The Chips Are Down - Paul Krugman
Paul Krugman
SubscribeSign in
Playback speed
1×
Subtitles
Share post
Share post at current time
Share from 0:00
/
Transcript
SPEAKER 2
Hi, Paul Krugman here. I'm recording this on Tuesday afternoon. I just won't have time to write a normal post for tomorrow when you'll see this. And I would take the day off, except it seemed to me as if people might want some reaction to the carnage that's been going on,
at least in part of the tech sector and stock markets around the world, which has been pretty remarkable. It's really tempting to say that it's deeply meaningful. In general, you want to be very cautious about putting too much stake in stock market events. I'll come back to that in a minute.
But it is striking enough that it does seem to be worth commenting on. So what's happened? There's been a fall in tech stocks very much concentrated in semiconductors. The Philadelphia Semiconductor Index was down almost 8%. on Tuesday. The Kospi Korean Index, which is largely a semiconductor index,
982
126
123
The Chips Are Down
What, if anything, are the markets telling us?
Jun 24, 2026
982
126
123
Transcript
A short talk in lieu of a post. Back on full duty tomorrow.
Transcript:
Hi, Paul Krugman here.
I’m recording this on Tuesday afternoon. I just won’t have time to write a normal post for tomorrow when you’ll see this. And I would take the day off, except it seemed to me as if people might want some reaction to the carnage that’s been going on, at least in part of the tech sector and stock markets around the world, which has been pretty remarkable.
It’s really tempting to say that it’s deeply meaningful. But in general, you want to be very cautious about putting too much stake in stock market events. I’ll come back to that in a minute. But it is striking enough that it does seem to be worth commenting on.
So what’s happened? There’s been a fall in tech stocks very much concentrated in semiconductors. The Philadelphia Semiconductor Index was down almost 8%. on Tuesday. The KOSPI Korean Index, which is largely a semiconductor index, was down just about 10% sort of the previous day or the same day, you know, time zones. And there was a 2.2% fall in the NASDAQ. We’ve seen a lot of decline in tech stocks, things related above all to chips. What’s going on there?
Part of the answer is that trying to understand why the market does what it does is, generally speaking, a mug’s game. In this case, however, it does seem that part of what’s happening, probably a large part of what’s happening, is that the tone, the rhetoric surrounding use of AI, and hence the demand for compute, has really shifted quite a lot just very recently.
All of a sudden, we have a spate of studies that seem to show that, yeah, AI models allow people to churn out a lot more stuff, but the actual payoff to that stuff is much, much smaller than the volume of stuff that they’re churning out, most obviously lines of code, but just in general. AI lets you do much more, but how productive that is in terms of the ultimate goals of a business, let alone economic growth and quality of life is much more doubtful.
On top of that you have a rather abrupt, jarring turn in business strategy. Up until just the other day a lot of businesses were more or less whipping their workers into using AI — you know, we’re going to judge you on how much you’re using AI whether or not you really want to whether or not you yourself think it’s valuable. We’re actually going to score you, we’re going to require that you do tokenmaxxing.
And then, with compute getting scarce and with the price of chips having gone through the roof, suddenly the AI companies began charging and the marginal cost of using a lot of tokens became really, really very high. And suddenly companies were saying, oh wait, stop. We want you to economize on your use of tokens and hence to ultimately reduce the demand for compute. And that’s a sudden U-turn.
This is part of a broader phenomenon, which I’m going to write about very soon, which is that there is a kind of lack of organicness to the AI boom.
There are people who are using it because it looks great. They’re using it because it’s fun. I have colleagues who are just mucking around with Claude and finding some uses for it. But there’s also a large amount of Corporate America that thinks that this is the way it has to go. Fear of missing out, not by the individual investor, but by the corporate bureaucracy. And then pressure from the financial markets, saying, you know, your company better be on the cutting edge of AI or else. All of which is very fragile. It’s a kind of a bubble, but not in the normal sort of asset price form. It’s more of a kind of fad, almost a social delusion. And that, it seems likely, certainly got ahead of itself.
Now, I’m reading way too much into these stock prices. And so let me give you a little bit of a caution on all of that. So yeah, the Philadelphia Semiconductor Index was down 8% in a day, which is one hell of a drop. But it was up 157% over the past year.
So you want to have some perspective here. This is a stunning setback, but the fact of the matter is that over the course of a year, these stocks have been incredibly high-performing. The KOSPI, the Korean index, was down 10%, strictly speaking, 9.99%. But anyway, it was down 10%.
But after that 10% fall, it was up 172% over the year. So we’re not talking about a catastrophe. We’re not yet talking about, we aren’t even talking about a Bitcoin level of disappointment for investors. But okay, it’s a break in the trend.
The other thing we should say: the famous old line by my teacher and colleague, Paul Samuelson, was that the stock market had predicted nine of the last five recessions. There’s many more than that now. In fact, just over the course of the past year and a half, we’ve had two major stock market declines that turned out to be false alarms.
There was a big decline in April of 2025 after Liberation Day, the Trump tariffs, because there was a lot of people just sort of, it’s chaos, terrible things may happen. While the tariffs have been a bad thing, they did not cause an economic catastrophe and stocks recovered the losses that they experienced then.
And then there was another round of major stock declines associated with the Iran war. Of course, the Iran war has been a complete debacle and a disaster, and we’ll be paying a price for that for a very long time. But the consequences for short-run macroeconomics were more modest than many people, myself included, expected. And it appears that the Strait of Hormuz is going to gradually open because the United States basically said, okay, you win. It won’t literally say that, but in practice, that’s what we’re doing. So that is going to be over.
So it’s not that uncommon for the markets to react as if something terrible is about to happen and be wrong.
And so you really don’t want to assume — there’s a real temptation to assume — that because there’s so much money involved, a big decline in markets must be signaling that something is really very much amiss in the fundamentals, that where there’s smoke, there’s fire. And sometimes, no, there’s just smoke, no fire.
So this might not be that big a deal. But it comes at a moment when the rhetoric really has shifted. You can see that there’s just a kind of a walking back.
There was a really striking interview just the other day with Satya Nadella of Microsoft. Microsoft is actually a consumer of AI, rather than a producer. They have tools you can use within Microsoft products, but I think they run basically off OpenAI.
And Nadella was pretty scathing about saying, you know, we can’t give all of this power and all this money to the big AI companies, and we should be using cheaper models. And hinted that Microsoft may start making use of DeepSeek, the Chinese model, which is less comprehensive. In general, the Chinese models are less comprehensive, but immensely cheaper, and among other things, just do a lot less computation. That’s kind of the core of why they’re cheaper.
And in that case, the picture changes a lot.
What bearing does all of this have on AI and the future of the economy and AI and the future of humanity? Well, part of what we’re seeing may not be so much disappointment in what AI can do as realizing that this extremely compute-intensive AI is not essential.
And maybe you can still get whatever the big productivity benefits are and still possibly the big labor-displacing effects without quite so much compute. But it’s not entirely separate either. I think we need to be saying that this is what a quasi-bubble quasi-bursting might look like.
Take care.
Get more from Paul Krugman in the Substack app
Available for iOS and Android
982 Likes∙
Discussion about this video
CommentsRestacks
4h Edited
Thank you, Professors Krugman and Wells, for posting today, not because you have the time and energy to do it but because of your compassion and commitment to your readers. We’re raw and overwhelmed and your explanations of the way things are help us see things more clearly. That’s the first step towards rebuilding the country we were promised 400 years ago. I’m updating to include stickers and a link to an awesome photo, since sending happy notes was my very favorite task as a guidance counselor. 🙏🏅🏫 https://witnessla.com/paul-krugman-wins-the-nobel-in-economics/
Like (89)
Reply (3)
Thanks ISO, I too feel that way about professor K and appreciate your expressing them so well in this dark time. I'm 81 and although he may not be the light at the end of the tunnel, he is certainly a bright lamp.
Like (38)
Reply (1)
💯
Like (8)
Reply
I appreciate this comment ISOequanimity. Clarity is often underrated. Before we can solve problems collectively, we need a shared understanding of what is happening and what is not happening.
One of the most valuable things thoughtful writers and teachers provide is context. They help us distinguish signal from noise, fear from fact, and temporary turbulence from deeper trends. That kind of understanding is not the whole solution, but it is often where meaningful solutions begin.
Like (9)
Reply
Loved that link. Laughed out loud at "happily scruffy, wildly intelligent Paul Krugman". Certainly not the description I would have written, but kinda cute. 😃 And this quote: "Mr. Krugman’s fame as a public intellectual should not lead anyone to think that they understand his contributions to economic research just because they regularly read his columns." But we're TRYING to understand, with the help of Dr. Krugman's clear writing!
Like
Reply
This point made by Paul is spot on, that this is not so much about AI bubble bursting because there is nothing there--there *are* some interesting applications of it.
Instead, this moment proves that our "Masters of the Universe" in Silicon Valley gambled wrong, on believing that AI could only be effective with massive compute--with huge sprawling datacenters churning away--while China has been building what we ought to have pursued: smart, smaller, more efficient models and targeted use applications.
China, by the way, just announced some careful loosening of regulations they set up early on for use of AI and sharing of the open source LLMs (data models, the core of good AI) between their AI companies. The clarity of how their state will be treating AI legally and where guardrail and enforcement of law and protections for consumers is very clear to companies. Not so much here, which ironically is making our companies more cautious as well as exposing us to some really ugly and stupid results, including pumping out child porn and Nazi sentiments.
Our greedy little men in Silicon Valley--no geniuses to be found among them--went for the big iron servers, because they wanted to pump up NVidia and their hardware investments and try to lock in America's (as it turned out, very brief) lead. They resisted deregulation, even when it involved national security. They created proprietary LLMs, so there is a lot less cooperation and sharing of results between the companies competing in this arena. And they created fat, bloated AI systems that spit out a pittance. This, by the way, was being pointed out by talented programmers and techs who know their stuff for quite a while now. Media in their frenzy, and bought out by the same tech tycoons, dismissed the obvious.
We see the same thing in the way our military is forced to wage war, lumbering in with big heavy expensive, unwieldy machines instead of light, cheap, swarms of drones.
The United States is no longer light on our feet and we are not being lead by smart men (much less women).
It is time to change things up. Starting with our leaders.
Like (55)
Reply (5)
It’s hard not to think of how Wiley Coyote can keep running for a while after going over the edge of the cliff. The value of all crypto is about 1.8 $tr. The ballpark debt and equity of all companies on NYSE and NASDAQ is 90 $tr and 90 $tr. So crypto is 1% of that 180 $tr total. It could easily go to zero without doing any real harm.
The role of the Roadrunner is played by the investment banking industry, and it’s the real value of that 180 $tr that’s the issue.
Like (17)
Reply
AI is nothing new. Remember ELIZA and Joseph Weizenbaum! The current brute force approach to the mimicking of natural human behaviour may turn out to be a massive wrong turn in a story that had been evolving in the interface between computer science and neurology for over 50 years!
Like (11)
Reply (2)
2h Edited
True, AI has been around for a long time, but mainly as a curiosity.
It's turned a corner now, but I don't think any of us quite understand yet exactly how best to use it. I've also noticed a lot of these 'geniuses' keep claiming wins for AI that really are at heart not much more than the grinding work done by data analysis that has been around...well, forever? Just now scaled up to the point where it kind of merges with what AI does with pattern recognition.
Totally agree about the wrong turn. Now we have to claw back money and resources these idiots threw at the Data Centers, and demand common sense government regulations and guardrails. It wouldn't hurt to use the open source LLMs that China's DeepSeek and their AI companies are using either, and contribute our own into the mix.
These days I think a lot about the early days of the Internet with Yahoo and AOL, then Facebook, Google, and Microsoft fighting over owning people's online "identity"--as if they could own that. All that produced was a mountain of proprietary lists of users and multiple logons for each individual as they fought over whose user lists should 'rule'. What a mess that yielded for the end user. It is why our individual data is constantly threatened, adding to the confusion over who owns what.
Might be a good time as part of the AI regulation push to standardize digital identities and return the control of each person's online identity to the individual user, not a corporation or even the government. Government can maintain the infrastructure, but the individual controls permissions, decides and controls who gets to use their identity when (including all data generated by that identity that AI would love to use), and for what purpose.
Would greatly increase security and simplify *all* our lives. Stop all the unnecessary middle man churn, with companies continually reaching in and grabbing our data and daily activities and then using it for things we never agreed to at any point in our lives. They are so damn sloppy about storing the copies of our activities and data, and will never care as much as we do about the consequences of using it.
Like (4)
Reply (1)
Totally agree with you point on AI regulation & digital identities. I hope that is where this is going.
Like
Reply
Just as the Neanderthals made a right turn into frozen Europe - younger more adaptable homo sapiens superceded them - now the Neanderthals are just interesting DNA history...
Like (3)
Reply
Totally agree. Seems no sector of America is untouched by our assumption that and adoration toward “bigger is better”. Calling it “AI”, or should I say being *allowed* to call it AI, has also been a driving factor in the bubble. I have wondered if “LLMs” would have had the same rapid uptake. Doubtful. It has very much felt like a fad from the beginning. I have been shocked by the FOMO though, particularly by businesses. You can hardly turn around without bumping into it. Like protein. It’s been added to everything whether it makes any sense or not. None of us asked for protein in our water or “AI” in a fridge. Some of what it has done is cost a bunch of money, suck up resources, cause unnecessary stress for employees and customers, slowed down websites, made up endless incorrect answers and slop, and stressed communities. I, for one, am not at all surprised that “AI” is not what these tech giants tried to make it out to be. Unfortunately, we’re all living through this stupidity.
Like (6)
Reply (1)
I believed this until recently when my skeptical coding progeny told me it was now really what it had been cracked up to in productivity terms. Both are very concerned about the societal implications.
Like
Reply
well said
Like (1)
Reply
Good corporate governance relies upon good leadership.
We inhabit an adversarial system. Regulation offers by negation, guideposts that inform corporate governance. In a word, leadership.
Thoughtless deregulation is therefore the abandonment of leadership.
In its absence, corporations act stupidly. They’ll readily follow one another over a cliff. Indeed, they must!
Otherwise they’ll be eaten, and their new owners will drive them there anyway.
Like
Reply
I think of AI like the advent of the MIDI and music synthesizers in my lifetime: while one person can now imitate both Oboe and flute, both trumpet sound and percussion, full scale operas are not dropping from the trees. Its hard to usefully use something as capable as the current frontier AI models. Its interesting, while actual thinking and promotion of public schools, research and higher learning are under direct attack from Pro-Wrestling profit takers, like Toll Troll Trump, the 'magic window' the idea that 'you can do it all', and upscaling your business and personal life from the Imaginariaum of your own ad hoc creation-- surprise, does not monetize. In the end, it is end users that will scour and explore and find $ uses. It is greedy monopolies and monopsonies that will harvest the $ windfall. We even have to use their platforms to have access.
Like (30)
Reply (1)
I suspect that true professional craftspeople, like the folks in opera, will actually become more in demand as people see real, lived experiences in communal settings.
A lot of tech, especially social media is terribly toxic. I've got a kid immersed in social media hype and nastiness. It's bad, really bad.
AI shouldn't be allowed to use personal pronouns.
Like
Reply
Thanks very much for comments. My gut tells me that AI is probably both overrated and somewhat dangerous to a healthy future for humanity. I also think humans should be much more focused on addressing the really dangerous things that we are doing to the biosphere!
Like (25)
Reply (1)
I appreciate the comments that I have been reading. Obviously AI is one elephant in the room. Hard to imagine how many more can fit in one room, but the moron's occupying the oval office will come up with something soon. This may be my opinion and people may not agree with it and I'm fine with that, just let me know how you don't agree. (I believe that's the Barrack Obama approach)
AI centers that want to move into communities should be allowed to do so, provided that they provide all of their own infrastructure required to operate them without encumbering the communities in which they want operate in. For a long time, maybe forever, tho modus operandi of business is to use whatever they need to produce whatever they produce without having to replace or repair whatever their business destroyed doing business. A couple of examples: PFAS. Just let them go into the environment because it's cheaper than taking care of the problems they cause. Plastic bottles and bags. The astounding detrimental affect on the environment because over use with no end game to eliminate the possibility of this problem occurring. The use of oil gas and coal. Emitting toxic chemicals and CO2 into the atmosphere causing climate change and health catastrophes. To my point; AI has many issues that will someday result in some sort of tragedy that will force us to deal with. Arguably, all advanced technologies have had their tragedies. Nuclear electrical generation, chemical manufacturing and storage, use of oil and coal. flying into space. I am not suggesting that these new technologies should not exist or not be used. I am suggesting that these advanced technologies, from past experiences, will indeed have a catastrophe attached to it in one way or the other. Back to my point, One way of avoiding catastrophic effects in communities where these centers are placed is to require them provide all of the infrastructure required to operate them. Both the effects of the technology itself and the environmental impact of the technology must be dealt with. Both need to be talked about. The same applies to bit coin mining operations and probably other business endeavors.
Like (2)
Reply (1)
Nothing has changed - the polluters reap the profits and externalize the huge environmental costs.
Like
Reply
Well to the extent AI /Grok has been running the Iran war would a business want to go that route? Also given the vulnerability of Taiwan betting on chips sure seems optimistic
Like (19)
Reply (1)
Right on!!
Like
Reply
Before AI, we had to produce a lot of business documentation that nobody ever read. AI helps us write documentation that nobody will read much faster. So there's that.
Like (15)
Reply
The AI market is maturing. Like all other products, there will be different ones for different price points. The market acted like everyone needed the AI that the tech sector needs. They, don’t, and the market will adjust. Dot Com bubble (n).0 has just started to burst.
Like (12)
Reply (1)
My grandson graduated high school yesterday. I was struck by a comment the valedictorian made, "a key part of maturing is answering the question, 'what do you want to be when you grow up'"
I suspect that a key aspect of the AI market maturing is that the market is getting a better focus on what AI will do profitably when it grows up.
AI that Translates English to Spanish, transforms words into visuals and visuals into words, navigates terrain, and teases out patterns from complex data, provides real benefits. But these benefits can be accomplished with AI models that focus on their specific functions.
So AI maturing as a market means that AI investment will become more focused, not that it will burst. AI investment schemes that were a scam to start with are an exception.
Like (3)
Reply
Did the timing happen to coincide with the headlines about Meta’s internal data breach?
Also, thank you for the word “tokenmaxxing;” just stored in my personal lexicon.
Like (12)
Reply
Wouldn't it be ironic if expensive AI were just another fad? A Pet Rock of the computer era?
Like (11)
Reply (1)
Like (5)
Reply
Crypto - mainly for crime
LLMs - mainly for cheating in school and burning silicon in the desert
Great work O genius capitalists and financial wizards, what wonders you have wrought for we W2 normies
Like (10)
Reply
An old friend's tongue-in-cheek motto for his construction business is, "We're good enough for people like you." Sounds like that might apply to the AI whoopdedo, also?
Like (10)
Reply
The Chinese models will prevail because they’re more efficient… similar to drones being more effective than fighter jets and battleships
Like (9)
Reply (1)
CHEAPER!
Like (1)
Reply
Perfect timing on this essay. Investors wanted to believe that AI is the next big tech thing, that it will be another world wide internet and create the next Google, or Amazon but I'm not sure it will. For one, it's not promising to create lots of jobs but to prune them, which might excite investors but doesn't send good vibes in general to our society.
Like (8)
Reply
Thank you for the excellent analysis. It seems that parents and teachers in LA are against screen time for their kids. It’s related to AI also. LA school district/system is banning screen time really limiting the time on the screen( banned from PreK to 2nd Grade)
Like (8)
Reply
What are the markets telling us about the tech sector (AKA AI)? Basically that they haven't a clue as to what's going on. Folks who can't write a line of Python code are second-guessing the geniuses who founded and run these companies. You know, the guys who daily meet with tech experts, business and government leaders, CFOs and CTOs; who get financials daily, in short, have up to the minute complex and important data. These are the guys who take their vaccination advice from RFK, Jr. and celebrities touting dietary supplements for memory improvement. Guys who don't know the difference between a NAND gate and a NOR gate architecture for memory chips, and how they differ really know what they are doing. Sigh!
Like (8)
Reply (1)
1h Edited
I agree. I dont think anyone else has a clue about what AI is about. Paul Krugman admitted he had no clue recently. Clueless County Councils dont care, they have approved thousands of massive data centers without the slightest clue about AI, without any environmental review or transparency, investors without the slightest clue about AI poured money into these apparently purely on speculation, thinking only about short term. All of these players in my view are verging on criminal recklessness. Maybe they think whatever bad may happen its ok -the gov will bail it out?
Like (1)
Reply (1)
No, sorry you are wrong. I and many others have more than a clue, we have knowledge. The true potential of AI is its fusion with robotics. Machines with near-human reasoning capabilities. It will end the 300 millenium era of human labor.
To prove my point, I encountered Nvidia in 2015, and had a vision close to what is now happening (it helps to have a strong background in physics and mathematics). I was so sure that I cashed out 25% of a portfolio that was generating 35-38% annual returns and placed the money on Nvidia's Pass Line. Left my chips there, and now that investment is worth over 220x what I invested. Went from the 2%-ile in net worth to the 99.96th %-ile. Today, roughly 140,000 Americans have a greater net worth than me (0.0004 x 340MM = 136,000). Wanna get rich? Invest in NVDA, AVGO, SMCI, MU and SK Hynix. And AI, with or without robotics will not be a threat to humanity, but a boon. Why, because we won't let it cause harm. The cottage industry folks caught in the Industrial Revolution of mass production suffered, but today that development has lifted much of humanity out of poverty. Read "The Midas Plague" by Frederik Pohl. PDF online available.
Like
Reply
How about the availability of memory? What impact will that have? Read an article where Tim Cook was saying that there will be increases on the price of iPhones now because AI is eating up all the memory chips and they're less available for other devices.
Like (7)
Reply
There is so much wrong with the stock market, but more wrong with the SEC. It's the Wild West, with Elon leading the pack; valuations in the trillions with revenues in the low billions. Now Antropic et al saying that they'll go bankrupt without hundreds of billions in revenue: are they lookiing to the government for that? Most likely.
Like (6)
Reply
Typical American corporate strategy: rush into something that is fashionable, make a mess and then collapse.
Saw it with Ford, GM and Stellantis writing off tens of billions of dollars in EV investments.
saw it with sub prime mortgages. and now AI.
in the 1980's Chase's Nick name in South Africa was Chase in, Chase out.
Like (5)
Reply
Weren't tulips in 1637 more or less a fad and a social delusion?
Like (5)
Reply
The big AI wave was around bigger/better/faster which demand huge resources. China, with its limited access to the latest&greatest chips (but plenty of electricity), went for efficiency with DeepSeek. And as Krugman points out, that may be the winning strategy going forward.
(classic product development is about growth at all costs, then consolidate and win by being more efficient. Seems that is what we are seeing now.)
Like (5)
Reply
Thank you for your analysis. I still find it unsettling that AI got past the research stage without the tech companies figuring out how to scale up without using such enormous amounts of our natural resources. I am not going to subsidize these companies. I switched to a different browser and opt out wherever possible. Small things in the grand scale, I know...
Like (5)
Reply
Do I really have to start using "compute" as a noun?? Sigh. Ugh.
Like (4)
Reply
When did "compute" become a noun?
Like (4)
Reply (1)
The first I remember "compute" being used as a noun was with the advent of cloud computing in the early 2000's. The cloud separated buying the capacity to do what computers do ("compute") from buying computers (the things that provide that capacity).
However, I was surprised to learn from the Meriam-Webster dictionary that "compute" was first used as a noun in 1483.
Like (1)
Reply (1)
As a verb or noun?
Like
Reply
Absent AI investment, the markets would be down. This includes investments in building things like data centers and some power generation. So without AI the economy is sucking wind right now. Chip indices on the skids suggest a pullback in demand for compute which is a stand-in for economic growth. Me thinks, this is the start of another dot.com bust.
Like (4)
Reply (1)
And just like that: poof.
Like (2)
Reply
The Nadella/DeepSeek angle is the real story buried here. We might be looking at a 'DeepSeek disruption' — companies pivoting toward cheaper models because they've realized productivity gains only make sense if the cost to achieve them is economically sustainable. Take Oracle, for example: it just cut 21,000 jobs and posted -$23.7B free cash flow to fund AI infrastructure — backed by a $638B backlog, mostly tied to OpenAI, that assumes premium compute pricing holds for years. If cheaper models like DeepSeek shift enterprise demand toward lower-cost compute, what happens to companies that fired people to build capacity priced for a market that might not pay premium rates much longer?
Like (4)
Reply
Market drop? Must be the algae.
Like (3)
Reply
Thank you, Professor Krugman. I've always felt that AI was not much of anything.... maybe a C plus on an English report card.
There is no humanity or sincere emotion to it - just a boring recitation of "facts" that may or may not be true.
Like (3)
Reply (1)
You may want to download Claude and just take it for a spin.
Like
Reply
Could there be some of this as well? Not just my experience, of course, but accumulating doubts about the reliability of AI?
I'm just writing history - its stunning to consider an AI agent flat out lying and making stuff up when a doctor asks it to read an x-ray.....
Like (3)
Reply
Is it possible there was some insider market manipulation?
Like (3)
Reply (1)
There is some market manipulation. I don't know how much of the AI investment is accounted for by Tech giants passing the same dollars back and forth amongst themselves Enron style, but I suspect the DOGE gutting of the SEC Sarbox enforcement is meant to ensure we can't see it.
I try to read news from different parts of the world in English, French, Spanish. I see a marked difference in the analysis of AI inside and outside the US.
Outside the US there is recognition that AI is a really big tech change, but it is on the same curve as tech in general.
Inside the US, AI is seen as the monster that reversed a recession by massively building data centers and will take everyone's job.
Like (1)
Reply
What do you mean by the use of the word 'compute' I'm not understanding how you're using it there at all.
Like (3)
Reply (1)
Computational power to do operations
Like (9)
Reply (2)
Debased language for the AI generation
Like (4)
Reply (1)
Not unlike using impact instead of effect or affect - too lazy to aptly use affect or effect
Like (2)
Reply (1)
I think for most people it's not laziness - it's not being sure of the difference between affect and effect...
Like
Reply (1)
Not learning the difference is lazy as is "more unique" - there are no degress of uniqueness
Like
Reply
Used to call it "capacity" also. Throughput was used to describe speed. Compute is another shorthand.
Like
Reply
I have read many glowing articles about AI in places like “The Economist” and dabbled a bit with a couple of the models. The fundamental problem I see is that you can never be sure when these tools are going to hallucinate (i.e. make stuff up out of thin air). That means that whatever output you get from AI needs to be rigorously cross-checked for accuracy before using it in any meaningful way. That takes time which obviously detracts from the productivity gains one would normally expect to see. When/if that problem is solved by the various devlopers, I suspect we’ll start to see more tangible benefits, but until then I put most AI in the category of “interesting science fair projects.”
Like (2)
Reply (1)
I have had the same experiences with the AI hallucinations.
Like
Reply (1)
I started out the way I suspect most people do. I asked questions to which I already knew the answers. To be fair, most of the time I got back about what I expected but perhaps 10% of the time what I saw didn’t fit with my knowledge base. When I delved into those findings, sometimes I got legitimate information I had not seen before. However, other times when I tried to check references, I learned that the sources the AI model cited simply did not exist. That was a useful lesson for me. Now, if I query an AI model about anything of importance, I always check the references before trusting the AI’s response.
Like
Reply
I am surprised that the professor has not considered a different perspective on this situation. Capitalism in its most ruthless form—toward which American capitalism often seems to move asymptotically—has historically sought to corner and monopolize emerging technologies. One only has to look at the histories of railroads, automobiles, petroleum, and countless other industries. The pattern is familiar: exuberance, concentration of power, and speculative excess, followed eventually by a collision with reality.
In the end, however, reality catches up with everyone. We are all in this together.
Like (2)
Reply
In the end, it is human creativity, ingenuity, ethics, and how we put tools to use to serve humanity that will matter. AI is useful, like email is useful. Email started out expensive, and now it is free. I suspect the same thing will happen to AI, and the compute data centers will shrink to useful size here in these United States. They have already shrunk in China to being useful, not overwhelming. China leads the way now on the AI front, because they are pragmatic, not massively, overwhelmingly, insanely greedy--which US-based businesses absolutely are where AI is concerned.
Like (2)
Reply
3h Edited
Yes, I agree that there is a wave of negative sentiments around the use of AI. As an AI user, as l see the field improving every day. I can be more creative and find more ways to work with a smart thinking partner (AI), creating spreadsheet models, analysis frameworks, and improving these every day. Bottom line, I think we are at the beginning of an odyssey here. Not everyone will be a power user, but the results will ultimately be transformational. The major labs are learning all along too, and a few are atune to what users are looking for and doing with it, pushing for more innovation.
Like (2)
Reply (1)
How are the results validated or does one assume the output is correct?
Like (1)
Reply (1)
TIP: I have found Gemini very good are writing formula and macros (JavaScript) for Google sheets. These are very ordinary tasks like assembling accounts payable, and preparing a schedule for volunteers at church. The macros boost productivity by providing popup input forms and formula that process arrays rather than individual cells.
KEY POINT: I am not using AI to generate the results. I am using AI to improve the logic that gets the results. If Google suddenly boosts the charge for using Gemini... I can keep using formula and JavaScript even if I stop using the AI.
Like (2)
Reply
When personal computers arrived they were supposed to make earth-shattering changes to the way we work. They did. But not overnight. The same will be true of AI. Right now people are playing with it, experimenting with it, throwing it at tasks for which it is well-suited and for which it is poorly suited. That will even out, people will find their footing with AI, and it will transform how we work and how we live. Some of that transformation will certainly be disruptive just as PCs were. If you're old enough you remember steno pools and offices full of accounting clerks. That's all history now. Most of those jobs were held by women and the disappearance of those jobs has helped women blossom in other areas and become much more important members of the business community.
Satya Nadella may want to cozy up to some bargain basement AI, but that says more about his lack of vision than it does about the future of AI. I'll stick with the AI agent I use and pay for out of my own pocket. If Nadella finds quality AI too expensive, maybe he ought to spend some time looking at how his employees are using ... and misusing it. For me AI is a bargain. If you're judging workers by how many tokens they're using, you are a moron.
Like (2)
Reply (1)
A bit defensive?
Like
Reply (1)
I have no dog in the fight. My portfolio doesn't even include any AI companies or semiconductor manufacturers other than through index funds. I'm just calling it as I see it. I've used AI, understand its limitations, but also its strengths. One can cover their head and hope it goes away, but that just isn't going to happen.
The industry will shake out, just as the computer industry did. Remember Gateway? Digital Equipment? But the best and most adaptable will survive and grow because the technology has real benefits to offer.
Finally, I've run businesses large and small. The most important lesson I learned is to be very careful what you incentivize because that is exactly what you are going to get. When companies incentivize their employees to burn tokens, guess what? They're going to burn tokens. That doesn't mean they're using those tokens effectively. The second most important lesson I learned is that you don't win the World Series by hiring the cheapest pitching staff. You hire the best people you can find, place them where they can be most effective, and incentivize them to achieve the company's goals. The same will prove out for AI.
Like
Reply
3h Edited
Thanks Paul, I very much appreciate your well informed reality-check on this. And yes, a friend pointed out the Semi’s are now back to where they were in May.
Like (2)
Reply
“It’s more of a kind of fad, almost a social delusion.” you perfectly described AI in one sentence.
Like (2)
Reply (2)
“Well, part of what we’re seeing may not be so much disappointment in what AI can do as realizing that this extremely compute-intensive AI is not essential.
Like (1)
Reply
2h Edited
Yes, but! AI hasn’t changed overnight. Humans are doing what humans do: question. Which is not to say the questioning is frivolous. It’s all part of our process. Which is not to say questioning might not be dangerous. We might question ourselves into a dark corner.
Like
Reply
For those of us who remember when every day the evening news would report why the market was up or down, those of us who worked in financial services mostly ignored the why and focused on the long term. So it is too early to panic but it is also useful to not what Professor Krugman said here - especially that companies may be backing off AI just a little.
Like (2)
Reply
AI stocks have had a good run. Valuations require continued rapid profit growth to sustain them. The fear is that lower energy, lower computing power AI models can provide 80% of the benefit at much lower cost. This is yet to be proved, but someone will, soon. In which case, the massive data centre with its own grid generators, will be a white elephant.
Like (1)
Reply
I suppose a bigger question to those of us who aren’t meme stock buyers is the degree of collateral damage to ‘normal’ shares - how far the ai bubble tentacles have spread into the markets in general
Like (1)
Reply
Please remember that both patent law and copyright law say that what AI produces isn't protected if the AI does the heavy lifting creatively. AI can be used as a really good quality reviewer though: "Please describe the strengths and weaknesses of this document. Additionally, from the perspective of a college level composition instructor please identify how the document could be strengthened and corrected." Asking the AI to do the writing though generates something that the AI would get the patent rights on except that we haven't had our Commander Data moment yet, so the product is not protected.
If AI produces unhackable code, it may as well be published in a journal article for all to see. Good luck trying to keep it hidden as a trade secret.
Like (1)
Reply (1)
And AI can hallucinate and create non-existent data and journal articles. Who guards the hen house?
Like (1)
Reply
I agree that trying to understand market moves is generally a mug's game, but it keeps drawing suckers because, a few times a year, the explanation is obvious. (This is not one of those times)
Like (1)
Reply
We have to remember there’s nothing inevitable or innate about an economy, let alone a particular flavor of it. It’s a human construct and as such extremely subject to the will of those humans. We want to believe that something has value and will continue to have value. Until we don’t or it’s ruined. The IPO of SpaceX reaffirms this. So what? So a slight dip in the market aka traders aka gamblers means they’re betting they can get money a little differently today. But I don’t see any crack yet in the will for AI to be a business magic bullet.
Like (1)
Reply (1)
Except that the veracity of the various AIs are now suspect with their accompanying disclaimer “AI results could be wrong, please check”. The reality is setting in: the product is derivative/imperfect. That’s no panacea.
Like (1)
Reply (1)
I've written this here many times before, but it's fun to read about lawyers who have used AI to write their briefs. Esp. when you're a retired lawyer who can't imagine just off-loading your job to a technology and then not even bothering to see if the technology knows what it's doing... Judges are pissed......
Like (1)
Reply
Ed Zitron has been explaining the empty promises of AI for some time. It doesn’t actually do much of use and the business model is impossible. Gonna be a lot of cheap warehouse space on the market in a couple of years.
Like
Reply
Thank you for being clear ☺️
Like
Reply
As one who spent a career in computers and biotechnology and loves new, creative ideas, I have to say that the AI craze reminds me of the California gold rush. The people who will ultimately make money off this will ultimately be the equivalent of those who sold blue jeans and shovels—chip manufacturers and builders of data centers.
As many others have noted, once the real accounting is done and businesses realize how much has to be charged to make the system profitable, AI doesn’t seem like a such a business-friendly replacement for skilled workers.
Like
Reply
Seattle has been on the cutting edge/bleeding edge for information technology for the last 30 years. Wildly hundreds of thousands of highly paid IT people. AI is widely believed to be involved in the massive job loss that is going on in our region. Almost all of us know lots of people laid off because of what is suspected AI related job-loss. Note - no real drop in housing prices, but the market is just a bit softer than before.
Like
Reply
Reading this makes me very happy that I am retired as I know that the corporation for which I worked would have jumped on the AI bandwagon and we'd be using it even if it didn't make the work flow faster or add value to the product.
Like
Reply
For a contrarian suspicion...there's info coming out in places like the Economist that the Korean chips...and other chips are in high demand and it exceeds supply and will do so for a couple of years. There's a possibility that the big boys are cratering the market so they can swoop in and buy up stuff cheaper.
Like
Reply
Aligning products/services to be offered with actual customer needs is a good idea! LOL.
Like
Reply
Call me dumb, but we weren't really very good at getting a lot out of the personal computer early in its life. As I recall, Excel & Word & Powerpoint were the key to opening up the 'personal computer' everybody was forced to have at there desk prior to the interwebs. And then when that came along, Katie bar the door, so did the completely surperfluous and unnecessary webpages - remember the dancing beaver?
I only say this because with AI it's going to be the same - a lot of mis-steps I use it to help with my coding, and it has been incredibly useful for that, producing routines I can modify based on ideas that I never implemented on my own. It's good for that.
Much like the computer contained the structure for what would become the internet, I predict that whatever significant 'product' AI gives us has yet to be imagined. But it is there, inherent in its conceptualization.
Like
Reply
When it comes to technological predicts versus the real world I often the fact that, back in the 70’s and 80’s, engineering models of nuclear plant failures, that depended on failure rates of valves and other components propsgsting through the system, predicted an accident once in a thousand operating years (or at least very small.) The reality was much more accurately predicted by insurance underwriters who looked at the reactors as complex industrial systems, like refineries or chemical plants, not as a collection of components, and accurately predicted that accidents would occur much more frequently.
The essential difference in the predictions is that accounts, perhaps indirectly, for human error and systemic deficiencies that are not reflected in a concatenation of component failure probabilities.
Predictions of AI productivity gains run up against the same problem: people. People must direct the AIs and decide when to use them.
Humans are creatures of habit as well as being highly fallible. I’m not sure AI can overcome that quite yet.
Like
Reply
My view on AI effectiveness changed two weekends ago when my family got together, including my son and his wife who are both coders/builders. They had started using the newer Claude models at work and playing with it at home. Both of them said it was fantastic. They could do far more work than before. Much of what they were doing was learning how to use the prompts. I watched my son making an application to help him keep track of things in a book he's writing. It wasn't something I could do, but he has the expertise to know what it's doing and know what he wants it to do. Both of them don't like the societal implications and have safety concerns but last January they were very unenthusiastic and now it's a big, exciting tool.
Both of them think it could replace a lot of workers.
Like
Reply
Chips were seen as a "shovels in a gold rush" business, a safe bet, hence that +157%.
But, what if they're beginning to just pile up in the warehouse? Data center build out is lagging. Look up up backlog in key components like diesel and gas turbine generators. Then there's serious and rising public opposition, power and water constraints. I wouldn't "buy the dip" here.
Like
Reply
"part of what we’re seeing may not be so much disappointment in what AI can do as realizing that extremely compute intensive AI is not essential."
Although I am not an academic or analyst, but rather an old(er) retired firefighter with this economics itch, personally I'm having a harder and harder time keeping up with the fire hose of change and information. I have come to the same conclusion as you, professor, and Satya Nadella of Microsoft, just from using the simple tools of communication and word processing that Nadella's Microsoft keeps jamming full of more and more features that are not needed - at least by me and I suspect by many others. Nobody wants to be left behind. That's why I follow you religiously.
Interesting discussion.
Like
Reply
Paul, Thank you for posting. You are always a welcome voice.
Like
Reply
I am a mechanical engineer /contractor. 70, still working . I have just started using AI for day to day reference type things, refreshingy memory on math concepts plus things in my personal life ( two cycle outboard repair, piano practice tips for particular passages) ....it is incredibly useful. It helps to know about what the answer should be , which is an engineering thing....does this sound right? This Is something an engineer asks everyday because it is so easy to slip a digit or fat finger something , or face it , just be wrong with the wrong approach to a problem .
The AI is like a sounding board that allows you to make way more in depth decisions and analysis, because , face it , an individual only knows a very tiny sliver of what is globally known on a topic AND what is globally known is a very tiny sliver of what is "out there ".
Like
Reply
Can I just say, the title. 🔥
Like
Reply
Or people are dumping tech stock with the expectation that SpaceX will be in their index so to avoid overweighting.
Like
Reply
How much does market manipulation play in these ups and downs?
Like
Reply
Thank you. I sincerely believe that it was forced upon us by Trump. That brand new shiner thing in the window, so to speak. The tech bros got him back into office. Promises were made, but nothing was to scale yet.
The worst case scenario is the land grab, the water and power expense and communities disrupted, all to bring about something that was not ready for prime time.
The American public has been fooled again, and sadly, we will be paying for a long time as these people, those responsible, charge off to private islands and bunkers. May they all be received as the Kushner were in Albania.
Like
Reply
The biggest problem with AI is that the purveyors/profiteers are shoving it down our throats without consent.
We need to slow down the implementation to get a grip on the various impacts it will have on our society, and figure out how to deal with them.
Like
Reply
Very insightful professor krugman.
Like
Reply
Thank you Paul.
Like
Reply
I love the Samuelson quote about stock market based predictions of recession.
Like
Reply
It feels like the stock market has gotten too big to fail in a truly major way, given the mind-boggling amount of money that's been created out of thin air - money that is now sustaining our economy, pensions, house values, etc. However, minor failures have become almost routine, , and - for the financially nimble at least - reliably profitable. With most of the market controlled by a relatively small number of sophisticated investors (Vanguard, Blackrock, etc.) this no longer seems impossible.
Like
Reply
Paul krugman predicted 50 of the last 2 crashes.
Like
Reply
Fantastic presentation, Professor Krugman. As always, clearly presented for the novice crowd of which I am front and center. In typical fashion, its written in a Jack Webb style, "All we want are the facts, ma'am." No clairvoyance, but just solid explanations to help guide us toward what to expect in this increasingly fast moving and sometimes uncertain future. I think independent media suits you even better than the Times quite honestly, and think it provides greater reach. You became familiar to me through frequent Bill O'Reilly appearances enduring his pompous spectacle. And now what happened to him? Fired from Fox, ostracized for a while, and now on the third rate televised scandal rag / Fox Lite "News Nation" with other disgraced individuals, including being paired with Chris Cuomo. The lesson from your success is as Mrs. Obama wisely said, "When they aim low, you aim high." Cheers, Professor. Without knowledge, we are nothing.
Like
Reply
Wait a minute, MicroSoft has Co-Pilot. That’s a full on AI LLM. Everytime we have a TechHub presentation at my work that’s all they want to talk about aside from cybersecurity.
Like
Reply
Thanks Dr Krugman
Like
Reply
"This is part of a broader phenomenon, which I’m going to write about very soon, which is that there is a kind of lack of organicness to the AI boom."
I can't wait for this, because this is a thing I've been thinking for a while. Everybody's just sort of decided based on no evidence that AI is the future.
I talk to some lawyers who swear that AI makes them more productive and basically none of them can ever explain what they actually DO with AI (that is actually part of a lawyer's job description; I've heard it does a lot of things that most lawyers would be outsourcing to interns, for instance.)
Like
Reply
Way back in the 1980s when the design professions were presented with the unproven, and costly, choice to move from drawing boards, lead holders, and erasing machines to computer aided drafting (“CAD” — so called because that was really all it was) the choice was driven not by the practitioners themselves, but demand expectations of clients. The transition was painful. Another 20 years would pass before design productivity using CAD reached the level of productivity “on the boards.”
Like
Reply
2h Edited
I realize this is off topic nevertheless appropos the dire straits we're in because of the self indulgence of Mango Loco.
A moment to reflect with a bit of humor during the machinations of madness besetting us:
The following is of course from the original cartoon genius: Bill Watterson
Calvin: you see Hobbes I have a water balloon, and u don’t
Calvin: therefore I have weapons superiority. You have to do what I say
Hobbes: I think I’ll take this stick and poke your balloon
Calvin: that’s the trouble with weapons technology.. it becomes obsolete so soon
Like
Reply
2h Edited
A couple of decades ago, I began applying chaos mathematics to study stock-market volatility, and published the main results in 2023 in the Journal of Strategic and International Studies [ https://www.researchgate.net/publication/371445241_Efficient_Markets_and_Chaos_A_Perturbation_Approach ]. I specifically looked at DJIA historical data from just after the Great Depression to just before the COVID-19 disruption.
I assumed there was a baseline of exponential growth from steady technological development, and looked at actual deviations from that baseline, then used a Zipf's-Law test, which is one of the oldest, simplest, and most robust tests for chaos (i.e., the butterfly effect), and found that approximately two-thirds of the index's behavior can be attributed to technological advancement, and one-third to pure chaos.
The thing about chaos is that it is predictable in the short run, but increasingly unpredictable as time goes on. That's what confuses market watchers--short-term tests of market-prediction strategies work, but fail when applied over the longer term, and it's the longer-term predictions that promise the biggest gains. OOPS!
This result confirms that the long-term value-stock strategy with a diversified portfolio is the most likely to provide positive returns because it's a bet on human civilization's more-or-less steady technological development.
The relevance to today's essay is that short-term movements of stock prices tell you nothing, and long-term movements of less than 30% are similarly low on information value.
Like
Reply
2h Edited
Yesterday 8 Prairieland defendants - anti ICE protestors - were sentenced in federal court today to sentences ranging from 30-100 years in prison. Convicted on trumped up charges including terrorism and conspiracy. One defendant Daniel Rolando Sanchez Estrada, was sentenced to 30 years despite not been at the protest. He was indicted because he was moving political literature in his car at the request of his wife. He got 30 years in jail. You can find out ways to help free them at support the Prairieland defendants. It's horrifying and obviously is designed to criminalise dissent.
Like
Reply
Thanks Paul! Sometimes I wonder how my State Sponsored DB Pension is doing, but it is quite opaque to its members on market strategies. And the Plan now requires your picture uploaded to access account info. I will never do that.
Best,
tv
Like
Reply
Many thanks!!
Like
Reply
Nice explanation of what may or may not be happening to the industry. Wonder if the opening of the Straight of Hormuz has anything to do with this (by potentially increasing supply, predicting a drop in price/profits)?
But I’m more interested in the fundamentals: Is any AI company actually making a profit? If not, is there any hint that they are like early Amazon: losing money but building infrastructure and gaining market share so that eventually they will be immensely profitable?
Like
Reply
thank you
Like
Reply
This is really interesting. I have a question about China's use of AI: China is putting AI in everything it can. In part, this is because it is all open source so everyone is using it, but it is also being trained on its access to vast information. However, my observation is that Chinese companies are actually using AI very effectively. In China, AI companies really are turning a profit and, most importantly, AI is being used in many industrial sectors to vastly improve efficiency. Is this because China is simply an industrialized manufacturing economy in which AI has a very practical use? Are economies in the West less industrialized and so where they are using AI is less obviously useful or productive?
Like
Reply
I work with several engineers who are about as skeptical of AI as I am. And we are in this kind of doomloop of none of this stuff gives consistent results...
But because we make products for clients, and the clients want to know we are doing cutting edge stuff, we have to wrap stuff in ai making a worse thing.
Just do a regression. You probably only need a regression.
Like
Reply
“He whose name shall not be mentioned.” 🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣
Like
Reply
Paul Krugman Podcast
Notes on economics and more
Notes on economics and more
Subscribe
Listen on
Substack App
Apple Podcasts
Spotify
YouTube
RSS Feed
Appears in episode
Recent Episodes
Jun 20•Paul Krugman
Power and Geopolitics After Trump
Jun 18•Paul Krugman
Lunch Money with Paul Krugman and Heather Cox Richardson
Jun 18•Paul Krugman and Heather Cox Richardson
Jun 13•Paul Krugman
Jun 8•Paul Krugman
Jun 6•Paul Krugman
Learning from a Mentally Ill President
May 31•Paul Krugman
Ready for more?
Subscribe