I know it's just surface level, but at least you can see that it's possible here. I'm gonna attempt to do my own surface-level valuation of NVDA pretty soon, and I'll try to remember to report back here. Thanks for helping get that ball rolling!
Great piece. I like your conclusion "However, it surprises me that even at a face-ripping $3T market cap, it remains theoretically possible to make a case for NVDA being fairly valued today. I honestly wasn’t expecting that at all."
I wrote a similar article (https://oninvesting.substack.com/p/the-price-is-right) although a lot less complex and came to a similar conclusion to yours. Although probably way too high, it's not as crazy as it seems.
I've thought about it some more. You're right that when competition shows up in 5 years, the margins will drop. The question I'm wondering about is volumes.
This setup looks like the 90s when Intel had dominant market share (90%ish) and AMD was playing catchup. Though chips were a commodity at that point, through increased volumes Intel was able to spend more on R&D and create better chips than AMD. And even when AMD was able to create a better chip, the switching costs for most users was not worth it. Intel was the brand and customers were willing to wait for a few months for Intel to catch up. All Intel had to do was create a chip that was equal to AMD, not necessarily better (lowering their R&D costs even more), and still maintain market share. This created a flywheel effect where Intel was able to just keep on going and AMD had a very hard time ever catching up.
Even if competitors show up, NVIDIA will be the gold standard and switching costs will be high. So even if margins fall, volume should be huge for NVIDIA which should allow them to keep the growth machine going. I think that projecting industry growth rather than NVIDIA's growth might be the way to do this, and just give NVIDIA 90% market share after 5 years. I don't know how different that is for your numbers, but I think this starts to veer off in the direction of projecting future AI uses, such as whether AGI is possible etc.
Definitely agree with the overall theme and takeaway of this post, and I think you did a good job of keeping it balanced.
Just a couple things to mention - first, I would have thought the main cause of their rise in margins in recent years is not operating leverage (which is generally understood as spreading fixed costs over a greater number of units) but rather price hikes due to the excess demand.
Second, I still find your valuation models very hard to interpret (is your earnings payback model undiscounted?). I continue to think if you're gonna go to the effort of doing all the numbers (and not just going to give a normalised current PE, which to be honest is almost always what I do) then it only makes sense to present it as a DCF, or in this case a reverse DCF. So that would be "current earnings are $x, current market cap is $y, what combination of next 5/10 year growth, perpetuity growth, and discount rate reconciles those two numbers?" - and then it gives you a pretty good indication of what the market is thinking, and what has to go right/wrong for the market to change its mind.
But this is not intended as an insult/argument, just a couple pointers.
Yes you're right, that's actually the typical definition of operating leverage. But I was trying to think of something which didn't require a mouthful to describe, and that term just came naturally to communicate that metric.
Well to keep it short, I've basically simplified the entire DCF model down to a PE ratio. There's usually no point being ultra-precise about these things anyway, as Howard Marks said, it's better to be approximately correct than precisely wrong.
What I've done instead is show a range of results (i.e. scenario analysis), so that readers can get a sense of where my head is at regarding valuation. Also helps readers do the extra legwork of looking at multiple valuations under different assumptions.
And no worries, I totally appreciate where you're coming from! Happy to take more questions.
Understood. I do think possibly in simplifying it down the way you have it becomes a bit hard to explain what a certain number actually means - I don't really have any framework to evaluate "PE 15 in 2030" - but maybe that's just me.
On the subject of Nvidia, one topic which I would be interested to hear more on is CUDA, as it often comes up as the thing that's gonna ensure Nvidia can remain way ahead of all the other chipmakers. If you're planning a follow up to this one I think that would be a great thing to talk about.
Sure, points noted. 15x PE in 2030 simply means the forward PE in 2030, i.e. today's market cap / est. earnings in 2030 = 15x. It's useful for companies with nonsensical or negative trailing PEs, like NVDA's 100x today.
I think someone else actually wrote a pretty great article about CUDA recently! I can't seem to find it now, but maybe do a search for CUDA on Substack (and perhaps Google).
I know it's just surface level, but at least you can see that it's possible here. I'm gonna attempt to do my own surface-level valuation of NVDA pretty soon, and I'll try to remember to report back here. Thanks for helping get that ball rolling!
Great piece. I like your conclusion "However, it surprises me that even at a face-ripping $3T market cap, it remains theoretically possible to make a case for NVDA being fairly valued today. I honestly wasn’t expecting that at all."
I wrote a similar article (https://oninvesting.substack.com/p/the-price-is-right) although a lot less complex and came to a similar conclusion to yours. Although probably way too high, it's not as crazy as it seems.
Great! Saved for later reading!
I've thought about it some more. You're right that when competition shows up in 5 years, the margins will drop. The question I'm wondering about is volumes.
This setup looks like the 90s when Intel had dominant market share (90%ish) and AMD was playing catchup. Though chips were a commodity at that point, through increased volumes Intel was able to spend more on R&D and create better chips than AMD. And even when AMD was able to create a better chip, the switching costs for most users was not worth it. Intel was the brand and customers were willing to wait for a few months for Intel to catch up. All Intel had to do was create a chip that was equal to AMD, not necessarily better (lowering their R&D costs even more), and still maintain market share. This created a flywheel effect where Intel was able to just keep on going and AMD had a very hard time ever catching up.
Even if competitors show up, NVIDIA will be the gold standard and switching costs will be high. So even if margins fall, volume should be huge for NVIDIA which should allow them to keep the growth machine going. I think that projecting industry growth rather than NVIDIA's growth might be the way to do this, and just give NVIDIA 90% market share after 5 years. I don't know how different that is for your numbers, but I think this starts to veer off in the direction of projecting future AI uses, such as whether AGI is possible etc.
Just my 2 cents. Otherwise great article.
Definitely agree with the overall theme and takeaway of this post, and I think you did a good job of keeping it balanced.
Just a couple things to mention - first, I would have thought the main cause of their rise in margins in recent years is not operating leverage (which is generally understood as spreading fixed costs over a greater number of units) but rather price hikes due to the excess demand.
Second, I still find your valuation models very hard to interpret (is your earnings payback model undiscounted?). I continue to think if you're gonna go to the effort of doing all the numbers (and not just going to give a normalised current PE, which to be honest is almost always what I do) then it only makes sense to present it as a DCF, or in this case a reverse DCF. So that would be "current earnings are $x, current market cap is $y, what combination of next 5/10 year growth, perpetuity growth, and discount rate reconciles those two numbers?" - and then it gives you a pretty good indication of what the market is thinking, and what has to go right/wrong for the market to change its mind.
But this is not intended as an insult/argument, just a couple pointers.
Yes you're right, that's actually the typical definition of operating leverage. But I was trying to think of something which didn't require a mouthful to describe, and that term just came naturally to communicate that metric.
Well to keep it short, I've basically simplified the entire DCF model down to a PE ratio. There's usually no point being ultra-precise about these things anyway, as Howard Marks said, it's better to be approximately correct than precisely wrong.
What I've done instead is show a range of results (i.e. scenario analysis), so that readers can get a sense of where my head is at regarding valuation. Also helps readers do the extra legwork of looking at multiple valuations under different assumptions.
And no worries, I totally appreciate where you're coming from! Happy to take more questions.
Understood. I do think possibly in simplifying it down the way you have it becomes a bit hard to explain what a certain number actually means - I don't really have any framework to evaluate "PE 15 in 2030" - but maybe that's just me.
On the subject of Nvidia, one topic which I would be interested to hear more on is CUDA, as it often comes up as the thing that's gonna ensure Nvidia can remain way ahead of all the other chipmakers. If you're planning a follow up to this one I think that would be a great thing to talk about.
Sure, points noted. 15x PE in 2030 simply means the forward PE in 2030, i.e. today's market cap / est. earnings in 2030 = 15x. It's useful for companies with nonsensical or negative trailing PEs, like NVDA's 100x today.
I think someone else actually wrote a pretty great article about CUDA recently! I can't seem to find it now, but maybe do a search for CUDA on Substack (and perhaps Google).
Excellent write-up!
Can't say I've ever seen a FY30 P/E : )