You're right about the vibes. I would say 2021 had to have gotten close to dotcom levels of euphoria. NFTs were the beanie baby equivalent. SPACs had crazy valuations. But I guess that market run didn't go on long enough for those crazy companies to actually infiltrate the index level. So when they burst (and that bubble did burst), it was its own small pocket.
And as far as a broad bubble bursting, I just don't know where I see the catalyst. Again, you're right that earnings expectations are pretty nuts. But even if the AI spending spree comes to a halt and the mag 7 rerate and get cut in half, that's still only a 18% decline in markets. But even in that case, I just don't see a world where Microsoft and Google are trading at 15x earnings...even if AI flames out.
I feel like we need some type of credit event to see a 'real' crash. We basically need to see huge swaths of companies disappearing overnight to get anything beyond a moderate correction. But ALL of these companies are so strapped with cash that I just don't see how that could happen. Even the meme stocks and former SPACS have tons of cash. Zoom has $10 Billion in net cash. Docusign has over a billion. Peloton has a good bit of debt but can certainly tread water for years with their manageable debt levels. Gamestop's business model should have it on life support, and yet, even they have close to 4 billion in net cash. They can basically keep operations running for another 4+ years and have no worries about liquidity.
Peter Lynch said "It's very hard to go bankrupt when you don't have any debt."
CAPE and treasury yields have been my barometer for valuation and expected returns.
This was my impetus to buy stocks in 2020. In March of that year, equity earnings yields were almost 5% while bonds were basically zero. It was (by this metric) the best time to buy since the bottom of the GFC. This metric has been flashing red since 2023.
But there's a lot of things that could bring that overvaluation down. One would simply be bonds yields falling. If the 10-year rate happened to fall back under 2%, stocks wouldn't really be expensive (this would be great for bond holders obviously). The other is equities could simply trade sideways while earnings growth and inflation erode the CAPE ratio.
I think it's no use calling for a bubble yet. Just own a lot of different types of assets and avoid FOMO. I don't see a world where we regret holding some Bonds and International equities at these levels.
I guess we know more or less the S&P 500 companies are not overpriced as a whole, except for the mag 7, which account for 35% of the 500 companies' total market capitalization now.
In order to have a good idea of whether the mag 7 is overvalued, I looked into each one of their financial statements and obtained the R&D figures from 2014 to 2023.
Total R&D was $41bn in 2014 and $239bn in 2023, which translates into 22% annual growth. If we amortize it by 7 years with zero residual value, the amortization cost would be $144bn in 2023, $95bn lower than 2023's R&D expenses.
Assuming the growth in 2024 is also 22%, total R&D would be $291bn in 2024 and amortization cost would be $175bn, $116bn lower than R&D costs.
Mag 7's total net profit TTM is $496bn, and current market cap is $18.2tn, so PER is 36.7x. If we adjust the net profit to $613bn (496+116), the PER is 30x. Given today's 10y treasury yield of 4.6%, it is not cheap. But it is also not extraordinary expensive, considering how fast their profits have been growing.
I am also an old style value investor, and just start to learn about the tech companies. So far I only know about google, because I use its products a lot. They are really sticky. Based on my old style value investment judgement, Google has a ecosphere with strong moat. How come I miss something so close to me?
As part of the learning journey, I recently tried different AI assistants, and am amazed by the quality of their services and products, and the speed they are improving. That enhanced my confidence that AI will tremendously increase the productivity of society, which would most likely contributes to the earning of mag 7.
I recently read a book, Where the Money is - Value Investing in Digital Age. It has a very interesting opinion, which I tend to agree: the R&D expenses of the tech companies should be capitalized. As we know, R&D is usually 100% expensed in P&L in the year it happens, although it generates earnings over a number of years, same as an investment in tangible fixed assets. As a matter of fact, since 2022, US IRS required the tech companies to capitalize R&D expenses and amortize them over 5 years. When I used deferred tax assets to get capitalized R&D, and use that to adjust the net earning, the PER TTM of Alphabet dropped below 20x. Alphabet's market capitalization has been growing at a pace of 20% p.a. since 2016, in line with its net profit (not adjusted with R&D capitalization).
I have not looked into the rest of the mag 7, because for a long time I thought they were too expensive, especially Amazon, which had a long history of losing money. Has Mr. Market really been irrational for so long? I think they deserve a refreshed look.
By the way, the same goes for a country. When the GDP does not include the intangible economic benefits, Buffett Indicator may not be reliable any more.
Yeah that point on capitalisation of software development is definitely worth making. I agree - a lot of software development is an investment and should be treated as such. I'm not sure how much of a difference this would make to the numbers though - here's an estimation:
The Mag 7 spent $242b on R&D in 2023 (https://www.ft.com/content/95065b2b-901d-4f77-aab1-6e489e04d1a3), 40% of the S&P500 total. That has compounded at an 11.5% CAGR since 2014. Non-M7 companies obviously also spend on software R&D, but also, not all of that $242b will be software development, and not all of it deserves to be capitalised, so we'll stick with $242b.
We also need to decide what timeframe to depreciate it over. I'll pull a number out of my arse and say 7 years. Based on the 11.5% growth rate, that would result in $160b of depreciation in 2023, leaving an $82b difference to be added to earnings. We'll assume that's 11.5% bigger again for 2024, so $91b.
S&P500 market cap is $52.2t, multiple is 29, so earnings are about $1.8t. Hence this $91b would reduce the multiple from 29 to 27.6.
But, you also have to consider that in past period there was also significant research expense - though less on software, more on industrial products - which was expensed instead of capitalised back then. So the "true" difference is probably less than 1 point on the multiple, based on my approximations.
All of those are subject to debate though so let me know if you disagree.
In addition to my reply just now, I would like to add that the Information Technology sector, as defined by the Global Industry Classification Standard (GICS), includes 69 companies within the S&P 500. That doesn't include Amazon (Consumer Discretionary) and Alphabet (Communication Services). So if we want to calculate the actual R&D capitalization impact on the SP500 PER, adjusting only the R&D expenses of mag 7 could end up underestimating the impact.
Please allow me to make the opposite case. Kindly indulge me. Historical 10 year return: SPX: 13.5%, Equal weighted SP500 10%, EFA and EEM 5%. So with the sole exception of SPX all longer term returns globally seem low to OK. They dont look inflated. What looks very inflated is the MAG7. The MAG 7 has compounded at 45% p.a for the last three years while the S&P 493 has just done low single digit return annually for the last two years. All of the above seems to indicate that the broad market seems fine while MAG 7 has rocked. Ironically the MAG 7 may be some of the best companies in the world where it is tough to poke holes at a 14-25X EBITDA and 25-40X earnings multiple. Tesla is the only exception at much higher multiples. So I look at eveything and end up feeling it all looks OK. Why should great companies be cheap? Market always has pockets of extreme optimism and pessimism but that is just normal. My take is that without a recession market returns will likely be average not low.
I think pretty much everything you said would have applied in the dot-com bubble too. A small number of large companies leading the charge, check. The vast majority of companies performing only averagely, check. Those large companies being, for the most part, very high quality businesses - check.
Of course, it wasn't then and isn't now only the top 7. There's all the other AI stocks, all the quantum computing stocks, many restaurants (as well as Cava you have Wingstop, Chipotle, Sweetgreen). Anything crypto. There's a quality bubble too - Costco at 55x, for example. Oh, and nuclear stocks. Am I missing anything?
I don't think what we're seeing right now is a normal quantity of pockets of extreme optimism; nor a normal degree of overvaluation within those pockets.
But I appreciate you sharing your view - it's always good to hear the other side. And I certainly don't want to imply I don't think there are any undervalued US companies. (You probably have to dig into the cyclicals for most of those, though.)
If AI brings the improvements in productivity that we hope, I agree that it's possible that the next 10 years deliver average returns.
Great note, very informative.
You're right about the vibes. I would say 2021 had to have gotten close to dotcom levels of euphoria. NFTs were the beanie baby equivalent. SPACs had crazy valuations. But I guess that market run didn't go on long enough for those crazy companies to actually infiltrate the index level. So when they burst (and that bubble did burst), it was its own small pocket.
And as far as a broad bubble bursting, I just don't know where I see the catalyst. Again, you're right that earnings expectations are pretty nuts. But even if the AI spending spree comes to a halt and the mag 7 rerate and get cut in half, that's still only a 18% decline in markets. But even in that case, I just don't see a world where Microsoft and Google are trading at 15x earnings...even if AI flames out.
I feel like we need some type of credit event to see a 'real' crash. We basically need to see huge swaths of companies disappearing overnight to get anything beyond a moderate correction. But ALL of these companies are so strapped with cash that I just don't see how that could happen. Even the meme stocks and former SPACS have tons of cash. Zoom has $10 Billion in net cash. Docusign has over a billion. Peloton has a good bit of debt but can certainly tread water for years with their manageable debt levels. Gamestop's business model should have it on life support, and yet, even they have close to 4 billion in net cash. They can basically keep operations running for another 4+ years and have no worries about liquidity.
Peter Lynch said "It's very hard to go bankrupt when you don't have any debt."
CAPE and treasury yields have been my barometer for valuation and expected returns.
https://riskpremium.substack.com/p/to-bond-or-not-to-bond
This was my impetus to buy stocks in 2020. In March of that year, equity earnings yields were almost 5% while bonds were basically zero. It was (by this metric) the best time to buy since the bottom of the GFC. This metric has been flashing red since 2023.
But there's a lot of things that could bring that overvaluation down. One would simply be bonds yields falling. If the 10-year rate happened to fall back under 2%, stocks wouldn't really be expensive (this would be great for bond holders obviously). The other is equities could simply trade sideways while earnings growth and inflation erode the CAPE ratio.
I think it's no use calling for a bubble yet. Just own a lot of different types of assets and avoid FOMO. I don't see a world where we regret holding some Bonds and International equities at these levels.
I guess we know more or less the S&P 500 companies are not overpriced as a whole, except for the mag 7, which account for 35% of the 500 companies' total market capitalization now.
In order to have a good idea of whether the mag 7 is overvalued, I looked into each one of their financial statements and obtained the R&D figures from 2014 to 2023.
Total R&D was $41bn in 2014 and $239bn in 2023, which translates into 22% annual growth. If we amortize it by 7 years with zero residual value, the amortization cost would be $144bn in 2023, $95bn lower than 2023's R&D expenses.
Assuming the growth in 2024 is also 22%, total R&D would be $291bn in 2024 and amortization cost would be $175bn, $116bn lower than R&D costs.
Mag 7's total net profit TTM is $496bn, and current market cap is $18.2tn, so PER is 36.7x. If we adjust the net profit to $613bn (496+116), the PER is 30x. Given today's 10y treasury yield of 4.6%, it is not cheap. But it is also not extraordinary expensive, considering how fast their profits have been growing.
I am also an old style value investor, and just start to learn about the tech companies. So far I only know about google, because I use its products a lot. They are really sticky. Based on my old style value investment judgement, Google has a ecosphere with strong moat. How come I miss something so close to me?
As part of the learning journey, I recently tried different AI assistants, and am amazed by the quality of their services and products, and the speed they are improving. That enhanced my confidence that AI will tremendously increase the productivity of society, which would most likely contributes to the earning of mag 7.
I recently read a book, Where the Money is - Value Investing in Digital Age. It has a very interesting opinion, which I tend to agree: the R&D expenses of the tech companies should be capitalized. As we know, R&D is usually 100% expensed in P&L in the year it happens, although it generates earnings over a number of years, same as an investment in tangible fixed assets. As a matter of fact, since 2022, US IRS required the tech companies to capitalize R&D expenses and amortize them over 5 years. When I used deferred tax assets to get capitalized R&D, and use that to adjust the net earning, the PER TTM of Alphabet dropped below 20x. Alphabet's market capitalization has been growing at a pace of 20% p.a. since 2016, in line with its net profit (not adjusted with R&D capitalization).
I have not looked into the rest of the mag 7, because for a long time I thought they were too expensive, especially Amazon, which had a long history of losing money. Has Mr. Market really been irrational for so long? I think they deserve a refreshed look.
By the way, the same goes for a country. When the GDP does not include the intangible economic benefits, Buffett Indicator may not be reliable any more.
Yeah that point on capitalisation of software development is definitely worth making. I agree - a lot of software development is an investment and should be treated as such. I'm not sure how much of a difference this would make to the numbers though - here's an estimation:
The Mag 7 spent $242b on R&D in 2023 (https://www.ft.com/content/95065b2b-901d-4f77-aab1-6e489e04d1a3), 40% of the S&P500 total. That has compounded at an 11.5% CAGR since 2014. Non-M7 companies obviously also spend on software R&D, but also, not all of that $242b will be software development, and not all of it deserves to be capitalised, so we'll stick with $242b.
We also need to decide what timeframe to depreciate it over. I'll pull a number out of my arse and say 7 years. Based on the 11.5% growth rate, that would result in $160b of depreciation in 2023, leaving an $82b difference to be added to earnings. We'll assume that's 11.5% bigger again for 2024, so $91b.
S&P500 market cap is $52.2t, multiple is 29, so earnings are about $1.8t. Hence this $91b would reduce the multiple from 29 to 27.6.
But, you also have to consider that in past period there was also significant research expense - though less on software, more on industrial products - which was expensed instead of capitalised back then. So the "true" difference is probably less than 1 point on the multiple, based on my approximations.
All of those are subject to debate though so let me know if you disagree.
In addition to my reply just now, I would like to add that the Information Technology sector, as defined by the Global Industry Classification Standard (GICS), includes 69 companies within the S&P 500. That doesn't include Amazon (Consumer Discretionary) and Alphabet (Communication Services). So if we want to calculate the actual R&D capitalization impact on the SP500 PER, adjusting only the R&D expenses of mag 7 could end up underestimating the impact.
Please allow me to make the opposite case. Kindly indulge me. Historical 10 year return: SPX: 13.5%, Equal weighted SP500 10%, EFA and EEM 5%. So with the sole exception of SPX all longer term returns globally seem low to OK. They dont look inflated. What looks very inflated is the MAG7. The MAG 7 has compounded at 45% p.a for the last three years while the S&P 493 has just done low single digit return annually for the last two years. All of the above seems to indicate that the broad market seems fine while MAG 7 has rocked. Ironically the MAG 7 may be some of the best companies in the world where it is tough to poke holes at a 14-25X EBITDA and 25-40X earnings multiple. Tesla is the only exception at much higher multiples. So I look at eveything and end up feeling it all looks OK. Why should great companies be cheap? Market always has pockets of extreme optimism and pessimism but that is just normal. My take is that without a recession market returns will likely be average not low.
I think pretty much everything you said would have applied in the dot-com bubble too. A small number of large companies leading the charge, check. The vast majority of companies performing only averagely, check. Those large companies being, for the most part, very high quality businesses - check.
Of course, it wasn't then and isn't now only the top 7. There's all the other AI stocks, all the quantum computing stocks, many restaurants (as well as Cava you have Wingstop, Chipotle, Sweetgreen). Anything crypto. There's a quality bubble too - Costco at 55x, for example. Oh, and nuclear stocks. Am I missing anything?
I don't think what we're seeing right now is a normal quantity of pockets of extreme optimism; nor a normal degree of overvaluation within those pockets.
But I appreciate you sharing your view - it's always good to hear the other side. And I certainly don't want to imply I don't think there are any undervalued US companies. (You probably have to dig into the cyclicals for most of those, though.)
If AI brings the improvements in productivity that we hope, I agree that it's possible that the next 10 years deliver average returns.
SP493 has done mid to high single digit returns annually for the last two years not the low single digit as stated above. That is an error!