A stock market veteran’s view on the AI bubble

AI adoption will cause job displacement and potential social unrest as productivity gains exceed the economy's ability to absorb and redeploy workers into new jobs, creating 'a lot of social unrest and a lot of pushback and a lot of strange things that we haven't any will experience with.'

“The increase in productivity throws a lot of people out of work. There may be redeployment opportunities for them that are respectable and decent paying, but there may not be. And uh if they if the productivity is faster than we can absorb the new jobs, you will get a lot of social unrest and a lot of pushback and a lot of strange things that we haven't any will experience with.”

Atmospheric CO2 has climbed remorselessly from 280 parts per million (pre-industrial) to 430 parts per million (current), and will likely peak at 550 ppm even with serious effort to reduce emissions, requiring eventual reduction back to 300 ppm to prevent long-term ocean level rise.

“There is data on CO2 particles in the atmosphere. Uh and uh they've climbed remorselessly from 280 parts per million to 430. And uh we think if we really try a bit harder and the technology stays good and improves, uh we'll be lucky to peak out at 550 parts per million. And uh incidentally, we have to get that back to 300 or the ocean levels continue to rise slowly all the time into the dim dim distant future.”

Electric vehicle sales in Europe are up 40% year-over-year, indicating market acceptance of EV technology, whereas the US is displaying an almost unique ability to miss obvious trends and be self-destructive by not pursuing EV adoption.

“sales of EVs in in Europe are up 40% year over year. Now, they're not up that much in the US. US is displaying a almost unique ability to miss the obvious and be self-destructive.”

AI investment capital expenditure has escalated from 2023 through 2026, remaining strong enough to shock the system and make economic performance appear unexpectedly strong, while also sustaining unusually high corporate profit margins for the Magnificent 7 companies.

“if you look at the CAPEX you'll see that 2023 is a compared to 2024 25 26. It's just escalated. So it's stayed strong enough to kind of shock the system and make it feel like it was unexpectedly strong. And with the corporate profits from the mag seven unexpectedly strong again.”

Voter behavior in recent elections shows that voters across countries punish incumbents regardless of party affiliation or whether they are right-wing or left-wing, and the 2020 US shift from Biden to Trump was actually smaller than the prior seven European elections.

“they vote against the party in power. Doesn't matter if they're right-wing like the conservatives, kick the rascals out. Left-wing in in in France, kick the rascals out. And ironic ironically the the shift from Biden to Trump, much analyzed by everybody, was less than the seven prior elections in Europe.”

Atmospheric CO2 has risen from 280 parts per million to 430 parts per million, and if current trends continue with aggressive technology improvements, it will likely peak around 550 ppm, requiring subsequent reduction to 300 ppm to prevent continued ocean level rise over hundreds of years

“There There is data on CO2 particles in the atmosphere. Uh and uh they've climbed remorselessly from 280 parts per million to 430. And uh we think if we really try a bit harder and the technology stays good and improves, uh we'll be lucky to peak out at 550 parts per million. And uh incidentally, we have to get that back to 300 or the ocean levels continue to rise slowly all the time into the dim dim distant future. Uh hundreds of hundreds of years”

Even though the political consensus around climate has shifted away from Grantham's position in recent years, companies globally have recognized the economic and security advantages of renewables over fossil fuels and are planning accordingly, as evidenced by 40% year-over-year EV sales growth in Europe

“Anybody who hasn't taken on board the vulnerability of oil and gas compared to the security of of solar wind and storage is brain dead. They all have and they're all planning accordingly. And sales of EVs in in Europe are up 40% year over year”

Wind, solar, and energy storage have made far more progress than expected when Grantham first wrote about these technologies 15 years ago, with storage costs falling to less than one-tenth of previous levels and continuing to decline

“Wind, solar, storage have all made much more progress, by the way, than we thought. When I first started to write some papers 15 years ago, we we never expected storage in particular to be Oh, if perhaps three times the current price would not have surprised me. It It's come down to less than a tenth since I wrote my first paper. Less than And it's still falling like a rock”

The AI investment boom has been so powerful and sustained that it has maintained corporate profit growth and prevented the market from sliding into recession that would have occurred without ChatGPT and AI investment

“Chat set off an amazing rally in the in the mag 7, plus or minus a few other companies. And uh secondly, the AI investment became so strong that it dragged an economy back from sliding into a recession. I think in a non-chat world, what would what would have happened is the market would have continued to go down. And in in 2022, we would have slipped into a mild recession”

Current AI investment is exhibiting the same pattern as railroads and internet bubbles: too much money being pushed into a transformative technology that is obviously powerful and life-changing

“and that is exactly as you can see dear listener. You can see that happening in spades almost today”

Heavy downpours caused by climate change have cost the global economy about half a percent of GDP in recent years, whereas 20 years ago such weather events did not cause serious economic damage, and prevention of natural disasters from fires, floods, droughts, and ocean flooding costs substantial money

“we can see in the last 2 or 3 years. Last couple of years. We reckon it's cost the global GDP about half a percent. We used to worry about it 20 years ago, but it didn't actually cost any serious money. And now it does. Preventing preventing all these natural disasters, fires, floods, and droughts. It's uh an occasional ocean floods. Cost a lot of money”

CO2 and methane have been known for over 100 years to create a heat-trapping blanket effect in the atmosphere, and the precise mechanisms of how temperature increases lead to more water vapor (about 15% more than used to exist) and subsequently more heavy downpours are well understood

“they've known the effect for over 100 years that CO2 has as uh creating a blanket to trap the heat, the same with methane and nitrous oxide. They know precisely uh uh reasonably precisely uh what is going on...The air carries more water vapor. That's just physics. We carry about 15% more water vapor than we used. What are the consequences of that? Well, one of them is more heavy downpours”

Technology is 'perhaps the last best hope' for solving climate change because humans do not appear sensible enough as a species to voluntarily sacrifice for the welfare of future generations.

“Technology it is perhaps the last best hope because it doesn't appear that we're sensible enough as a species to want to actually sacrifice to do the right thing for our grandchildren. Screw our grandchildren, you know.”

Oil supplies have been maintained through reserve stocks built into the system and logistical lags (long shipping times), but these buffers are being exhausted, meaning geopolitical oil risks are about to become much more acute.

“we've been floating along uh on on the reserves that were built into the system and the lags. It took weeks and weeks for the last oil tankers to just finish their journey and unload in Japan... now we're coming to the end of that.”

Grantham considers himself a 'fanatic on climate change' because he views it as life-threatening for the human species, potentially putting civilization in a 'very bad place' within 40-50 years without serious action.

“I suppose I have to think of myself as a fanatic on on climate change because I I consider it uh life-threatening for for our species. I think in as little as 40, 50 years, we could be in a very bad place.”

Warmer air carries about 15% more water vapor than previous conditions, which causes more heavy downpours visible in recent years, costing the global economy about 0.5% of GDP annually in disaster costs and recovery expenses.

“The air carries more water vapor. That's just physics. We carry about 15% more water vapor than we used. What are the consequences of that? Well, one of them is more heavy downpours, which we can see in the last 2 or 3 years... We reckon it's cost the global GDP about half a percent.”

The Federal Reserve and regulatory authorities took a passive stance toward the Magnificent Seven's monopolistic consolidation, not intervening like Teddy Roosevelt's antitrust actions against Standard Oil, which allowed these firms to become 'grand global monopolies' rather than local ones.

“They were enormously helped by the way in the Justice Department's attitude [11:34] and the administration's attitude to monopolies. They They weren't interested. So, they were not stepping in like Teddy Roosevelt to break up uh the great Standard Oil. So, these were not just local monopolies. They were often grand global monopolies.”

Technology is perhaps the last best hope for addressing climate change because the human species does not appear sensible enough to make voluntary sacrifices for the sake of future generations

“Technology it is perhaps the last best hope because it doesn't appear that we're sensible enough as a species to want to actually sacrifice to do the right thing for our grandchildren. Screw our grandchildren, you know”

Grantham considers himself 'a fanatic on climate change' because he considers it life-threatening for humanity, with potential for the species to be in a 'very bad place' in as little as 40-50 years without serious action

“I have to think of myself as a fanatic on on climate change because I I consider it uh life-threatening for our species. I think in as little as 40, 50 years, we could be in a very bad place”

Wind, solar, and energy storage technologies have made far more progress than expected 15 years ago; storage costs have fallen to less than one-tenth of what Grantham projected, and are continuing to decline, while fossil fuel costs are rising due to political exposure.

“Wind, solar, storage have all made much more progress, by the way, than we thought. When I first started to write some papers 15 years ago, we we never expected storage in particular to be Oh, if perhaps three times the current price would not have surprised me. It It's come down to less than a tenth since I wrote my first paper. Less than And it's still falling like a rock.”

The housing bubble was well-behaved in its pattern (rising for 3 years, then declining for 3 years) and reached 3-sigma (a one-in-100-years event), whereas the US housing market had never bubbled at that extreme before and required both Greenspan and Bernanke to create it.

“We had, uh, a a really splendid housing bubble, which went way over two sigma to three sigma, one every 100 years or longer. Uh, the US housing market basically had never bubbled like that. It it took Greenspan and Bernanke working hard to create that situation. And, uh, very well-behaved, incidentally. It went up for 3 years and then down for 3 years. It looked looked perfect.”

Federal Reserve policy since Alan Greenspan's tenure has made bubble occurrences more frequent than the historical 36-year rate, suggesting policy choices have increased market volatility and bubble frequency.

“There's some indication that more recently, after Alan Greenspan and his troops, they've, uh, initiated a style of Federal Reserve policy that would make them a little more often.”

CO2, methane, and nitrous oxide have all been known for over 100 years to trap heat in the atmosphere similar to a blanket, creating a greenhouse effect; the precise mechanisms and consequences can be calculated from this physical understanding.

“They know the effect. And they've known the effect for over 100 years that CO2 has as uh creating a blanket to trap the heat, the same with methane and nitrous oxide. They know precisely uh uh reasonably precisely uh what is going on.”

When dominant firms compete directly in the same market, they engage in aggressive, destructive capitalism akin to late 19th century practices where firms went 'for the throat' rather than maintaining oligopolistic equilibrium, leading to margin compression

“these seven uh firms are not your average firms. They are ultimately aggressive. We have actually learned to develop a very sharp-edged form of capitalism, not unlike the late uh 19th century, where uh Standard Oil and so on went for the throat. And and and bully the railroads into increasing increasing the freight rates on their competitors and so on. I mean, really dirty pool. And and we've gotten pretty sharp sharp-edged. And these guys are the winners. They They go for the throat”

The stock market is currently in a bubble and has been for 4-5 years, with clear indicators that it checked every marker of a 'super bubble' by the end of 2021.

“Starting at the end, yes, we are in one and have been for some time... We entered bubble territory quite a long time ago, um, this time, 4 or 5 years ago, uh, and it it bubbled nicely, uh, till the end of 20 21. The end of 2021 it it checked off every indicator of what you might call super bubble coming to the end.”

We are currently in a stock market bubble and have been for some time, entering bubble territory 4-5 years ago

“Starting at the end, yes, we are in one and have been for some time”

The United States displays 'an almost unique ability to miss the obvious and be self-destructive' in terms of adopting EVs and transitioning from fossil fuels compared to Europe

“US is displaying a almost unique ability to miss the obvious and be self-destructive”

The best historical comparison for understanding the current AI bubble is the tech/internet bubble of 1999-2002, which felt very similar up until the end of 2021 and early 2022, before ChatGPT introduced an unprecedented new variable.

“I think the most useful straightforward comparison is with between the tech bubble and uh the first half of this bubble we're in now. Up to the end of 2021 and even into the break of 2022, it felt very much like the uh internet bubble uh forming and breaking.”

Trump should be considered for major environmental awards because no one has done more good for the long-term health of the planet than him, as major corporations have recognized the vulnerability of oil/gas dependence and the security benefits of solar, wind, and storage.

“Trump should be considered for the great environmental prize that they give every year. Because no one has ever done this much a good for a long-term green planet. Anybody who hasn't taken on board the vulnerability of oil and gas compared to the security of of solar wind and storage is brain dead.”

A bubble can be statistically defined using two standard deviations from trend data; events exceeding two standard deviations occur approximately once every 36 years in the stock market (compared to the mathematical expectation of once every 44 years for random data), making this definition both statistically rigorous and practically useful.

“And uh being slightly nerdy, we looked around for a sta- a statistical way of defining it. And very conveniently, there was this concept of two sigma, two standard deviations. And uh a standard deviation just tells you how rare an outlier is in a in a field of data. And it turns out that it's uh every 44 years on an annual basis. A a two sigma event uh upside and downside is every 44 years uh you get an outlier. And in the stock market in real life, uh you get one every 36 years, but I was surprised how close that was to 44.”

The Magnificent 7 tech companies previously represented strong monopolies in seven different sectors (Microsoft, Apple, Google, etc.), each dominating their own field, but now they are all preparing to directly compete against each other in the same AI field, fundamentally changing their competitive dynamics.

“the mag seven used to represent really strong monopolies in seven different areas. They each control their field. Microsoft, Apple, Google, and so on. And and now they're all bracing to be the top dog in the same field. This is utterly different.”

The critical question about the current market is how much of the profit and PE gains is permanent versus temporary, and how long the Magnificent 7 can maintain their profit margins as competitive dynamics shift from regional monopolies to head-to-head competition.

“now the question is how much of that is permanent, how much of the PE move is justified. But I think most importantly how long can they maintain their profit margins? I have a very strong feeling on that.”

The vulnerability of tech companies' valuations to technological breakthroughs is high, as the assumption that they will maintain dominance in AI technology rests on the belief that no other breakthrough technology will disrupt their position.

“By the way, the vulnerability, the vulnerability of these people to a technological uh breakthrough of some other kind. You know, a new a new type of of chip, a new type of technology. Um there doesn't seem to be uh to me a death grip on on this technology the way people are assuming.”

The Magnificent Seven have developed and refined a 'sharp-edged form of capitalism' similar to the ruthless late 19th-century practices of Standard Oil (bullying railroads into raising freight rates on competitors), where the winners use aggressive tactics to maintain dominance.

“We have actually learned to develop a very sharp-edged form of capitalism, not unlike the late uh 19th century, where uh Standard Oil and so on went for the throat. And and and bully the railroads into [12:34] increasing increasing the freight rates on their competitors and so on. I mean, really dirty pool. And and we've gotten pretty sharp sharp-edged. And these guys are the winners. They They go for the throat.”

The cloud computing market had only three major competitors (unlike the seven in AI) and they maintained a 'very well-behaved oligopoly' by intuitively coordinating to take it easy on each other, but the AI competitors are openly 'screaming at each other,' indicating this level of coordination is unlikely to occur.

“they had a little go around on the cloud, but they're only three and they handled it very carefully and they either intuited that the others were you know, take it easy on me, I'll take it easy on you and they they shared the market, a [14:40] very well-behaved oligopoly. But this is not You didn't hear the cloud people screaming at each other, but you hear the AI people doing that.”

A bubble can be defined statistically as a two-sigma (two standard deviation) event, which occurs approximately once every 36-44 years in stock market data, representing an outlier far removed from the long-term trend

“we looked around for a sta- a statistical way of defining it. And very conveniently, there was this concept of two sigma, two standard deviations. And uh a standard deviation just tells you how rare an outlier is in a in a field of data. And it turns out that it's uh every 44 years on an annual basis. A a two sigma event uh upside and downside is every 44 years uh you get an outlier. And in the stock market in real life, uh you get one every 36 years”

When Grantham paid to acquire computers for his investment firm, it provided a competitive advantage for a couple of years, but once everyone had computers it became just 'a cost of doing business' producing only modest returns on capital with no upward pressure on profit margins

“When we finished paying for our computer, which we choked on, we had we it helped us have a an advantage in calculating everything for a couple of years, particularly over uh smaller firms like us. And uh and then everyone had a computer. It It was just a cost of doing business. All you should expect to make in that situation is a modest return on the capital involved”

Looking out 10 years from now and turning back, everyone will have AI, everyone will be buying whatever AI services they can afford, competitive advantage will be competed away, and AI will become something all firms have to do just to maintain parity

“If you look out 10 years and then turn around and look back, everyone will have AI. Everyone will be buying the service they can afford. And the com- competitive advantage will have been competed mostly away. It will be something you have to do. If you don't have an expert uh uh uh on your staff, you better get one. And uh everyone will have their expert”

During the tech bubble, Grantham's firm was making 6-7% annual returns while the market surged, which was enough to lose half its clients despite being right, because short-term underperformance is intolerable to clients regardless of long-term correctness

“All we did was under We didn't even lose money, of course. We were making 6 or 7 or 8% a year, just much less than than uh the high shots and much less than the market. 6% a year uh less than the market. And that was enough in two and a quarter years to lose half our clients. None of whom came back”

Unlike the cloud computing market where three major players (AWS, Azure, Google Cloud) engaged in careful, cooperative behavior and informally shared the market, the AI competition is characterized by public hostility and aggressive chest-beating, indicating a race dynamic unlike any in recent tech history.

“they handled it very carefully and they either intuited that the others were you know, take it easy on me, I'll take it easy on you and they they shared the market, a very well-behaved oligopoly. But this is not You didn't hear the cloud people screaming at each other, but you hear the AI people doing that.”

Technology companies may be vulnerable to disruption from unexpected breakthroughs in chip design or entirely new types of technology, meaning they do not have a 'death grip' on their dominant positions

“the vulnerability, the vulnerability of these people to a technological uh breakthrough of some other kind. You know, a new a new type of of chip, a new type of technology. Um there doesn't seem to be uh to me a death grip on on this technology the way people are assuming”

Grantham does not see evidence of the Magnificent Seven's aggressive AI competition being reflected in market commentary or analysis, despite it being a simple and straightforward logical outcome of their competitive situation.

“Even though that argument seems so simple and straightforward, I don't see it anywhere. People think all seven of them are going to make obscene profits. Well, that would be new. When does that ever happen in any market ever?”

The shift from Biden to Trump was less significant than seven prior elections in Europe, suggesting anti-incumbent voting is primarily driven by economic discontent rather than unique American political factors

“the shift from Biden to Trump, much analyzed by everybody, was less than the seven prior elections in Europe”

The recent oil crisis showed that the world is ending reliance on reserve buffers built into the system, with oil tankers taking weeks to unload and supply chain lags that previously provided time to adapt now coming to an end

“the oil crisis. People don't realize we've been floating along uh on on the reserves that were built into the system and the lags. It took weeks and weeks for the last oil tankers to just finish their journey and unload in Japan, you know. All of these things bought us bought us time and our reserves were pretty good depending on which country. And and now we're coming to the end of that”

Grantham's firm lost half its clients (and none came back) during the tech bubble of 1997-1999 by being underweight the market while it was still rising, losing 6-7% annually versus market gains, which was enough to lose clients despite ultimately being right.

“We'd done pretty well through '97. All we did was under We didn't even lose money, of course. We were making 6 or 7 or 8% a year, just much less than than uh the high shots and much less than the market. 6% a year uh less than the market. And that was enough in two and a quarter years to lose half our clients. None of whom came back.”

Grantham's firm managed to weather the career risk by being a private firm that could afford to lose business; they dropped from $30 billion to $20 billion in assets while competitors went from $30 billion to $50-60 billion, then recovered to $165 billion in the following 4 years after the market crashed.

“we were a private firm, and if we chose to lose half our business quickly, that was that which we did, that was our business... we went from uh 30 billion to uh 20. When everyone else was going from 30 to 50 or 60... then in the break, when they went back from 50 to 35 or 30, we went from 20 to 22... in the following 4 years, we went from 22 to 165.”

The tech bubble of 1999-2000 provides the most useful comparison to the first half of the current bubble (through 2021), showing similar patterns and validating Grantham's 2022 prediction that the market would break

“I think the most useful straightforward comparison is with between the tech bubble and uh the first half of this bubble we're in now. Up to the end of 2021 and even into the break of 2022, it felt very much like the uh internet bubble uh forming and breaking. And a lot of the nuances that we could discuss were were similar and uh made me very confident that in late 2021 uh the market would break, and it did”

Grantham was surprised that despite believing markets are 'totally inefficient', they almost obey statistical rules that would apply to a random series, as if markets were efficient, with outlier events (two standard deviations) occurring roughly every 36 years as predicted by theory.

“I'm a great believer that the market is totally inefficient. And the fact that it would almost obey a statistical rule that applies to a random series, uh, as if the market were efficient, uh, surprised me. But, uh, 36 years seems like for most people a pretty workable definition of an outlier, an extreme event, doesn't it?”

Grantham has donated over 90% of his wealth to his foundation for environmental protection, and the foundation operates as 'shock troops' making aggressive new investments, pioneering research, and helping new enterprises get off the ground.

“the great majority, well over 90% of everything I have has gone into into our foundation for the protection of the environment. And and and we try and operate there as kind of shock troops. We We want to be making aggressive new investments, helping new groups get going, whether they're profit or non-profit, uh helping to pioneer the research, and um and helping new enterprises get off the ground”

Each of the Magnificent Seven is willing to spend $200 billion or more on capital expenditure in a single year to achieve AI dominance, viewing it as a winner-take-all competition where being first to the next breakthrough is worth any investment.

“And here they are looking at each other now and saying, 'It's going to be vastly important to be the biggest and the best at AI. To be the first one to get the next generation breakthrough in in uh in AI. And it doesn't matter if it takes me 200 billion in cap eight this year, I'm going to do it.'”

The exact same dynamic of excessive capital deployment that occurred in railroads and internet is happening 'in spades almost today' in AI, with investors falling over themselves to fund AI ventures regardless of valuation, such as the SpaceX IPO.

“And that is exactly as you can see dear listener. You can see that happening in spades almost today... I was talking to someone the other day about some bankers about the SpaceX IPO and and the the sort of the the extent to which people are falling over themselves to put money into that IPO regardless of the valuation”

All seven Mag 7 firms view AI as a field where being the biggest and best at AI is 'vastly important' and winning the next AI generation breakthrough is critical enough to justify spending $200 billion or more in capex this year regardless of return calculations.

“they've learned to be the winner who takes all. And uh and here they are looking at each other now and saying, 'It's going to be vastly important to be the biggest and the best at AI. To be the first one to get the next generation breakthrough in in uh in AI. And it doesn't matter if it takes me 200 billion in cap eight this year, I'm going to do it.'”

On January 15, 2022, Grantham wrote a quarterly letter stating 'Let the wild rumpus begin,' predicting the market would decline if it were listening to history, as all indicators showed the bubble was peaking.

“I wrote a quarterly letter on January the 15th, 2022, saying, 'Let the wild rumpus begin.' In other words, very rare for me, the timing is now. The market should go down if it's listening to history.”

Being wrong for 2.25 years during a bubble was acceptable to Grantham because his firm was private and willing to accept the risk; most firms cannot and will not accept such pain, choosing to 'want the quiet life' instead.

“Only only a an independent firm with a willingness to be uh exposed to the possibility of a semi-disastrous drop in business... We were only wrong for 2 and 1/4 years... who who wants to play that game? Um everyone wants the quiet life. I I strongly recommend you don't do it.”

Both financial market analysis and climate analysis are fundamentally data-driven, examining data and making logical conclusions, allowing similar analytical approaches despite different subject matter

“I would say the similarities are that they're both data-driven. You You look at the data and you try and make the logical conclusion”