The capex spending at MSFT, META, GOOG will eclipse $150 billion this year alone. These businesses are much different than they were 5 years ago with perhaps different return on capital profiles.
I haven't read the book, but I do think the moats could be getting stronger. My main question is what the returns on capital will be on their capital expenditures.
Companies have personalities like human beings. So, if you are an organization that was born in the 70s or early 80s, and have gone through a crisis or two, that memory is seared in the persona of the organization and how they allocate capital. You can put companies like Microsoft, Oracle, Apple in that genre.
Then you have those that only saw interest rates go one way (AMZN, META, GOOG, CRM) and they are only just learning the discipline of capital allocation, albeit reluctantly. I worked for four of them and as an employee you can totally feel the mindset. The famous one was Oracle where margins ruled supreme, and everyone knew management would gladly sacrifice revenue to keep margins. If you worked at CRM, you asked "what margins?"
I will make two interesting observations of this capex vis-a-vis internet build out:
1) How concentrated this AI capex cycle is and outside of a dozen companies, the rest can hardly afford it.
b) Not everyone needs AI Data center, and even if you do, you going to use a fraction of it. That means there is a high likelihood of a hiccup, and these investments can easily be put on hold.
Great comment. For b) by hiccup, you mean a slowdown in demand for AI? Not that I disagree, but I'd love to hear your views on why it would be easy for the big techs to easily cut back on these investments? Seems like a lot of these investments are going to carry a lot of recurring fixed costs to replace depreciated hardware. I could see how additional new growth spending could be cut, but the existing infrastructure they're building is going to carry much more fixed costs, no?
He makes the case that AI's impact on automation may be exaggerated, at least near term. I would concur with that opinion, not based on what he shares but from my personal and professional experience. The paper and the article in MIT Technology review is worth a read.
Yes, I meant slowdown in AI capex, specifically high end H100 or H200 type investments. The street is pretty unforgiving when margins go down and the hyperscalars are quite aware of it.
More broadly, there is a healthy skepticism of whether these broad use cases can be adequately monetized, both on the provider side as well as consumer (Enterprise). I dont mean to suggest that the AI data center build out will stop, but it won't happen at the pace it is happening right now and there will be lot of cheaper alternatives to high end GPUs that are purpose built.
I know this is anecdotal but just heard the Dell quarterly call and they added $1.7B of AI server revenue but zero margin expansion. And, promptly, the stock got crucified!
On a slightly diff note, I was looking at your 2017 and 2024 capex numbers for MSFT and the numbers you have are quite diff from what I get even if I add capex + R&D expenses. For example, for Microsoft, I see 2017 Capex + R&D in of $22B and $68B for 2024. Do you mind sharing the rationale for your numbers?
Are you referring to the table where I’m showing 2017 and 2024 capital invested? That’s a balance sheet number: the amount of capital that has been invested into the business to date. The capex is annual cash outflow number
Very insightful, John. I share your skepticism when I see those Capex numbers that need to translate to earnings. The good news is all four companies have very large user and customer bases to monetize the capex. Some have both.
Yeah the FCF already lags net income because of the capital expenditures. They are spending most of the FCF on buybacks, but at these valuations it won't do a whole lot to move the share count reduction needle
Yes, I've thought a lot about both those subjects. Probably a topic for another post, but yes I think there is competition in numerous places that all of these companies will have to deal with (hence the large capex spending on AI, META investing in entirely new businesses outside their core, etc...) On Apple, they have a great brand strength in China. I think the risk is more related to the geopolitical issues that no one really knows how to handicap. Apple's supply chain relies heavily on China (although that's a risk for lots of companies and not exclusively Apple)
Have you considered "The New Goliaths" by James Bessen? Seems like moats are being strengthened.
I haven't read the book, but I do think the moats could be getting stronger. My main question is what the returns on capital will be on their capital expenditures.
Companies have personalities like human beings. So, if you are an organization that was born in the 70s or early 80s, and have gone through a crisis or two, that memory is seared in the persona of the organization and how they allocate capital. You can put companies like Microsoft, Oracle, Apple in that genre.
Then you have those that only saw interest rates go one way (AMZN, META, GOOG, CRM) and they are only just learning the discipline of capital allocation, albeit reluctantly. I worked for four of them and as an employee you can totally feel the mindset. The famous one was Oracle where margins ruled supreme, and everyone knew management would gladly sacrifice revenue to keep margins. If you worked at CRM, you asked "what margins?"
I will make two interesting observations of this capex vis-a-vis internet build out:
1) How concentrated this AI capex cycle is and outside of a dozen companies, the rest can hardly afford it.
b) Not everyone needs AI Data center, and even if you do, you going to use a fraction of it. That means there is a high likelihood of a hiccup, and these investments can easily be put on hold.
Great comment. For b) by hiccup, you mean a slowdown in demand for AI? Not that I disagree, but I'd love to hear your views on why it would be easy for the big techs to easily cut back on these investments? Seems like a lot of these investments are going to carry a lot of recurring fixed costs to replace depreciated hardware. I could see how additional new growth spending could be cut, but the existing infrastructure they're building is going to carry much more fixed costs, no?
Thanks Murali
There have been some recent updates that are worth noting. The most notable is the work from MIT economist Daron Acemoglu in this paper: https://www.nber.org/system/files/working_papers/w32487/w32487.pdf
He makes the case that AI's impact on automation may be exaggerated, at least near term. I would concur with that opinion, not based on what he shares but from my personal and professional experience. The paper and the article in MIT Technology review is worth a read.
Yes, I meant slowdown in AI capex, specifically high end H100 or H200 type investments. The street is pretty unforgiving when margins go down and the hyperscalars are quite aware of it.
More broadly, there is a healthy skepticism of whether these broad use cases can be adequately monetized, both on the provider side as well as consumer (Enterprise). I dont mean to suggest that the AI data center build out will stop, but it won't happen at the pace it is happening right now and there will be lot of cheaper alternatives to high end GPUs that are purpose built.
I know this is anecdotal but just heard the Dell quarterly call and they added $1.7B of AI server revenue but zero margin expansion. And, promptly, the stock got crucified!
On a slightly diff note, I was looking at your 2017 and 2024 capex numbers for MSFT and the numbers you have are quite diff from what I get even if I add capex + R&D expenses. For example, for Microsoft, I see 2017 Capex + R&D in of $22B and $68B for 2024. Do you mind sharing the rationale for your numbers?
Thanks
Are you referring to the table where I’m showing 2017 and 2024 capital invested? That’s a balance sheet number: the amount of capital that has been invested into the business to date. The capex is annual cash outflow number
I see. Thanks for the clarification.
Very insightful, John. I share your skepticism when I see those Capex numbers that need to translate to earnings. The good news is all four companies have very large user and customer bases to monetize the capex. Some have both.
One possible implication of all this spend, is less for buybacks in the future. Buybacks have propelled the market.
Yeah the FCF already lags net income because of the capital expenditures. They are spending most of the FCF on buybacks, but at these valuations it won't do a whole lot to move the share count reduction needle
Have you considered China’s phone threat to Apple, Tic Tock to Google?
Yes, I've thought a lot about both those subjects. Probably a topic for another post, but yes I think there is competition in numerous places that all of these companies will have to deal with (hence the large capex spending on AI, META investing in entirely new businesses outside their core, etc...) On Apple, they have a great brand strength in China. I think the risk is more related to the geopolitical issues that no one really knows how to handicap. Apple's supply chain relies heavily on China (although that's a risk for lots of companies and not exclusively Apple)