 percent of GDP. And we've been through significant tax reform, right? We've caught personal income tax. We've cut corporate income tax. We've got level curve. And you've raised tax or whatever. Regardless of what we have done, the total tax receipt is about 17 to 20 percent. Now, the problem with that is that then the spending has been rising second.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Whether AI causes this inflation, we should actually be able to observe it, right? So wage growth should be coming down, leading wage growth, if you look at postal wages, indeed, you have all sorts of private measures that is not just sort of the slow government measures that allows you to see whether or not the labor market is indeed posting slow wage growth. And you can even try to dive it down into specific sectors that should be more or less affected by AI and generate evidence to argue that. And then to the extent that we know that rent inflation things, services inflation is very tied to W2 income because there's this integer problem that one person can only one meal at a time or say in one hotel room at a time that was what I refocused on rather than trying to resolve this impossible debate whether or not how much of this”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“So I wasn't, and then, of course, all the badgering of the Fed came along. And that really, I think, hurt the case, right? Because you don't want to undermine the credibility of the Fed and gave off an impression that the only reason they were cutting is because they got badgered into cutting of political influence. But apart from that, I think it's sort of, I think that's actually independent of the AI effect. It's okay to look through transitory supply shocks, especially of volatile components such as food or energy. That is classical economics monetary policy that we are quite used to. I don't know how you actually draw a direct link with AI as a factor, sort of the connection that I was just alluding to, right? Even if there is this unresolved debate about”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“I mean, honestly, you've seen me tweet about this as well. Like, I was actually quite okay with this idea of the Fed looking through some of the transitory supply shocks. Like, I was okay with the oil shocks being a transitory shock, although that one has clearly become a very dirty word and triggers a lot of people. But textbook economics does teach you that, right?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“So, okay, yeah. I mean, I think it's fair to say that it would be a mistake to preemptively do it. But maybe the other side of the argument is to be very acutely aware and finger on the pulse ready to cut if you start to see that, okay, we are getting a lot of growth, but, oh, look, it's not inflationary growth. Like the economy is actually not overheating. And then to lean to it. And that seems to be the argument they're making that like, okay, let's look at what Greenspan did during the dot-com boom, which is that he leaned into it instead of slamming the brakes on the economy. Do you think there's some validity to that approach?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Economy that is not very good at manufacturing stuff that has a lot of frictions that has all sorts of electricity shortages, you know, dilapidated grid system, you know, you have shortage of hardware. And that's literally being driven without being priced by the demand of AI that's causing its own inflation. I fail to see how we're going to over-index the”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“In labor market weakening. And if the labor market is not weakening, if you don't see growth showing that type of impact, for that matter, healthcare services, then I don't know whether it actually really matters for you to debate or speculate whether AI will have that disinflation effect. We just should just watch whatever way it should be always watching. So I'm quite skeptical of this argument that we should preemptively justify rate cuts on the anticipation, unproven anticipation that AI was going to have this miraculous and dramatic impact on inflation, if anything. I think the AI buildup, which still happens, you know, in order with all the age, we're talking about hundreds of billions of dollars of cutbacks in the US.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Hospital and healthcare went up secularly and manufactured goods went down over time. We are used to that type of inflation dynamics. What we are not quite able to, I think, ingest this time is this is a technology potentially able to impact services productivity, white collar work could become more productive, but how would that actually show up two-thirds of services inflation is what rent inflation and rent inflation is driven by wage growth. Maybe if we did have a massive slowdown in wage growth because the labor market becomes slackened so much by AA adoption that people don't get hired and labor market weakens so much that you actually get that disinflation but you will just see it right you would just see that in the label”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“I mean, it seems to me very courageous to make such an assertion with so little data, concrete data, even on a theoretical level, I find it difficult, there's a timing issue, right? Productivity gains through disinflation, social productivity gains through technological innovation was always a slow J-curve kind of a thing. And even if it does happen, there is the impact on goods versus services and we know that the stuff that we generally do worry about is really the bumble effect, the labor intensive, low productivity things that is not easily automated, right? We've always had good disinflation through automation across TV. You track the price has gone up. Everyone has seen that chart where college.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“So I don't know, is Kevin Wars, the first person to have been prominently brought up this idea of justifying red cards based on anticipated deflationary impact of that always felt, I mean, when I first heard, I was a little bit surprised. I didn't know whether it was meant as a joke or not, but once I realized it was actually serious.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Economic framework just on anecdotes, even though I do think you're right that they will lead. But that's the question at hand now. It's like, okay, there is, there is a lot of anecdotes out there of what's going on. But when you get to the Fed and you ask the Fed of like, okay, how do we decide how to set policy based on this whole thing? Because there is a potential tail outcomes that are very meaningful to how monetary policymakers should react to this. And it's hard to understand. To navigate this whole thing, so yeah, I'm curious how do you think”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Of the AI adoption, I just think that study will eventually sort of come from more and more companies adopting and then we'll see more economics research showing that we've had. But I think it's not something we're actually seeing aggregate labor.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“And more efficient, God knows how to actually measure that. But I think as such stories, I think sort of percolate and hear more of them, I think I find difficult to believe that everyone is just somehow aligned to themselves. There is also a degree of time to learn, right? So the economics, there's this concept of time of time to build, right? The idea is that any sort of change takes time for that adjustment to come for it to actually see the net impact within corporate setting, you know, people have their workflows and it's difficult to attribute. So forget about even AI, even sort of you think about you as an individual within a company, how do you actually find out whether or not you add a value to the overall company's P&L, right? It's pretty virtually impossible. It's a pretty difficult task. So now you sort of add on top.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“I think that will come. It's probably just not here yet, but eventually evidence will become quite overwhelming. Jeff Bezos has this famous quote where he says when anecdotes and data disagree that he thinks anecdotes are usually right and data are just mismeasured. We are very good at mismasuring stuff, especially with that availability bias where if there is stuff that's difficult to measure, then stuff that's easy to measure. put a lot of weight on the stuff that's easy to measure and try to do crazy amount of inference on the stuff that we can measure. The fact we don't observe, I think, cleanly identifiable labor productivity from AI is not to me evidence that AI is not giving off productivity. We hear anecdotal stories of people who have adopted AI tell us, oh, my work first become 20%.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“But of course, the more interesting question everyone's trying to really get at is some version of a cleanly identified causal link between adopting AI, making people more productive, like saying making widgets. I used to be able to make five widgets per hour. Now with AI, I can make eight, right? Or I was equal. And that is exactly why economics people are economists and so on obsessed with this idea of causal identification.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Owner of a company says, Oh, I'm just not going to hire you people, but I'm going to adopt this AI thing. I'm going to pay for the AI subscription and you guys will do more work. There won't be more new help. And whoever is there suddenly feels that pinch in the labor market is not as tight anymore. They work hard and they try to do more generate more value using the tools that's available to them. Observation, I don't know if there's actually a difference between AI having made everybody productive or not.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Exams, like you know, so the fact that we have all these new investments, you know, that's probably you could actually have a give you give of a mirage. And also, if you remember, if you look at the measured productivity to the last 10 years, when COVID first happened, productivity went through the roof, right? Because everyone got fired and output didn't really dip right away. So there is a certain degree of that going on as well, which I suspect is sort of in the labor market is becoming a little bit more productive in that people are not getting hired. It was over hiring, you know, during COVID. And as we know, every company, especially a big corporation in America, has probably some slack bloat in the first place. So I think this is a combination of mechanical factors inside of which, you know, AI probably plays a role and it's hard to tease out, right? Suppose your boss says, okay, so supposed to”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“And you basically sort of take a look at how total GPU growth is subjected out, but there's stuff that you can measure and you attribute the rest of it to labor productivity. I suspect a big chunk of it is actually composition bias that we sort of move towards, you know, if the labor intensive sectors are shrinking and then you have a lot more investments that is very capital intensive and then productivity is measured on dollar value generated. It's going to look as though, right, that the productivity has gone up.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“First of all, I think that's a great question, that's a really difficult question, and I think everyone is trying to figure out the answer to that question. The aggregate numbers we're looking at, my intuition is that it's almost certainly not reflecting AI productivity, right? We are seeing measured labor productivity, which is a residual, right?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“To get as much productivity out of it, you know, and maybe this current period is almost like a training ground for you to develop those skills. There will come a time when AI will become more expensive, at which point the people who are deploying AI most productively are going to be paid the most, right? Then you'll maybe even have like a chief token officer at every company who basically decide how much token, how much compute. How do you sort of decide between hiring human versus OPECs? And the substitution or the complementarity looks like. So that's how I think about this question.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Models and then maybe do a pay as you go and then maybe the lowest tier models will be more of a subscription based, something along the line such that you can actually manage anticipated compute ahead of time and then manage the demand. But regardless, I think to me that is perhaps a bigger question to think about is how would this market evolve as we are seeing now increasingly not just clawed but other models that are throttling or changing pricing and github just I think some changes to their pricing and this idea of a la carte or sorry so this idea of all you can eat and just pay some nominal monthly subscription and consumers as much tokens as you can is probably going away so how do we actually have efficiency how can we actually have people use ai models efficiently”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Right now, given the capability and there's certainly a sense of experimentation, remains to be seen as the price of AI increases towards its actual cost of solving those models and also even potentially the value that you provide to the users, how much demand will change and there's that sort of demand illicity question. I have this theory that I'm collaborating with someone on Alec Bimas who is a columnist at Chicago Booth and we have this idea that AI pricing, we eventually evolve towards a multi-tier system where you have to somehow meet the, so somehow you have to equilibrate market demand supply given that there's a shortage you may have to basically auction off the most powerful”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Succession of such news throughout the last couple few years. Generally speaking, I think indeed we have this issue that I think is actually unresolved is that the price isn't right, right? Currently, AI is too cheap, right? People are just out there consuming token because it's heavily subsidized and they ask the most powerful model for cooking recipe. That's sort of your analogy, except that it's actually being used on a gentle level as well. The agents are calling AI AI models inefficiently. Part of the reason why people run up these huge bills using open claw is exactly because the system has not been optimized for token efficiency. And part of the reason why is because price has not really been reflecting the genuine cost. So there is that question. There's clearly massive amount of”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, I mean, I assume you're talking about the open AI, you know, so Wall Street article, right? About how they are not meeting the expected growth. First of all, that's old news. And also, I think we have had, like I said,”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“It's a bit like watching tanners having never held a tanners racket yourself that the moves you watch, Roger Federer make looks a lot easier, right? So, I think that's part of the reason why personally I think the AI cycle has been very undemonstrated is that it's the simple fact that most people don't code, so they underestimate just how much agentic AI deployment was going to amplify AI demand. And people who actually do code themselves probably don't think about these questions enough because they're not economists and not finance people in finance. So you have that sort of in-between area where I think we're just about finding out.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“And you have pointed it out yourself, you know, the compute is underestimated because most people don't code, right? The vast majority of people don't code. And the fact that AI has enabled people who don't code to actually interact with code is actually a little bit misleading in that if you don't code, you wouldn't ask questions that you will ask if you get code, right?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Last year when Claude Cole came along and you had Sonnet and then we had Opus and that completely changed the air landscape moment 2.0 where suddenly you have this recursive calling of AI tools. So instead of query chatbot where it's all of one, we're talking about computer language like one order of magnitude of one now you have AI calling itself which changes the magnitude of AI compute demand probably by hundredfold or potentially more depends on just how crazy you let the AI agents go and that obviously changes the picture completely and I think one reason why a lot of people underestimated the most recent acceleration”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Right, so if you think about the catalyst, right, we had ChatGPT and then all this built out and then GIPSIC moment came, oh, what if people always had this lingering fear? What if the Chinese just came up with a way where we don't need any of this heavy compute and it's just all algorithmic black magic and we don't suddenly find all these things useless and then came the very powerful intuition concept, economic concept given paradox, funny enough brought up by Satyam Dadala, which turned out to be exactly correct, right? But the more cheaper you make it, the more people want to use it and the more highway lanes you have, the more people drive a car. And then we sort of kept on progressing along that narrative and finally we got to this destination of agentic AI, right?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“The nuclear scientists, because they went to the core of the Earth was dying, and they wanted to resuscitate Earth and they went down there with a huge nuke and they realized that the problem was actually a lot bigger than the one nuke could handle they split up the nuke into multiple nukes and according to the succession of explosions that will actually amplify each other and that's actually honestly how I thought about the timeline of AI, right along the way we have had multiple moments of progress, disillusionment and then another progress, another periodic disillusionment that sort of became his own sustaining catalyst including by the way things that had nothing to do with AI, the macro bearish stuff I would argue actually helped AI cycle the last as long as it did such as Liberation Day and most recently Iran conflict”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“The internet upward is that the AR bubble is very much adopted and used by everybody almost right away, unlike the dot-com bubble where you actually had a lot of unused capacity that was at the end built out. So this one, you reminds me of that movie, I forget the name of that movie where basically to rescue the Earth's core, they had to basically blow up a succession of nukes. Remember that the core, I think that's the name of the movie.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“All of the stocks, the big beneficiaries of this bubble are actually coming from specific components input into the construction of data center that nobody who's today talking about these stocks had ever even thought about them. These are the most random things we're talking about gluing agents that glue together the memory chip with GPU. We're talking about random wires that will handle a low latency. And there was a lot of physics that go into it and the evolution of the compute and so on and so forth that creates basically a succession of bottlenecks and so on, which of course on the mirror on the other side by the actual utility of these tools, if there were no practical use and demand for these things, you wouldn't have seen the construction of these stones either. And that's one big differentiation between, I think, the AI bubble.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“I mean, of course, it's a bubble. I mean, it's going to bubble. It was always going to be a bubble, right? Back in 2019, it's free. Now, the question is not, I think the interesting debate is not whether or not AI was the bubble, will become a bubble. Manifestly was and had become one, right? The question is, how long is the bubble? How big could the bubble get? And at what stage are we? And I think that is the part that was very underestimated by people, but especially map for people. Once they heard the word bubble, they started dismissing its significance. Just because something's a bubble doesn't mean that you couldn't have a very large sustained impact. And especially if one that's actually much larger and longer lasting than people had anticipated. So we had ChatGPT. And this bubble really ultimately is underestimated mainly because it's a very”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, so over the last couple years, obviously the build out phase has to come before the phase where everybody starts using it. And, you know, we go into a golden age. And that led to a lot of discussion about whether this is the next great dot-com bubble. And to your point earlier, you mentioned how it feels like people might have over-indexed that to that a little bit, a bit of this like hindsight bias going on. Yeah, so I'd just love to hear how you think about that. But like, is there still a route where this becomes a bubble? What is even the definition of a bubble in your mind? Like, is this even relevant to the conversation? Just how do you think about that whole idea of the detractor saying that, basically?”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT
“Equity in a booming stock market that's been driven by the AI narrative and investment and evaluation has obviously led to a certain degree of wealthy facts which has led to consumption of the top half of the K-shaped economy as many people have alluded to. So the combination thereof, there's a direct and there's an indirect contribution to growth platform from AI investment sector. There's just not a lot of about it.”
2026-04-29 · Forward Guidance · The AI Bubble Is Widely Misunderstood | Steve Hou · IDENTIFIED FROM THE TRANSCRIPT