Artificial Intelligence

[Longnose]

11 minutes ago, Dalal.Holdings said:

 

One of the big reasons why Buffett bought BNSF is because "the rails are already put down" and don't need to be replaced. It's also much harder to build new railways because of Nimbyism/property ownership/etc.

 

These things are not true for your AI:railroad analogy: the rails need to be replaced frequently ($$$ recurring capex), and it's not harder for a new entrant to come in and buy the latest NVDA chips (or maybe not: Deepseek) and replicate the LLMs and achieve much of the same things

 

Also, as I noted, it's like a bunch of railroad tracks that are duplicative and run the same routes next to each other...

 

Finally, Railroads were often terrible investments in the 1800s and much of the 20th century as well--despite railroads being game changers that advanced the world significantly...there is a parallel with airlines in that respect too

 

 

I get your points and agree with some and disagree with others. I think some of the biggest points tho is the "rails" might wear out, but Google and Microsoft have already bought all the best parts for years to come. Even if some new company had a ton of money, they couldn't even get the chips they'd need to compete. Plus, it's not just the hardware. These companies have been collecting and organizing data for decades, and that's a serious moat too. You can't just get that overnight. 

[Dalal.Holdings]

 

 

 

 

 

 

[Pelagic]

A fun read

https://www.astralcodexten.com/p/what-is-man-that-thou-art-mindful

[WayWardCloud]

44 minutes ago, Pelagic said:

A fun read

https://www.astralcodexten.com/p/what-is-man-that-thou-art-mindful

That was awesome 

[Zemergefen]

11 hours ago, Pelagic said:

A fun read

https://www.astralcodexten.com/p/what-is-man-that-thou-art-mindful

Great read!

[WayWardCloud]

I've been reading some versions of "it's just like the late 90's all over again" and "It's just like 2007 all over again" with a bunch of very alarming charts ever since I started investing 10 years ago. Listening to these experts who have kept predicting the last crisis to happen again and again is such a costly hobby that affects most investors, but especially the "value" ones. Because we looove the idea of being smarter than others, we love it more than making money. One day something bad will happen and the experts will claim they were right. In the meantime how much wealth has been left on the table out of fear? Just find a few things you'd like to own at prices you think are sensible and maintain a tactical allocation that you can sleep well at night with regardless of market conditions, then let go of overthinking and do something else more interesting with your life than doom-scrolling twitter for macroeconomic tea-leaves readings.

Edited September 18, 2025 by WayWardCloud

[Dalal.Holdings]

I hope ya’ll are paying subscription fees for your LLMs…

 

 

 

[Spekulatius]

I added all the stocks that are impacted by AI in the top 20. it’s currently 39.42% of the market cap. Almost 40% of the SP500 is and ai trade of some sorts. I haven’t done the QQQ yet. To be fair, not all of those business are just AI but they do trade on this single these now. we are now exceeding the internet and telecom bubble from 1998-2000in terms of thematic stock contribution as well as investment as a percentage of GDP by quite a bit.

 

[mattee2264]

There is an awful lot of circularity which adds to the risks.

 

Cloud companies are getting a big boost in revenues from the AI start ups (with as yet massively unprofitable business models) and a lot of it is investments in exchange for cloud credits. 

 

Similarly you are getting investments in exchange for purchase contracts/commitments. You don't have to expense the investment but you get to book the revenue. 

 

And NVIDIA gets most of its revenues from a handful of big Tech companies spending most of their operating cashflows on AI capex not because they have any great confidence in the returns on those investments but because they can afford to do so and feel the risk of underinvestment is far greater i.e. the investment is defensive to a large extent.

 

And the circularity extends to the economy and the stock market. Investment spending adds to GDP both directly but also via the wealth effects and increased consumption. 

 

So if everyone were to cut back considerably on investment and a lot of this projected revenue doesn't materialize and growth slows or even declines then all of this goes into reverse and even rate cuts and extra government spending might not be enough to prevent a very nasty bear market and recession. 

[Eldad]

3 hours ago, mattee2264 said:

There is an awful lot of circularity which adds to the risks.

 

Cloud companies are getting a big boost in revenues from the AI start ups (with as yet massively unprofitable business models) and a lot of it is investments in exchange for cloud credits. 

 

Similarly you are getting investments in exchange for purchase contracts/commitments. You don't have to expense the investment but you get to book the revenue. 

 

And NVIDIA gets most of its revenues from a handful of big Tech companies spending most of their operating cashflows on AI capex not because they have any great confidence in the returns on those investments but because they can afford to do so and feel the risk of underinvestment is far greater i.e. the investment is defensive to a large extent.

 

And the circularity extends to the economy and the stock market. Investment spending adds to GDP both directly but also via the wealth effects and increased consumption. 

 

So if everyone were to cut back considerably on investment and a lot of this projected revenue doesn't materialize and growth slows or even declines then all of this goes into reverse and even rate cuts and extra government spending might not be enough to prevent a very nasty bear market and recession. 

Had to laugh today when NVDA invested 100 MM in Open AI. 

[Spekulatius]

On 9/22/2025 at 7:54 PM, Eldad said:

Had to laugh today when NVDA invested 100 MM in Open AI. 

$100B. $100Mm buys you an AI guy who knows how to write code.

https://www.reuters.com/business/nvidia-invest-100-billion-openai-2025-09-22/

 

So $NVDA invests $100B in OpenAI, they buy $100B in chips from NVDA. Thats worth 10x revenues. The stake in OpenAI is worth another $500B next year. So $NVDA just created $1.4T in value. Thats how this works. 

Edited September 24, 2025 by Spekulatius

[Eldad]

27 minutes ago, Spekulatius said:

$100B. $100Mm buys you an AI guy who knows how to wrote code.

https://www.reuters.com/business/nvidia-invest-100-billion-openai-2025-09-22/

 

So $NVDA invests $100B in OpenAI, they buy $100B in chips from NVDA. Thats worth 10x revenues. The stake in OpenAI is worth another $500B next year. So $NVDA just created $1.4T in value. Thats how this works. 

Ha yes, my bad 

[Buckeye]

Also interesting is that OpenAI represents 300B of the 317B in “sales” backlog that Oracle just added in their most recent quarter, which subsequently caused its stock to rise ~40% (Open AI’s largest fund raising to date is 40B)

[Peregrine]

I must say - this market is utterly insane. No one has even figured out how to make any money off of this stuff and yet we're talking about trillions in spend. This coming from a country that can't even fix a highway.

[WFF]

With all the investment in AI, it really brings back memories of the 90s and this article.  
 

Warren 1999

 

Few key points: 

1. There were over 2000 auto companies and there are only 3 that survived.  
 

2. Even those that survived, it doesn’t mean that the economics are great. 
 

Perhaps we should invert and instead of who will win and who has the better LLM etc we should think of who will lose. 
 

This stuck from the article:

 

 Sometimes, incidentally, it’s much easier in these transforming events to figure out the losers. You could have grasped the importance of the auto when it came along but still found it hard to pick companies that would make you money. But there was one obvious decision you could have made back then–it’s better sometimes to turn these things upside down–and that was to short horses. Frankly, I’m disappointed that the Buffett family was not short horses through this entire period. And we really had no excuse: Living in Nebraska, we would have found it super-easy to borrow horses and avoid a “short squeeze.”

 

[thowed]

5 hours ago, Peregrine said:

a country that can't even fix a highway.

 

Yep - not the only guilty country - but it does drive me insane that foundational services (roads, education, healthcare) are being ignored over future speculation.

[Spekulatius]

10 minutes ago, thowed said:

 

Yep - not the only guilty country - but it does drive me insane that foundational services (roads, education, healthcare) are being ignored over future speculation.

Every generation gets to play a bubble. Mine was the telecom bubble from 1997-2000. I think this one is bigger in term off investment in this sector/ GDP and also our government is all in.

[dealraker]

8 minutes ago, Spekulatius said:

Every generation gets to play a bubble. Mine was the telecom bubble from 1997-2000. I think this one is bigger in term off investment in this sector/ GDP and also our government is all in.

[dealraker]

Just now, dealraker said:

14 minutes ago, Spekulatius said:

Every generation gets to play a bubble. Mine was the telecom bubble from 1997-2000. I think this one is bigger in term off investment in this sector/ GDP and also our government is all in.

Yep.

[CharlesMunger]

1 hour ago, Spekulatius said:

Every generation gets to play a bubble. Mine was the telecom bubble from 1997-2000. I think this one is bigger in term off investment in this sector/ GDP and also our government is all in.

 So how are you playing this one Spek?

[Marco Van Basten]

3 hours ago, thowed said:

 

Yep - not the only guilty country - but it does drive me insane that foundational services (roads, education, healthcare) are being ignored over future speculation.

There is plenty of money being spent on education, it is just being spent poorly.  NYC spends $44K per kid in public school per year and gets horrible outcomes.  Too many administrators, no accountability, declining standards, getting rid of things that work in the name of "those who lag behind should not feel bad", et cetera.  

[Spekulatius]

8 hours ago, CharlesMunger said:

 So how are you playing this one Spek?

I think the best play to buy what the current crowd thinks gets disrupted. Sectors like software, consulting, financial data.

Edited September 24, 2025 by Spekulatius

[Saluki]

20 hours ago, Spekulatius said:

$100B. $100Mm buys you an AI guy who knows how to write code.

https://www.reuters.com/business/nvidia-invest-100-billion-openai-2025-09-22/

 

So $NVDA invests $100B in OpenAI, they buy $100B in chips from NVDA. Thats worth 10x revenues. The stake in OpenAI is worth another $500B next year. So $NVDA just created $1.4T in value. Thats how this works. 

This reminds me of the tech book where a company who made routers or servers would "invest" in a startup by giving it product instead of cash. 

 

The potential problem is that rather than NVDA selling those chips for cash, they got shares. And for those shares to be worth more than cash, down the road, they need to be profitable. Otherwise what is the difference between the that and donating it to a university?

[Spekulatius]

42 minutes ago, Saluki said:

This reminds me of the tech book where a company who made routers or servers would "invest" in a startup by giving it product instead of cash. 

 

The potential problem is that rather than NVDA selling those chips for cash, they got shares. And for those shares to be worth more than cash, down the road, they need to be profitable. Otherwise what is the difference between the that and donating it to a university?

Yes, these circular relationships are nothing new. Lucent gave generous finance terms customers who probably could not afford the equipment otherwise.

[Peregrine]

17 hours ago, Spekulatius said:

Every generation gets to play a bubble. Mine was the telecom bubble from 1997-2000. I think this one is bigger in term off investment in this sector/ GDP and also our government is all in.

 

At least fiber optic lines lasts a long time and has a very obvious benefit that people clearly wanted (high speed internet). Over half of AI CAPEX is going into NVIDIA GPUs that are only useful for 3 years and on top of that no one even knows what it's good for and if anyone will pay for it.

Edited September 25, 2025 by Peregrine