Engines of growth reshaping the world, but then increasingly becoming rent collectors. The fortunes of the “Magnificent Seven”. What happens in the New Reality of Big Tech? And what are the answers...
As usual a lot of insight here. I was a little surprised though at the prominence of Azure in your MSFT story alongside the short shrift given to cloud at AMZN and GOOG. Amazon, not Microsoft, created the cloud infrastructure industry and remains its dominant leader. This is not controversial. Google's cloud unit is growing faster than either of the other two in the big 3, and faster than the rest of Google's business. (Also, as an aside, Alphabet happens to have a really expansive quantity of random innovation stuff happening outside its moneymachine ads business.)
Meanwhile, in contrast to the other two cloud units, Microsoft's Azure revenues/profits include a very high volume of screw turning in the form of licenses (esp SQL Server) that are now rolled up into "cloud" revenue for no real reason other than wanting to show growth in that unit. And MSFT still rests its dominance in software (and no small part of its Azure sales) on network effects of Windows and Office compatibility. I would submit that MSFT is at least as much, if not more, a rent-extraction enterprise as either of those two.
Brilliant. Insightful. But it ain't just the Mag 7. Everyone is taxing the choke points. Forget the name of the current occupant at 1600 Penn Ave. Do we believe the next Dem there will lower any of the tariffs? All up and down supply chains, there is more rent extraction at every node. Walmart used to get pilloried for squeezing its suppliers, now everyone does it. Firms find creative ways to use new tech to extract rent. Customers in every store happily scan and bag their own items while seeing no efficiency-induced reduced prices; they are now just unpaid employees. Universities now charge additional fees for things that used to be in tuition and bundle courses and outreach seminars to make fee-based "digital certificates" students and non-students can add to LinkedIn to signal some value that used to just be part of a transcript. DMV charges a convenience fee to register your car online even though a paper check is costlier to process. Rent is being extracted by everybody. It's actually quite astounding how this silent revolution came about, and Brad is dead-on...for investors , are you looking at valuations correctly?
Reminded me of what one of my work study students said about his father's engagement with MickyD. He developed a food that MD was interested in and had invested in equipment to produce it for MD. Then he read the contract!!!! They wanted all his profit. This was in the early 1980's.
Out of curiosity I asked ChatGPT what the situation was in 1880 and got this:
Rank Company Name Estimated Market Cap (1880 USD) Notes
1 Pennsylvania Railroad ~$120–130 million
2 New York Central Railroad ~$90–100 million
3 Illinois Central Railroad Railroads ~$70–80 million
4 Erie Railroad Railroads ~$50–60 million .
5 Union Pacific Railroad Railroads ~$40–50 million
6 Pullman Palace Car Company ~$40 million
7 Western Union Telegraph Co. ~$35–45 million
8 Standard Oil (Trust formed 1882) ~$40 million
9 American Bell Telephone Company ~$10–20 million
10 Delaware & Hudson Canal Co. ~$20–30 million
ChatGPT estimates these top ten were 60-70% of the stock market. I'm not up to fact-checking ChatGPT here. What is striking to me, however, is that, aside from Standard Oil, they are all communications industries that have platform/natural monopoly characteristics (counting Pullman as railroad-adjacent). Even the old tech of canals.
Also interesting to note is that many countries nationalized their railroad and later telephone systems, such was the political antipathy to these natural monopolies. The US is something of an outlier here.
I'll let others speculate on how this is related to the quality on modern US railroads compared to other industrialized nations, and what lessons we might draw about the long-term future of tech.
That sounds reasonable even if I don't trust ChatGPT to give correct answers. Those are mainly railroads. Thirty years or so later it was mainly steel companies. In the 20th century, railroads increasingly became real estate companies.
The top ten stocks account for a much bigger portion of the S&P 500 than ever. My guess is the suspension of the rule against buy backs. If you have the cash flow, you can support your stock price which is much more important than the company or economy in thirty years.
I have noticed that VC startups continue to define exit as being bought by one of the magnificent 7. Actually building a firm to enormous scale is no longer perceived as possible, or at least a waste of IRR.
In lines of the meme stock notion, how much does the savings glut play into these gaudy valuations? It's not like there are dividends to be had. MSFT wins with a whopping 0.65%, 3 offer zilch and NVIDIA almost zilch. Are people who don't need cash buying these so they can say "I'm in Amazon"?
Crypto is the invisible man here, with a valuation of $4 trillion more than any of the Seven except maybe Nvidia, despite being totally worthless. Tesla’s car and battery business also virtually worthless, and robotics is vaporware. That leaves six of the seven with profitable businesses, but no obvious way of achieving the growth implicit in their valuations. US financial market is one big meme stock.
Remember back in 1999 when Alan Greenspan was speaking of hedonic improvements and CNBC was deriding "bricks and mortar"? When the word "space" replaced "industry"?
The current ouroboros AI investment climate reminds me of that time.
The connective tether from the AI dream to our real sector reality is the electric wire powering that endeavor. Trump's turn away from Green Tech, at a time when China keeps dropping the cost of solar every 6 months is a disaster for the AI dream.
Electricity is the big input here.
So I asked Google AI in chrome:
What's the average price /kwh of Google data centers in North America
The search results do not provide the exact average price per kilowatt-hour (kWh) Google pays for electricity at their North American data centers. However, the search results do offer some insights into factors influencing data center electricity costs and general pricing trends:
Google's Electricity Consumption:
Google's data centers consumed 30.8 million megawatt-hours (MWh) of electricity in 2024, representing 95.8% of the company's total electricity consumption.
This is a significant increase, more than double the consumption in 2020 and seven times the consumption in 2014.
Data Center Power Costs in North America:
Electricity rates vary considerably depending on the location. For example, some regions have low industrial electricity rates under $0.047/kWh, particularly those with hydroelectric power. Other areas, particularly in the Northeast, can have rates exceeding $0.10/kWh.
Data centers are major electricity consumers, with some hyperscale data centers using as much power as 40,000 homes.
The increasing demand for data center power, driven by AI and other computing-intensive processes, is putting pressure on electricity grids and is expected to lead to higher energy prices.
As usual a lot of insight here. I was a little surprised though at the prominence of Azure in your MSFT story alongside the short shrift given to cloud at AMZN and GOOG. Amazon, not Microsoft, created the cloud infrastructure industry and remains its dominant leader. This is not controversial. Google's cloud unit is growing faster than either of the other two in the big 3, and faster than the rest of Google's business. (Also, as an aside, Alphabet happens to have a really expansive quantity of random innovation stuff happening outside its moneymachine ads business.)
Meanwhile, in contrast to the other two cloud units, Microsoft's Azure revenues/profits include a very high volume of screw turning in the form of licenses (esp SQL Server) that are now rolled up into "cloud" revenue for no real reason other than wanting to show growth in that unit. And MSFT still rests its dominance in software (and no small part of its Azure sales) on network effects of Windows and Office compatibility. I would submit that MSFT is at least as much, if not more, a rent-extraction enterprise as either of those two.
Brilliant. Insightful. But it ain't just the Mag 7. Everyone is taxing the choke points. Forget the name of the current occupant at 1600 Penn Ave. Do we believe the next Dem there will lower any of the tariffs? All up and down supply chains, there is more rent extraction at every node. Walmart used to get pilloried for squeezing its suppliers, now everyone does it. Firms find creative ways to use new tech to extract rent. Customers in every store happily scan and bag their own items while seeing no efficiency-induced reduced prices; they are now just unpaid employees. Universities now charge additional fees for things that used to be in tuition and bundle courses and outreach seminars to make fee-based "digital certificates" students and non-students can add to LinkedIn to signal some value that used to just be part of a transcript. DMV charges a convenience fee to register your car online even though a paper check is costlier to process. Rent is being extracted by everybody. It's actually quite astounding how this silent revolution came about, and Brad is dead-on...for investors , are you looking at valuations correctly?
Reminded me of what one of my work study students said about his father's engagement with MickyD. He developed a food that MD was interested in and had invested in equipment to produce it for MD. Then he read the contract!!!! They wanted all his profit. This was in the early 1980's.
Out of curiosity I asked ChatGPT what the situation was in 1880 and got this:
Rank Company Name Estimated Market Cap (1880 USD) Notes
1 Pennsylvania Railroad ~$120–130 million
2 New York Central Railroad ~$90–100 million
3 Illinois Central Railroad Railroads ~$70–80 million
4 Erie Railroad Railroads ~$50–60 million .
5 Union Pacific Railroad Railroads ~$40–50 million
6 Pullman Palace Car Company ~$40 million
7 Western Union Telegraph Co. ~$35–45 million
8 Standard Oil (Trust formed 1882) ~$40 million
9 American Bell Telephone Company ~$10–20 million
10 Delaware & Hudson Canal Co. ~$20–30 million
ChatGPT estimates these top ten were 60-70% of the stock market. I'm not up to fact-checking ChatGPT here. What is striking to me, however, is that, aside from Standard Oil, they are all communications industries that have platform/natural monopoly characteristics (counting Pullman as railroad-adjacent). Even the old tech of canals.
Also interesting to note is that many countries nationalized their railroad and later telephone systems, such was the political antipathy to these natural monopolies. The US is something of an outlier here.
I'll let others speculate on how this is related to the quality on modern US railroads compared to other industrialized nations, and what lessons we might draw about the long-term future of tech.
That sounds reasonable even if I don't trust ChatGPT to give correct answers. Those are mainly railroads. Thirty years or so later it was mainly steel companies. In the 20th century, railroads increasingly became real estate companies.
For an interesting chart of market concentration see https://www.visualcapitalist.com/charted-sp-500-market-concentration-over-145-years/
The top ten stocks account for a much bigger portion of the S&P 500 than ever. My guess is the suspension of the rule against buy backs. If you have the cash flow, you can support your stock price which is much more important than the company or economy in thirty years.
I have noticed that VC startups continue to define exit as being bought by one of the magnificent 7. Actually building a firm to enormous scale is no longer perceived as possible, or at least a waste of IRR.
Ed Zitron's been writing about this.
https://www.wheresyoured.at/the-haters-gui/
Thx much... - B.
A nice chart showing stock market value concentration over the last 145 years. We are in uncharted territory.
Has anyone ever constructed a sort of Gini index for the market over time?
https://www.visualcapitalist.com/charted-sp-500-market-concentration-over-145-years/
A very nice Goldman Sachs report on market concentration, with a comparison of other country stock markets. Surprisingly, the USA is near the bottom, i.e., it is less concentrated. https://www.goldmansachs.com/pdfs/insights/pages/top-of-mind/market-concentration-how-big-a-worry/report.pdf
In lines of the meme stock notion, how much does the savings glut play into these gaudy valuations? It's not like there are dividends to be had. MSFT wins with a whopping 0.65%, 3 offer zilch and NVIDIA almost zilch. Are people who don't need cash buying these so they can say "I'm in Amazon"?
Crypto is the invisible man here, with a valuation of $4 trillion more than any of the Seven except maybe Nvidia, despite being totally worthless. Tesla’s car and battery business also virtually worthless, and robotics is vaporware. That leaves six of the seven with profitable businesses, but no obvious way of achieving the growth implicit in their valuations. US financial market is one big meme stock.
Remember back in 1999 when Alan Greenspan was speaking of hedonic improvements and CNBC was deriding "bricks and mortar"? When the word "space" replaced "industry"?
The current ouroboros AI investment climate reminds me of that time.
The connective tether from the AI dream to our real sector reality is the electric wire powering that endeavor. Trump's turn away from Green Tech, at a time when China keeps dropping the cost of solar every 6 months is a disaster for the AI dream.
Electricity is the big input here.
So I asked Google AI in chrome:
What's the average price /kwh of Google data centers in North America
The search results do not provide the exact average price per kilowatt-hour (kWh) Google pays for electricity at their North American data centers. However, the search results do offer some insights into factors influencing data center electricity costs and general pricing trends:
Google's Electricity Consumption:
Google's data centers consumed 30.8 million megawatt-hours (MWh) of electricity in 2024, representing 95.8% of the company's total electricity consumption.
This is a significant increase, more than double the consumption in 2020 and seven times the consumption in 2014.
Data Center Power Costs in North America:
Electricity rates vary considerably depending on the location. For example, some regions have low industrial electricity rates under $0.047/kWh, particularly those with hydroelectric power. Other areas, particularly in the Northeast, can have rates exceeding $0.10/kWh.
Data centers are major electricity consumers, with some hyperscale data centers using as much power as 40,000 homes.
The increasing demand for data center power, driven by AI and other computing-intensive processes, is putting pressure on electricity grids and is expected to lead to higher energy prices.