Hi guys, interesting article, I enjoyed the case study to explain these concepts. However, I do have a few questions/observations that perhaps can help going into more detail and improve our understanding.
1. Over how long did you amortize the capitalized intangibles for the NOPAT calculation? Im curious as I know there's no "correct" way of doing this although Mauboussin uses a study-backed method.
2. When you say "the rapid expansion of the asset base has been matched by NOPAT driven by AI-enchanced monetization". I think the analysis isn't really giving you the true picture of the returns of the data centers, and more generally the new AI investments. To get an accurate idea of the ROIIC, you would need to know exactly to what extent the new CapEx has contributed to the new NOPAT - which is difficult to obtain the data for. I think it's difficult to come to the conclusion that the new investments have paid off to the extent you say. I think it's more likely that the NOPAT can be attributed to the other reasons you mention i.e. growth in their other activities + cost cutting... I believe we still have quite a lot of time in front of us before seeing any material impact of the AI investments currently being made, don't you think?
3. Also, specifically with Meta - in case this is of interest to you - it may be worth reminding that a significant part of their investments in data centers are being done through joint ventures (see their recent collab with Blue Owl) as a way of not taking on more debt on their consolidated FAs. If the goal is to estimate the true returns of such activities, perhaps this would be worth taking into account... as an idea.
Let me know your thoughts. In any case, interesting article! Keep up the work!
Thank you for your feedback, always happy to talk investing. I’ll address a few of your points from my side, with the caveat that The Pursuit of Compounding knows Meta much better than I do and may add to or correct anything I miss.
On your second point, I fully agree: we can’t really isolate “the ROIIC of the data centers” from the outside. Only an insider could do that properly. We could have pushed further by building scenarios, tweaking assumptions, and trying to back into a more precise estimate, but we made a deliberate choice to keep the focus on the ROIC/ROIIC framework itself and to use Meta as a concrete illustration of how the combined metrics can tell slightly different stories. Put simply, the subject of the article is more the framework than Meta.
On the JVs, it’s a similar story: you’re right. Even though they sit outside our analysis window of 2020/2024 (there were no meaningful AI JVs for Meta over that period), they clearly become important once you start assessing the 2025+ investment cycle. But again, we’ll run into the same issue: limited disclosure. It’s very likely that what we’ll mostly see is Meta’s investment line in these JVs, without standalone financials for the vehicles themselves. So we’ll still have to balance accuracy, usefulness, and simplicity if we want to arrive at numbers that are at least directionally exploitable.
Thanks again for taking the time to write such a detailed comment. This kind of exchange is exactly what makes Substack worthwhile.
Great post. I'm going to add links to this write-up whenever I touch on ROIIC in my posts! This is a definitive piece of literature on both metrics.
In terms of the invested capital issue with capital light companies like software and pharmaceuticals, Yefei Lu discusses how to add back accumulated amortization in the book, Inside the Investments of Warren Buffett. That said, I think you handled it well in the META example.
I like how you points out the flaws of both metrics. There's no God equation in investing. Enterprising investors have to be ready to make necessary adjustments.
Finally, I have a question. The operating/excess cash item in your calculations. I can see you used 2% of revenue to arrive at operating cash. Is this a rule of thumb or the result of a calculation?
I really like that line: “There’s no God equation in investing. Enterprising investors have to be ready to make necessary adjustments.” It would have made a great conclusion.
Thanks for these insights, and thank you for your feedback and for taking the time to write it. I’m glad you liked our post!
Thorough and methodical breakdown—especially the note on Meta’s shift to negative net working capital as a funding strategy. That operational lever (stretching payables while collecting receivables quickly) is a classic trade credit and liquidity tactic that many B2B finance teams monitor closely. TCLM often explores how such working capital discipline intersects with broader capital allocation and risk. A valuable read for finance leaders.
Hi guys, interesting article, I enjoyed the case study to explain these concepts. However, I do have a few questions/observations that perhaps can help going into more detail and improve our understanding.
1. Over how long did you amortize the capitalized intangibles for the NOPAT calculation? Im curious as I know there's no "correct" way of doing this although Mauboussin uses a study-backed method.
2. When you say "the rapid expansion of the asset base has been matched by NOPAT driven by AI-enchanced monetization". I think the analysis isn't really giving you the true picture of the returns of the data centers, and more generally the new AI investments. To get an accurate idea of the ROIIC, you would need to know exactly to what extent the new CapEx has contributed to the new NOPAT - which is difficult to obtain the data for. I think it's difficult to come to the conclusion that the new investments have paid off to the extent you say. I think it's more likely that the NOPAT can be attributed to the other reasons you mention i.e. growth in their other activities + cost cutting... I believe we still have quite a lot of time in front of us before seeing any material impact of the AI investments currently being made, don't you think?
3. Also, specifically with Meta - in case this is of interest to you - it may be worth reminding that a significant part of their investments in data centers are being done through joint ventures (see their recent collab with Blue Owl) as a way of not taking on more debt on their consolidated FAs. If the goal is to estimate the true returns of such activities, perhaps this would be worth taking into account... as an idea.
Let me know your thoughts. In any case, interesting article! Keep up the work!
Hi Paul,
Thank you for your feedback, always happy to talk investing. I’ll address a few of your points from my side, with the caveat that The Pursuit of Compounding knows Meta much better than I do and may add to or correct anything I miss.
On your second point, I fully agree: we can’t really isolate “the ROIIC of the data centers” from the outside. Only an insider could do that properly. We could have pushed further by building scenarios, tweaking assumptions, and trying to back into a more precise estimate, but we made a deliberate choice to keep the focus on the ROIC/ROIIC framework itself and to use Meta as a concrete illustration of how the combined metrics can tell slightly different stories. Put simply, the subject of the article is more the framework than Meta.
On the JVs, it’s a similar story: you’re right. Even though they sit outside our analysis window of 2020/2024 (there were no meaningful AI JVs for Meta over that period), they clearly become important once you start assessing the 2025+ investment cycle. But again, we’ll run into the same issue: limited disclosure. It’s very likely that what we’ll mostly see is Meta’s investment line in these JVs, without standalone financials for the vehicles themselves. So we’ll still have to balance accuracy, usefulness, and simplicity if we want to arrive at numbers that are at least directionally exploitable.
Thanks again for taking the time to write such a detailed comment. This kind of exchange is exactly what makes Substack worthwhile.
Great post. I'm going to add links to this write-up whenever I touch on ROIIC in my posts! This is a definitive piece of literature on both metrics.
In terms of the invested capital issue with capital light companies like software and pharmaceuticals, Yefei Lu discusses how to add back accumulated amortization in the book, Inside the Investments of Warren Buffett. That said, I think you handled it well in the META example.
I like how you points out the flaws of both metrics. There's no God equation in investing. Enterprising investors have to be ready to make necessary adjustments.
Finally, I have a question. The operating/excess cash item in your calculations. I can see you used 2% of revenue to arrive at operating cash. Is this a rule of thumb or the result of a calculation?
Thank you for the feedback, happy it created some value! Thanks for the suggestion about Yefei Lu, I will check that out.
To answer your cash question, we used Mauboussin as the source material for the rationale in doing the adjustments and the calculations.
He defines excess cash as any cash greater than 2% of revenue, so that’s where that comes from. Great question!
I really like that line: “There’s no God equation in investing. Enterprising investors have to be ready to make necessary adjustments.” It would have made a great conclusion.
Thanks for these insights, and thank you for your feedback and for taking the time to write it. I’m glad you liked our post!
Thorough and methodical breakdown—especially the note on Meta’s shift to negative net working capital as a funding strategy. That operational lever (stretching payables while collecting receivables quickly) is a classic trade credit and liquidity tactic that many B2B finance teams monitor closely. TCLM often explores how such working capital discipline intersects with broader capital allocation and risk. A valuable read for finance leaders.
(It’s free)- https://tradecredit.substack.com/
Thank you for this answer.
The excess cash question had been bouncing around in my head for a couple of years now.
https://open.substack.com/pub/absolutetoal/p/roic-1gni1gic1-and-roiic?utm_source=share&utm_medium=android&r=5g11d4