1. The 10-year tsy yield just dipped below Fed Funds. Yield curve inversion, again. I'm still short TLT as the Fed is unlikely to cut anytime soon, and negative term premium is an unnatural state.
2. Data centers can (and eventually will) have built-in natural gas turbines and generate their own electricity rather than being grid-tied. One can also reduce the electric needs of data centers by building them in the arctic, since a significant % of electricity is used for cooling.
Hi Russell, a big fan of your Substack here (I’m a South Korean econ student, studying in the Netherlands!). After listening to your podcast, I’ve been trying to make sense of current market dynamics regarding the overall market sentiment and the concentrated interest in AI stocks (as it pertains to the US); why can these stocks get such high valuations, even after having been priced in the conventional reasoning that goes about fundamentals, growth, and investor psychology?
Talking without any industry experience, my theory goes:
Background) Global Liquidity levels are still rising, the real economy is showing signs of weakness, and market breadth is getting thinner.
1) In a late-stage bull market, liquidity in the market is still on an increasing trajectory, but certain stocks begin to fall out as earnings start to disappoint.
2) Investors / Traders still want to stay in the market. So they “switch lanes” and latch onto the momentum of the strongest stocks (think NVDA, PLTR).
3) As a result, certain “winning” stocks get a stronger tailwind. The price chart of these stocks begin to look parabolic while general market breadth slowly fades. Increasingly, as negative market sentiment seeps in, there are fewer and fewer winners, but the remaining winners win stronger and stronger. Technically, indices can still be hitting new highs, since investors just “switched lanes,” and net flows into the market are still positive.
4) At last, when negative market sentiment dominates, traders pull their investments and seek cash. Consequently, the last of the winning stocks also become the largest losers, because most stocks’ recent runup were led by momentum investors. The fall of these stocks mark the end of a bull run.
According to my theory, we would currently be in step 3) of the bull market. I think this could make some sense especially if we consider how the economic cycle tends to move -> When the real economy begins to deteriorate, certain economy-sensitive sectors (e.g. housing and durable goods manufacturing) tend to show the earliest signs of decline (e.g. falling new investments), whereas service sector stocks — which also comprise of bubbly tech stocks — are often the last to be hit by bad economic conditions. So under this lens, the most bubbly stocks would also be the last to be exposed under the hood (e.g. due to bad earnings).
I’m wondering if any of this makes sense to you, and I’d also like to ask if this has been true in your experience, back in other lead-ups to stock market bubbles and crashes of the past. Thanks!
P.S - I’d just like to show my appreciation for your articles and how much it has helped me in becoming a better reader of the markets, though I'm still very new to this. It would be a dream-come-true to one day work and learn under your belt!
I think the biggest driver is that we no longer really see price wars... Apple could cut product prices and still be profitable. Same with cloud computing - but big tech has learnt not to compete this way anymore. Contrast to China is quite interesting. Usually countries with little compeition tend to see their currencies weaken - which may be happening now.
Hi Russell, Thanks for the article. However, In this ZeroHedge article, https://www.zerohedge.com/markets/morgan-stanleys-shocking-math-paying-ai-capex-will-need-over-1-trillion-new-debt-2028 , Morgan Stanley thinks there will be $2.9 Trillion in Global Cap Ex between 2025 and 2028, and that about $1.15 Trillion of that will come from the debt markets. $800 Billion will come from Private Credit, $200 Billion in Corporate Debt, and $150 Billion from ABS/CMBS. (MS does estimate $1.4 Trillion from Hyperscaler cash flow funding and $350 Billion from PE/VC/ Sovereign.)
Does this not contradict your thesis that there cannot be a bubble because the Cap Ex build-out will not touch the credit markets? What about the leverage in Private Credit?
If you look across the private space, they have trillions in "dry powder" - and if the government is offering a backstop to the data centre build out, then I can see the money being there.
I think the best example was in World War II - when the government funded all military equipment innovation until the war was won. Then once it was done, a whole bunch of planes were scrapped without ever been used.
I'm not thinking of the price of electricity, but that the cost of running trained AI models will converge with the cost of electricity. Cheap to free, open source, trained models will deliver consumer surplus. At some point, if AI finds all the low hanging fruit and has built out electricity for intensive training, then eventually there should be a glut down the road. Maybe coal is a winner right now.
so a bit like the overbuild in broadband caused prices to collapse in the 2000s. Definitely possible - but generally speaking the cost to me of using ChapGPT and VEO is not that high. I feel both are loss making business at the moment
Brad Safalow pointed out Rocket Mortgages saving 20k person hours using AI. From the CC, "AI-Driven Productivity: Automated refinancing client follow-ups rose by nearly 20%, with AgenTeq AI processing over 80% of earnest money deposit validations, and saving an estimated 20,000 hours annually."
If there is more electricity consumption, you would expect more electrical grid instability. This should drive backup generator growth such as Generac (GNRC). I went long 2 months ago
Interesting. MS were pointing out that some bitcoin miner have value due to long term contracts with access to cheap electricity. They seemed to imply that was the logic behind CoreWeave buying Corse Scientific.
1. The 10-year tsy yield just dipped below Fed Funds. Yield curve inversion, again. I'm still short TLT as the Fed is unlikely to cut anytime soon, and negative term premium is an unnatural state.
2. Data centers can (and eventually will) have built-in natural gas turbines and generate their own electricity rather than being grid-tied. One can also reduce the electric needs of data centers by building them in the arctic, since a significant % of electricity is used for cooling.
Hi Russell, a big fan of your Substack here (I’m a South Korean econ student, studying in the Netherlands!). After listening to your podcast, I’ve been trying to make sense of current market dynamics regarding the overall market sentiment and the concentrated interest in AI stocks (as it pertains to the US); why can these stocks get such high valuations, even after having been priced in the conventional reasoning that goes about fundamentals, growth, and investor psychology?
Talking without any industry experience, my theory goes:
Background) Global Liquidity levels are still rising, the real economy is showing signs of weakness, and market breadth is getting thinner.
1) In a late-stage bull market, liquidity in the market is still on an increasing trajectory, but certain stocks begin to fall out as earnings start to disappoint.
2) Investors / Traders still want to stay in the market. So they “switch lanes” and latch onto the momentum of the strongest stocks (think NVDA, PLTR).
3) As a result, certain “winning” stocks get a stronger tailwind. The price chart of these stocks begin to look parabolic while general market breadth slowly fades. Increasingly, as negative market sentiment seeps in, there are fewer and fewer winners, but the remaining winners win stronger and stronger. Technically, indices can still be hitting new highs, since investors just “switched lanes,” and net flows into the market are still positive.
4) At last, when negative market sentiment dominates, traders pull their investments and seek cash. Consequently, the last of the winning stocks also become the largest losers, because most stocks’ recent runup were led by momentum investors. The fall of these stocks mark the end of a bull run.
According to my theory, we would currently be in step 3) of the bull market. I think this could make some sense especially if we consider how the economic cycle tends to move -> When the real economy begins to deteriorate, certain economy-sensitive sectors (e.g. housing and durable goods manufacturing) tend to show the earliest signs of decline (e.g. falling new investments), whereas service sector stocks — which also comprise of bubbly tech stocks — are often the last to be hit by bad economic conditions. So under this lens, the most bubbly stocks would also be the last to be exposed under the hood (e.g. due to bad earnings).
I’m wondering if any of this makes sense to you, and I’d also like to ask if this has been true in your experience, back in other lead-ups to stock market bubbles and crashes of the past. Thanks!
P.S - I’d just like to show my appreciation for your articles and how much it has helped me in becoming a better reader of the markets, though I'm still very new to this. It would be a dream-come-true to one day work and learn under your belt!
I think the biggest driver is that we no longer really see price wars... Apple could cut product prices and still be profitable. Same with cloud computing - but big tech has learnt not to compete this way anymore. Contrast to China is quite interesting. Usually countries with little compeition tend to see their currencies weaken - which may be happening now.
Sometimes things just take time.
Hi Russell, Thanks for the article. However, In this ZeroHedge article, https://www.zerohedge.com/markets/morgan-stanleys-shocking-math-paying-ai-capex-will-need-over-1-trillion-new-debt-2028 , Morgan Stanley thinks there will be $2.9 Trillion in Global Cap Ex between 2025 and 2028, and that about $1.15 Trillion of that will come from the debt markets. $800 Billion will come from Private Credit, $200 Billion in Corporate Debt, and $150 Billion from ABS/CMBS. (MS does estimate $1.4 Trillion from Hyperscaler cash flow funding and $350 Billion from PE/VC/ Sovereign.)
Does this not contradict your thesis that there cannot be a bubble because the Cap Ex build-out will not touch the credit markets? What about the leverage in Private Credit?
If you look across the private space, they have trillions in "dry powder" - and if the government is offering a backstop to the data centre build out, then I can see the money being there.
I think the best example was in World War II - when the government funded all military equipment innovation until the war was won. Then once it was done, a whole bunch of planes were scrapped without ever been used.
My bear thesis for AI is that it's going to rapidly converge with the cost of electricity. Deflation.
So AI will not push up the price of electricity?
I'm not thinking of the price of electricity, but that the cost of running trained AI models will converge with the cost of electricity. Cheap to free, open source, trained models will deliver consumer surplus. At some point, if AI finds all the low hanging fruit and has built out electricity for intensive training, then eventually there should be a glut down the road. Maybe coal is a winner right now.
so a bit like the overbuild in broadband caused prices to collapse in the 2000s. Definitely possible - but generally speaking the cost to me of using ChapGPT and VEO is not that high. I feel both are loss making business at the moment
Brad Safalow pointed out Rocket Mortgages saving 20k person hours using AI. From the CC, "AI-Driven Productivity: Automated refinancing client follow-ups rose by nearly 20%, with AgenTeq AI processing over 80% of earnest money deposit validations, and saving an estimated 20,000 hours annually."
If there is more electricity consumption, you would expect more electrical grid instability. This should drive backup generator growth such as Generac (GNRC). I went long 2 months ago
Interesting. MS were pointing out that some bitcoin miner have value due to long term contracts with access to cheap electricity. They seemed to imply that was the logic behind CoreWeave buying Corse Scientific.
If the AI trend will push electricity prices, any other ideas to best play this trend?
I am thinking about this myself.
Like OKLO and NVTS?
Nuclear power and a company that develop energy efficiency on chips