Artificial Intelligence

[treasurehunt]

1 hour ago, Milu said:

Maybe we are confusing each other here, maybe I'm confusing myself .

 

Is your opinion that the market is currently pricing in epic returns from the massive cap ex build out. If you look at the valuation multiples of Alphabet or Meta do they look elevated in anyway against their average multiples over the last 5 years, or last decade. If the market was pricing them in a bubbly way I'd have expected significantly higher multiples to earnings, cash flow etc, but I don't see anything different to valuations before this AI capex ramped up. I can't speak to Nvidia or Microsoft as I don't own those stocks but maybe they are bubblier.

 

I'm not sure whether we are agreeing with each other or not.

 

I don't think we are in agreement.

 

I'm not saying the market is pricing in epic returns from capex; just that it is not pricing in low returns. There is a lot of room between low and epic.

 

Also, as Eldad pointed out with excellent examples, P/E is just one measure of valuation. I'm sure you know that P/E can be misleading for various reasons.

 

Finally, I would generally expect valuation measures based on sales or earnings (P/E, P to FCF, P/S etc) to decrease over time for growth companies, all else equal. For example, Google is almost five times as big today by revenue as it was 10 years ago. It will be tough for future growth to match past growth unless something big and new is on the horizon. The market seems to think AI is such a thing.

[Milu]

2 hours ago, Eldad said:

PEs are maybe similar but price to FCF is massively higher and every year going forward you get a bigger and bigger depreciation hit from the servers. The Capex is effectively lowering future earnings while the profit side is still in doubt. 

Yup aware of the free cash flow, but again if the capex can be deployed at high rates of return then the optimum strategy would be to invest all excess cash and run a free cash flow of zero. It all depend on how much of today's current capex you view as growth vs maintenance capex and whether the growth capex generates the incremental ROIC or not. And as you said the earnings each year will be factoring in more depreciation so wouldn't the earnings continue to be a good yardstick to look at for multiples?

[Eldad]

11 minutes ago, Milu said:

Yup aware of the free cash flow, but again if the capex can be deployed at high rates of return then the optimum strategy would be to invest all excess cash and run a free cash flow of zero. It all depend on how much of today's current capex you view as growth vs maintenance capex and whether the growth capex generates the incremental ROIC or not. And as you said the earnings each year will be factoring in more depreciation so wouldn't the earnings continue to be a good yardstick to look at for multiples?

I’m just not convinced it will materialize into a good investment. These companies are so good that a lot mistakes and inefficiencies were covered over in the past. I am especially concerned as a META shareholder. When Zuck believed we would all be buying digital real estate by now, the capex was a lot smaller. 

[Milu]

14 minutes ago, Eldad said:

I’m just not convinced it will materialize into a good investment. These companies are so good that a lot mistakes and inefficiencies were covered over in the past. I am especially concerned as a META shareholder. When Zuck believed we would all be buying digital real estate by now, the capex was a lot smaller. 

Ya you could certainly be right, nobody knows how these things play out. I think some of Zuck's logic is that the payoff is so large that it's worth making the bet, and if it doesn't work out a lot of the data centre infrastructure can be used anyway to improve it's ad engine and as compute for other areas of the business. I have faith that he is a good operator and is not just flinging money into a pit without assessing returns and downside. Doesn't mean he's always going to be right.

[Dalal.Holdings]

5 hours ago, FiveSigma said:

 

Pay attention.

 

Words of caution are for you, when you say and I quote "with likely lower future returns on capital".

 

When a business can lower a key input costs by 5x in a single year (source below) and another key one (power) by 33x, it should give a prudent person pause when assuming ROIC declines.

 

You move the goal posts again, now claiming the argument is about whether AI being "the next big thing" is consensus.

 

Also, nowhere did I state that "sky is the limit".

 

I'm saying it's a very difficult question to throw uneducated takes on.

 

Source:

https://www.seroundtable.com/google-sge-ai-answers-cost-80-less-37326.html

 

The fact that it has become much cheaper to do these things is not the positive development you think it is. It means new entrants can easily duplicate these efforts for a fraction of the costs/resources (as deepseek showed) and the moats prove illusory.

 

It also means all the high capex being deployed now will be unlikely to generate future returns to justify the expenditures. It means it pays to wait on AI and let someone else spend the money on it now (a strategy that Apple is pursuing).

 

[Dalal.Holdings]

I like simplicity above all. The law of large numbers (of investing, not probability and statistics) tells me that $50-100B in annual capex that grows every year is likely to result in returns going south.

 

Many of these firms are in the mentality that they have to run the race with rival firms (and they think it will be winner-take-all) and it’s leading to capital incineration in my view.

 

Apple is wise in sitting it out.

 

Meanwhile, the consensus view (via everyone’s 401k going into the S&P and disproportionately these stocks) is that these firms will win big from these big investments.

[Eldad]

14 minutes ago, Milu said:

Ya you could certainly be right, nobody knows how these things play out. I think some of Zuck's logic is that the payoff is so large that it's worth making the bet, and if it doesn't work out a lot of the data centre infrastructure can be used anyway to improve it's ad engine and as compute for other areas of the business. I have faith that he is a good operator and is not just flinging money into a pit without assessing returns and downside. Doesn't mean he's always going to be right.

Agree there. These guys are visionaries and have had mostly good judgement as far as what is coming next. They are totally immersed in that world and I am just a guy in the stands yelling at the quarterback. 

 

I think one of the big blind spots of these guys is sort of their love / belief in / reverence for science fiction. When I read some of their statements on the near future. When I read that the Ex Uber CEO thinks that he had Gronk on the verge of a major physics breakthrough when he was “talking” to “him” the other night. When I read that an MIT CS student dropped out of school because the world will be over in 5 years due to AGI and it was a waste of time. And on and on and on complete BS. My horse sense says this is one of the most massive delusions I have ever witnessed. 
 

We will see. 
 

 

[vinod1]

In 2011 or 2012, or somewhere around that time, the law of large numbers was thrown out for exactly these companies on this forum. "These are already hundreds of billon dollar companies, do you really think they would go to trillion?" And here we are.

 

The last 15 years we have witnessed companies growing from already large bases at extraordinary rates. In ways most of us thought is just not possible. If the last 15 years were such a surprise, there is an even chance that it might be true going forward. 

 

Vinod

 

 

 

[Eldad]

2 minutes ago, vinod1 said:

In 2011 or 2012, or somewhere around that time, the law of large numbers was thrown out for exactly these companies on this forum. "These are already hundreds of billon dollar companies, do you really think they would go to trillion?" And here we are.

 

The last 15 years we have witnessed companies growing from already large bases at extraordinary rates. In ways most of us thought is just not possible. If the last 15 years were such a surprise, there is an even chance that it might be true going forward. 

 

Vinod

 

 

 

Now they are 30+% of the S&P

 

US GDP in 2024 was 29T. Let’s say if the US was a stock it would be worth 350T give or take (29T x 12). So the Market Cap of NVDA, MSFT, AAPL, AMZN, META, GOOGL is 5% of the entire US economy. 
 

Do you think that is sustainable? Populism rises when there are monopolies and crony capitalism. 1890s and today. If they were to accomplish their wildest AI dreams it would just assure their own demise politically. 

[Milu]

23 minutes ago, vinod1 said:

In 2011 or 2012, or somewhere around that time, the law of large numbers was thrown out for exactly these companies on this forum. "These are already hundreds of billon dollar companies, do you really think they would go to trillion?" And here we are.

 

The last 15 years we have witnessed companies growing from already large bases at extraordinary rates. In ways most of us thought is just not possible. If the last 15 years were such a surprise, there is an even chance that it might be true going forward. 

 

Vinod

 

 

 

Yup and my thinking is that people are making the same mistake again. The whole world is continuously moving into the digital space, before these companies were going after the consumer and advertising markets, now with the advent of AI and Robotics this is opening up the whole labour market to be disrupted. If you think about all the jobs out there going from entry level things like a barista, to factory workers, to junior lawyers, to accountants, all the way up. Over the coming decades a decent proportion of these could likely be replaced/augmented. It may happen faster or slower than people think but the direction of travel is only going one way.

[Eldad]

2 minutes ago, Milu said:

Yup and my thinking is that people are making the same mistake again. The whole world is continuously moving into the digital space, before these companies were going after the consumer and advertising markets, now with the advent of AI and Robotics this is opening up the whole labour market to be disrupted. If you think about all the jobs out there going from entry level things like a barista, to factory workers, to junior lawyers, to accountants, all the way up. Over the coming decades a decent proportion of these could likely be replaced/augmented. It may happen faster or slower than people think but the direction of travel is only going one way.

Seems right but probably what the Romans used to say. Then came 1000 years of dark ages. You never know.  
 

I will predict the next great Democrat will be a trust buster. Gavin isn’t the one. 

[Dalal.Holdings]

46 minutes ago, vinod1 said:

In 2011 or 2012, or somewhere around that time, the law of large numbers was thrown out for exactly these companies on this forum. "These are already hundreds of billon dollar companies, do you really think they would go to trillion?" And here we are.

 

The last 15 years we have witnessed companies growing from already large bases at extraordinary rates. In ways most of us thought is just not possible. If the last 15 years were such a surprise, there is an even chance that it might be true going forward. 

 

Vinod

 

 

 

 

So how big will they get ? 10 Trillion? 100 Trillion?

 

Saying there is an even chance is not correct. For example, as BRK has grown bigger, the chances that returns go down increases. The bigger, the lower the chances of high returns.

 

As you start approaching sizes of major GDPs, the ceiling gets closer and closer

 

 

Edited September 4, 2025 by Dalal.Holdings

[Dalal.Holdings]

52 minutes ago, Eldad said:

Agree there. These guys are visionaries and have had mostly good judgement as far as what is coming next. They are totally immersed in that world and I am just a guy in the stands yelling at the quarterback. 

 

I think one of the big blind spots of these guys is sort of their love / belief in / reverence for science fiction. When I read some of their statements on the near future. When I read that the Ex Uber CEO thinks that he had Gronk on the verge of a major physics breakthrough when he was “talking” to “him” the other night. When I read that an MIT CS student dropped out of school because the world will be over in 5 years due to AGI and it was a waste of time. And on and on and on complete BS. My horse sense says this is one of the most massive delusions I have ever witnessed. 
 

We will see. 
 

 

 

We've already seen them fail spectacularly in other areas and spend lots of money doing so: Metaverse being a big example for Zuck. He changed the name of the company if that gives you an idea of how confident he was in metaverse. And even Apple tried following him to that category and failed.

 

You can also include some of these guys getting sucked into NFTs, etc

[Eldad]

19 minutes ago, Dalal.Holdings said:

 

We've already seen them fail spectacularly in other areas and spend lots of money doing so: Metaverse being a big example for Zuck. He changed the name of the company if that gives you an idea of how confident he was in metaverse. And even Apple tried following him to that category and failed.

 

You can also include some of these guys getting sucked into NFTs, etc

Yeah but the bet was small and Wall St immediately pushed back. They are now making a huge bet and Wall St. is supporting it for now. 

[Milu]

27 minutes ago, Eldad said:

Seems right but probably what the Romans used to say. Then came 1000 years of dark ages. You never know.  
 

I will predict the next great Democrat will be a trust buster. Gavin isn’t the one. 

Yes wise words, I do try to caution myself with any of my opinions/predictions. Anybody who has read any amount of human history has be aware that we are forever humbled and there are never any guarantees. Hence what keeps this game of investing so interesting.

[Dalal.Holdings]

36 minutes ago, Eldad said:

Yeah but the bet was small and Wall St immediately pushed back. They are now making a huge bet and Wall St. is supporting it for now. 

 

The bet was small because Metaverse and the rest of tech does not require the capex AI requires.

 

That fundamental nature of AI is what makes it so likely to blow up spectacularly in terms of firmwide returns

[Spekulatius]

On 9/2/2025 at 11:39 PM, LC said:

 

They're probably betting it will play out like search. In the early days of search, results were not really different between the various search engines. Eventually, one search engine took all the economics. 

This is not correct. there were a bunch of different search engines like Lycos, Yahoo and Altavista and the result were very differnt, more so than Bing and Google. Altavista was the Google at that time and far better than anything else until Google came around.

[Spekulatius]

3 hours ago, Dalal.Holdings said:

 

The bet was small because Metaverse and the rest of tech does not require the capex AI requires.

 

That fundamental nature of AI is what makes it so likely to blow up spectacularly in terms of firmwide returns

The Metaverse bet wasn’t small. He put tens of billions of $ in it and thousands of people were working on it. It was a Manhattan size project for a company like Meta.

[Dalal.Holdings]

8 minutes ago, Spekulatius said:

The Metaverse bet wasn’t small. He put tens of billions of $ in it and thousands of people were working on it. It was a Manhattan size project for a company like Meta.

 

Relative to the capex needed to build an AI model, it is quite small. In 2021 (the year Facebook changed its name to META, annual capex was $18B. The past twelve months it’s $52B.

 

AI requires more capex because these firms are doing it for themselves and the chips and data centers require vast amounts of capital. The energy requirements and talent to do it is also expensive which crushes margins.

 

Compare that to Meta’s Oculus or Apple’s iPhone: these tech companies design the product and outsource the capital intensive manufacturing to third party manufacturers like Foxconn. So Apple’s and Meta’s ROIC remain high as they are merely designers while Foxconn is doing the capital intensive work.

 

In effect, AI is making these businesses worse…

 

Edited September 5, 2025 by Dalal.Holdings

[Eldad]

I play Wordle most days. Today I got it in 3 tries which is rare for me. 
 

I put my first two guesses with the colors for each letter into Chat GPT and Gemini. 
 

They both got the same wrong answer with the same obvious mistake. Today’s answer is DRIFT. I told both of the systems that “I” was green in the third position. Yet they both answered with DIRTY. Very odd. Not sure this is taking over the world anytime soon. 

Edited September 5, 2025 by Eldad

[vinod1]

On 9/4/2025 at 3:16 PM, Eldad said:

Now they are 30+% of the S&P

 

US GDP in 2024 was 29T. Let’s say if the US was a stock it would be worth 350T give or take (29T x 12). So the Market Cap of NVDA, MSFT, AAPL, AMZN, META, GOOGL is 5% of the entire US economy. 
 

Do you think that is sustainable? Populism rises when there are monopolies and crony capitalism. 1890s and today. If they were to accomplish their wildest AI dreams it would just assure their own demise politically. 

 

I am wary of using macroeconomic variables to infect the valuation of individual companies or make proclamations of bubbles for several reasons:

 

1. It never worked in the past. When it did seem to work, it is mostly because some other event outside of that macroeconomic variable impacted the market. Dot com bubble did not pop because Market Cap to GDP exceeded some percentage, but corporate scandals at Enron, Worldcom, 9/11, Anderson, etc. shed light on the bubbly nature of many dot com businesses and investors turned skeptical. Same with profit margins - they are supposed to be at 6.5% or "else capitalism is broken". Dividend yield of stocks is supposed to always be higher than bonds, etc. 

 

2. We still do not yet have a good understanding of macroeconomic variables. Something as common as GDP calculation itself is highly subjective. If you want to take a look at GDP calculations and the discussions around what to include, read "GDP: A Brief but Affectionate History". We might be underestimating GDP for a whole lot of very benign reasons (no conspiracy theory). Google search had a dramatic impact that improved the welfare of a lot of people, same with Internet, mobile, etc. None of this is captured in GDP. Dont want to digress too much on this topic, but at some point we might end up revising what goes into GDP calculations.

 

So why mix what you know with a high degree of certanity - current Mag 7 profits with something we know vaguely and imprecisely - GDP calcuations?

 

 My point being, don't use these percent of GDP numbers to influence your investment decisions. 

 

Vinod

[vinod1]

On 9/4/2025 at 3:42 PM, Dalal.Holdings said:

 

So how big will they get ? 10 Trillion? 100 Trillion?

 

Saying there is an even chance is not correct. For example, as BRK has grown bigger, the chances that returns go down increases. The bigger, the lower the chances of high returns.

 

As you start approaching sizes of major GDPs, the ceiling gets closer and closer

 

 

 

Hi Dalal,

 

Please see my response to Eldad. 

 

Vinod

[WayWardCloud]

The chatbots you are using are very generic tools and not at all optimized for wordle solving. If either Alphabet or OpenAI decided to they could in just a few days slap together a specialized wordle agent using their underlying LLM which would be the best wordle player in the world ever. Remember the history of AI development has been to apply it to games so we know that works very well. It's not that the models aren't good enough it's actually the other way : the challenge of getting wordle right is so trivial none of the developers are even bothering with it.

[Eldad]

1 minute ago, WayWardCloud said:

The chatbots you are using are very generic tools and not at all optimized for wordle solving. If either Alphabet or OpenAI decided to they could in just a few days slap together a specialized wordle agent using their underlying LLM which would be the best wordle player in the world ever. Remember the history of AI development has been to apply it to games so we know that works very well. It's not that the models aren't good enough it's actually the other way : the challenge of getting wordle right is so trivial none of the developers are even bothering with it.

I understand that but they keep saying it is moving towards AGI or that it somehow models a human brain and has real intelligence. Completely not the case if it can’t figure out that the third letter I have given it is correct and then moves it to the 2nd letter.
 

My point is that it is still, and probably always will be, just a giant statistical calculator that badly needs human guidance. 

[Eldad]

4 minutes ago, Eldad said:

I understand that but they keep saying it is moving towards AGI or that it somehow models a human brain and has real intelligence. Completely not the case if it can’t figure out that the third letter I have given it is correct and then moves it to the 2nd letter.
 

My point is that it is still, and probably always will be, just a giant statistical calculator that badly needs human guidance. 

Basically any time I ask a super specific, esoteric question it makes up a complete lie. I wouldn’t be surprised if there is an army of humans feeding it boiler plate answers for the most common questions.

 

I mean when it first came out I remember a Google exec saying yes we have that as well but we didn’t release it because it lies constantly. When Open AI let the cat out of the bag, they were forced to play the game.