There's a quiet flaw in the holy grail right now. Diversification only works while your bets stay uncorrelated. But the thing driving this whole concentration, one technology repricing every sector at once, is also quietly correlating everything that touches it. Chips, power, the grid, even bonds through rates, all turning into the same leveraged bet on the AI capex cycle.
So the correlations climb toward one exactly when you most need them low. Finding 15 good bets was never the hard part. The hard part is that the thing inflating the index is busy making everything the same bet.
This is a good point. Because correlations within equites have risen. But I’d argue that isn’t a flaw in the holy grail. It just makes it more holy grail-ish. (Folk lore vs reality) I’d say it’s just much harder than normal. And as such, more important than ever To try and get it right. I agree 15 good ones seems aspirational. If not nearly impossible.
I think taking rays point here is two fold. 1. is to make sure you are spreadout even within your AI driven bets. So you can own chip stocks, and infrastructure build, energy/nuclear, or direct AI frontier lab, but as you rightly point out. Dont confuse those as separate bets. It’s the same thesis. So it forces your to evaluate your holdings. To me spreading those bets within AI is a way of hedging against being wrong on what still has good valuation. For example. Maybe the chip stocks are over done but there is still a lot of room to run in the energy aspect (personally I think Nuclear) Quantum is correlated too, but as it progresses there is massive upside there. But those are all the same bet. Maybe you can call them 2. we are all a little more leveraged to Ai at the moment, and that’s not necessarily a bad thing if spread well.
The second point of uncorrelation is to find assets with returns that don’t fall together when a thesis breaks. And this, as you highlight, is so really hard right now. Normally you can just run a backtest to see the ebbs and flows. But this one requires you to really think it out. It’s incredibly difficult run a “narrative” in a backtest. So you have to do it by logic, walking forward.
Right now you can look at the sectors that are historically stable but unloved at the moment. Take consumer staples for example. If AI disappoints. Those may not rise, but surely they will fall less. Biotech. Even if AI stalls here. there is still major advancements that we will see out of the Biotech sector. Healthcare is quite defensive and subsector areas of that should stay well bid over the next decade given the demographic landscape.
Bonds or cash. Ray hasn’t been a fan given the where the debt is and he’s not wrong. But you’ve gotta park some money somewhere to rotate around, and at 3.5%, short end duration isnt a bad value over the long end.
Real estate. It’s one thing you can look “at least that’s not ai”
Art. I’m not gonna dignify this one. If you have enough wealth to own 7 figure art pieces. You’re not reading ray dalios Substack.
Gold (which I’m not a fan) has offered excellent low to no correlation with the AI trade. Albeit toxic to your portfolio the last 6 months. People have mistaken it to trade as a risk asset with correlation to equities as of late. But if you look closer It’s traded purely as a high volatility alternative currency to the dollar. Silver and copper have been swooped into the Ai correlation because they are base metals in demand for the build out. And yes the counter argument there is that gold and silver have a near perfect correlation.
My point here is that I think you and Ray are both right. But that it’s more important than ever to reconsider what you have on and be willing to rotate around a bit more.
As we saw with the war in Iran when volatility hits there will be very few assets that do well or don’t go down. (Iran war - energy, dollar/cash, real estate… everything else suffered) and to me that’s the reason to have cash. To have dry powder.
Most investors seem to assume that uncertainty is a problem that must be resolved in order to make a move. It seems you're saying, in investing, we often have to act with substantial uncertainty that cannot be resolved. The same capacity that allows an investor to make a decision without complete certainty is also what allows them to hold a position and not sell at the wrong time when uncertainty inevitably returns.
Diversification then, is a way of investing taking into account the unknown, not just a portfolio balancing technique.
I see this as a form of intellectual honesty that is uncommon in investing, where there is often pressure to have an opinion on everything, but now seems as integral as data analysis itself.
15 uncorrelated bets sounds great in theory. My question is whether truly uncorrelated bets even exist when AI is driving so much of the market. Curious how you're defining "uncorrelated" here.
I too got amused, because equity markets too follow the bond market after some point of time. As bond market can tell you a lot about future of equity markets. But Thanks Ray for sharing his insights in this chaotic scenario. Always lovely reading his articles as it teaches something new!
With you so far. Since we all know different things to different confidence levels, we all have a different ideal diversified portfolio. For example, I would guess that I might be 9.9/10 on environmental/physical constraints and 9.9/10 on some technologies that I understand well. Investing in the intersection of these two seems natural and easy. My best returns are then in a few highly concentrated positions. I’m maybe a 5 or 6/10 on things like geopolitical conflict and monetary policy. It seems much harder to invest in those things that will balance my portfolio because I lack that knowledge or conviction. Is collaboration the answer or is there some other way to invest in what you don’t know effectively?
Ray, I respect the diversification gospel, but I think you're conflating two different risk profiles here. The "bubble" you're describing assumes AI capex is speculative—but $580B from Amazon, Google, and Microsoft isn't venture capital gambling. It's investment-grade infrastructure spending backed by contractually committed enterprise revenue (Oracle's $638B RPO isn't vaporware). The 2000 comparison breaks down because these aren't pre-revenue startups burning cash—they're cash-generative monopolies with 40%+ operating margins building the rails for the next industrial cycle.
Your China disruption thesis is interesting, but you're assuming Fortune 500 CIOs will swap Azure for DeepSeek the moment pricing drops—and that's not how regulated industries work. Compliance frameworks, vendor lock-in, and integration costs create switching friction that commodity pricing alone can't overcome. The real second-order play isn't avoiding concentration—it's positioning in the unsexy infrastructure (power, cooling, fiber) that benefits regardless of which hyperscaler wins. Diversification is prudent, but dismissing the durability of these revenue streams because they're concentrated misses how modern monopolies actually compound.
Thankyou for the valuable commentary. My concern is that debt vs prospective revenue is generally the weak link in these scenarios, and when debt collapses, a heck of a lot of asset classes become correlated quickly. Do you allow for that outcome in your weighting (with cash) or simply allow that wave to wash through?
There's a quiet flaw in the holy grail right now. Diversification only works while your bets stay uncorrelated. But the thing driving this whole concentration, one technology repricing every sector at once, is also quietly correlating everything that touches it. Chips, power, the grid, even bonds through rates, all turning into the same leveraged bet on the AI capex cycle.
So the correlations climb toward one exactly when you most need them low. Finding 15 good bets was never the hard part. The hard part is that the thing inflating the index is busy making everything the same bet.
This is a good point. Because correlations within equites have risen. But I’d argue that isn’t a flaw in the holy grail. It just makes it more holy grail-ish. (Folk lore vs reality) I’d say it’s just much harder than normal. And as such, more important than ever To try and get it right. I agree 15 good ones seems aspirational. If not nearly impossible.
I think taking rays point here is two fold. 1. is to make sure you are spreadout even within your AI driven bets. So you can own chip stocks, and infrastructure build, energy/nuclear, or direct AI frontier lab, but as you rightly point out. Dont confuse those as separate bets. It’s the same thesis. So it forces your to evaluate your holdings. To me spreading those bets within AI is a way of hedging against being wrong on what still has good valuation. For example. Maybe the chip stocks are over done but there is still a lot of room to run in the energy aspect (personally I think Nuclear) Quantum is correlated too, but as it progresses there is massive upside there. But those are all the same bet. Maybe you can call them 2. we are all a little more leveraged to Ai at the moment, and that’s not necessarily a bad thing if spread well.
The second point of uncorrelation is to find assets with returns that don’t fall together when a thesis breaks. And this, as you highlight, is so really hard right now. Normally you can just run a backtest to see the ebbs and flows. But this one requires you to really think it out. It’s incredibly difficult run a “narrative” in a backtest. So you have to do it by logic, walking forward.
Right now you can look at the sectors that are historically stable but unloved at the moment. Take consumer staples for example. If AI disappoints. Those may not rise, but surely they will fall less. Biotech. Even if AI stalls here. there is still major advancements that we will see out of the Biotech sector. Healthcare is quite defensive and subsector areas of that should stay well bid over the next decade given the demographic landscape.
Bonds or cash. Ray hasn’t been a fan given the where the debt is and he’s not wrong. But you’ve gotta park some money somewhere to rotate around, and at 3.5%, short end duration isnt a bad value over the long end.
Real estate. It’s one thing you can look “at least that’s not ai”
Art. I’m not gonna dignify this one. If you have enough wealth to own 7 figure art pieces. You’re not reading ray dalios Substack.
Gold (which I’m not a fan) has offered excellent low to no correlation with the AI trade. Albeit toxic to your portfolio the last 6 months. People have mistaken it to trade as a risk asset with correlation to equities as of late. But if you look closer It’s traded purely as a high volatility alternative currency to the dollar. Silver and copper have been swooped into the Ai correlation because they are base metals in demand for the build out. And yes the counter argument there is that gold and silver have a near perfect correlation.
My point here is that I think you and Ray are both right. But that it’s more important than ever to reconsider what you have on and be willing to rotate around a bit more.
As we saw with the war in Iran when volatility hits there will be very few assets that do well or don’t go down. (Iran war - energy, dollar/cash, real estate… everything else suffered) and to me that’s the reason to have cash. To have dry powder.
Good observation, Scenarios. Take farming, for example. It might seem that the farm is a long ways from AI/tech excitement.
But even the farmer needs to look into how AI/Tech is going to impact his farm.
Another thing: AI is having a massive effect on the economy. Talk about everything being connected!
Most investors seem to assume that uncertainty is a problem that must be resolved in order to make a move. It seems you're saying, in investing, we often have to act with substantial uncertainty that cannot be resolved. The same capacity that allows an investor to make a decision without complete certainty is also what allows them to hold a position and not sell at the wrong time when uncertainty inevitably returns.
Diversification then, is a way of investing taking into account the unknown, not just a portfolio balancing technique.
I see this as a form of intellectual honesty that is uncommon in investing, where there is often pressure to have an opinion on everything, but now seems as integral as data analysis itself.
It’s interesting how many investing mistakes start with a strong opinion and end with a lack of diversification.
15 uncorrelated bets sounds great in theory. My question is whether truly uncorrelated bets even exist when AI is driving so much of the market. Curious how you're defining "uncorrelated" here.
I too got amused, because equity markets too follow the bond market after some point of time. As bond market can tell you a lot about future of equity markets. But Thanks Ray for sharing his insights in this chaotic scenario. Always lovely reading his articles as it teaches something new!
With you so far. Since we all know different things to different confidence levels, we all have a different ideal diversified portfolio. For example, I would guess that I might be 9.9/10 on environmental/physical constraints and 9.9/10 on some technologies that I understand well. Investing in the intersection of these two seems natural and easy. My best returns are then in a few highly concentrated positions. I’m maybe a 5 or 6/10 on things like geopolitical conflict and monetary policy. It seems much harder to invest in those things that will balance my portfolio because I lack that knowledge or conviction. Is collaboration the answer or is there some other way to invest in what you don’t know effectively?
Ray Dalio delivering as always.
As a retiree who has to live off of our investment income I am extremely grateful for Ray sharing his experience and thoughts.
Thank you for providing your views on how to navigate this tumultuous time. We are finding it most helpful! Looking forward to hearing more.
Thank you for sharing this with us!
Ray, I respect the diversification gospel, but I think you're conflating two different risk profiles here. The "bubble" you're describing assumes AI capex is speculative—but $580B from Amazon, Google, and Microsoft isn't venture capital gambling. It's investment-grade infrastructure spending backed by contractually committed enterprise revenue (Oracle's $638B RPO isn't vaporware). The 2000 comparison breaks down because these aren't pre-revenue startups burning cash—they're cash-generative monopolies with 40%+ operating margins building the rails for the next industrial cycle.
Your China disruption thesis is interesting, but you're assuming Fortune 500 CIOs will swap Azure for DeepSeek the moment pricing drops—and that's not how regulated industries work. Compliance frameworks, vendor lock-in, and integration costs create switching friction that commodity pricing alone can't overcome. The real second-order play isn't avoiding concentration—it's positioning in the unsexy infrastructure (power, cooling, fiber) that benefits regardless of which hyperscaler wins. Diversification is prudent, but dismissing the durability of these revenue streams because they're concentrated misses how modern monopolies actually compound.
Everyone is so excited with greed. Personally, I see this AI/TECH mania to be another dot.com fiasco.
When the big crash comes, no one will care how much energy that chip draws. Trust me. Greed, greed, greed.
DIVERSIFY.
Thank you for this wise perspective.
Thankyou for the valuable commentary. My concern is that debt vs prospective revenue is generally the weak link in these scenarios, and when debt collapses, a heck of a lot of asset classes become correlated quickly. Do you allow for that outcome in your weighting (with cash) or simply allow that wave to wash through?
useful information
Great advice, thank you Ray.
Looking forward to your next communication.