Artificial Intelligence

[Longnose]

With the ventures that Musk pioneers. Is with almost all of them. I would be like that's absurd and likely impossible. Then... mad man Musk makes the impossible possible AND starts scaling it to affordability.  I would never bet against any of that mans ventures. 

[formthirteen]

1 hour ago, Longnose said:

With the ventures that Musk pioneers. Is with almost all of them. I would be like that's absurd and likely impossible. Then... mad man Musk makes the impossible possible AND starts scaling it to affordability.  I would never bet against any of that mans ventures. 

 

Don't forget Google:
https://services.google.com/fh/files/misc/suncatcher_paper.pdf

 

 

Can you take me higher, Rudolph:

 

[gfp]

Data centers in space sounds bonkers until you consider how much better solar energy is when almost all of it isn't being blocked by our atmosphere and it can be a 24/7 resource.  Cooling is also simpler up 'dere.

 

The rest of it I have no clue about but I heard there is a lot of energy up there in that big fireball that we are missing out on down here.

[treasurehunt]

5 hours ago, gfp said:

Data centers in space sounds bonkers until you consider how much better solar energy is when almost all of it isn't being blocked by our atmosphere and it can be a 24/7 resource.  Cooling is also simpler up 'dere.

 

The rest of it I have no clue about but I heard there is a lot of energy up there in that big fireball that we are missing out on down here.

 

Are you sure about cooling being simpler? Only radiative cooling is more effective in space than on earth and I have seen many reasonable-sounding arguments that cooling overall is actually more complicated in space. Like this post on X.

 

 

[gfp]

10 minutes ago, treasurehunt said:

Are you sure about cooling being simpler?

 

of course not I'm not sure about it! 

[treasurehunt]

2 minutes ago, gfp said:

 

of course not I'm not sure about it! 

 

Yes, since you are Global Financial Partners I figured space-based cooling is not your area of expertise, but I didn't want to make any assumptions.

[gfp]

We are strictly global down here in south Louisiana treasurehunt.  Nothing past that is the rule

[nsx5200]

I remember when Satellite phones like Globalstar and Iridium came out, and people just extrapolate the growth/coverage to imply that cellular phone was going to go the way of the Dodos.  Instead of going away, cellular phone coverage just kept building and increasing capacity.  Satellite phones remained a niche product.  Even now, the famed Starlink is still a niche product, when compared to traditional terrestrial products, despite all its success in displacing the incumbent in that niche.

 

The issue with listening to business executives, even CIOs, that have little to none technical background in how AI really works is that they repeat the same mistake of taking the current capability growth, and project it out to infinity.  In the case of AI, that is the end goal of super intelligent GAI.  They will even adopt the mainstream "tech" buzzwords like "first principal" without actually apply the concept.

 

Be wary of AI maximalists/minimalists that hand-wave a lot of the details and throws around a lot of buzz words.  Any good engineer knows that the devil is in the details, especially in tech.

[nsx5200]

 

 

It's interesting to see that the people that started "modern" AI are mostly not in the top right corner (AI maximalists).  At least not on it's current LLM path (Lecun).

[MungerWunger]

https://joincolossus.com/episode/nvidia-v-google-the-economics-of-ai/

[Libs]

Steve Eisman was just on CNBC, making the AI bear case, almost exactly as I see it. It was a little eerie.

 

In a nutshell, while he's long the MAG 7, he started reading Gary Marcus 9 months ago (the leading, lonely skeptic of the efficacy of LLM scaling). Lo and behold, Marcus is now being joined by other skeptics, many of them converts, essentially saying we can't reach the promised land of AGI without other tools to complement LLM scaling. Eisman, in so many words, says, "I'm still long, but increasingly wary. I'm looking for the day, if it comes, when one of the Mag 7 states they are cutting back on chip buying because the gains are leveling off."

 

That's when the dam will crack.

 

Eisman: 'Just as the GFC hinged on one idea- that national housing prices could not all fall together- so the AI bet hinges on LLM scaling indefinitely.'

 

Yes, I'm aware I'm chasing the high of calling a bubble like '08 again. But if it walks like a duck......

 

 

 

Edited December 11, 2025 by Libs

[Spekulatius]

On 12/10/2025 at 6:04 PM, treasurehunt said:

 

Are you sure about cooling being simpler? Only radiative cooling is more effective in space than on earth and I have seen many reasonable-sounding arguments that cooling overall is actually more complicated in space. Like this post on X.

 

 

Yes, cooling is more complex in space, since there is no convection cooling

and radiation cooling takes a lot of surface area. 

There is also the issue with launch cost (each kg of mass cost more than a thousand $ to launch even if Musk’s plans pan out) , destructive cosmic and solar radiation and the difficulty of repair and upgrades. 

Edited December 12, 2025 by Spekulatius

[sholland]

Datacenters in space are a terrible, horrible, no good idea.

TL/DR:  huge obstacles in power, thermal regulation, radiation tolerance, and communication.

https://taranis.ie/datacenters-in-space-are-a-terrible-horrible-no-good-idea/

 

[Milu]

13 hours ago, sholland said:

Datacenters in space are a terrible, horrible, no good idea.

TL/DR:  huge obstacles in power, thermal regulation, radiation tolerance, and communication.

https://taranis.ie/datacenters-in-space-are-a-terrible-horrible-no-good-idea/

 

I seem to remember how starting a private rocket company was a terrible idea, and starting an EV company was also a dumb idea.  The world is full of people, often well educated intelligent people with expertise in the specific area who loves nothing better than to expound on all the things that are impossible and how it will never work. Maybe this guy will be right, but if it's something SpaceX, Google, Nvidia are working on I'll put my faith in the smart leaders with demonstrated track record of success than this fellow.

 

Here is the conclusion of the article.

"I suppose this is just about possible if you really want to do it, but I think I've demonstrated above that it would firstly be extremely difficult to achieve, disproportionately costly in comparison with Earth-based datacenters, and offer mediocre performance at best."

 

Seems like pretty limited thinking if he's comparing the cost structure of already established earth based data centres which have had many years of cost improvements and optimisations with a brand new approach that doesn't exist yet. Obviously a space based approach will be much more expensive initially but like all new technological approaches costs will improve over time.

[Spekulatius]

@Milu you ignore physics. There is an easy way to do it and a hard way. Everything in space is hard and expensive and I don’t see a compelling reason to do this in the first place. Tell me one and I am ears.

 

Yes, I know heard that there is more sun radiation per unit area but did you know that you need to protect solar cells in space because or the hard radiation’s?How easy is it to install a solar cell ins pace versus on earth. What are launch cost to get a kg of any material in space versus some place on earth?

 

I think it makes more sense to put datacenters on the bottom of the ocean than in space , especially on the continental  shelf. Way cheaper and indeed cooling would not be an issue at all. Same with Antarctica.

Edited December 13, 2025 by Spekulatius

[UK]

https://www.bloomberg.com/features/2025-mustafa-suleyman-weekend-interview/?srnd=homepage-europe

[Libs]

I'm reading a VIC writeup on CoreWeave and the author has an interesting take.

 

1) Open AI's price cuts are undermining the competition;

 

https://techcrunch.com/2025/08/08/openai-priced-gpt-5-so-low-it-may-spark-a-price-war/#:~:text=The top-level GPT-5,popular for coding-related tasks.

 

This has repercussions for its competitors, like Anthropic. So if a price war is erupting for LLM's - 

 

2) it's evidence AI is moving to an ad model, similar to GOOG'S.

3) Hiring Kate Rouch from Meta, where she ran ads for 11 years, is further evidence.

 

This makes sense to me. That's a huge market, and the current pricing system is not working in terms of generating profit.

 

I hadn't read this before, what does the board think?

Edited December 13, 2025 by Libs

[UK]

43 minutes ago, Libs said:

I'm reading a VIC writeup on CoreWeave and the author has an interesting take.

 

1) Open AI's price cuts are undermining the competition;

 

https://techcrunch.com/2025/08/08/openai-priced-gpt-5-so-low-it-may-spark-a-price-war/#:~:text=The top-level GPT-5,popular for coding-related tasks.

 

This has repercussions for its competitors, like Anthropic. So if a price war is erupting for LLM's - 

 

2) it's evidence AI is moving to an ad model, similar to GOOG'S.

3) Hiring Kate Rouch from Meta, where she ran ads for 11 years, is further evidence.

 

This makes sense to me. That's a huge market, and the current pricing system is not working in terms of generating profit.

 

I hadn't read this before, what does the board think?

 

https://stratechery.com/2025/google-nvidia-and-openai/

 

 

[Spekulatius]

1 hour ago, Libs said:

I'm reading a VIC writeup on CoreWeave and the author has an interesting take.

 

1) Open AI's price cuts are undermining the competition;

 

https://techcrunch.com/2025/08/08/openai-priced-gpt-5-so-low-it-may-spark-a-price-war/#:~:text=The top-level GPT-5,popular for coding-related tasks.

 

This has repercussions for its competitors, like Anthropic. So if a price war is erupting for LLM's - 

 

2) it's evidence AI is moving to an ad model, similar to GOOG'S.

3) Hiring Kate Rouch from Meta, where she ran ads for 11 years, is further evidence.

 

This makes sense to me. That's a huge market, and the current pricing system is not working in terms of generating profit.

 

I hadn't read this before, what does the board think?

I think OpenAI spends $3 for every $ they take in currently, so how long can they do this? Google with Gemini can do this all day long and offer Gemini Pro in a package with others things for $20 and make a profit. How is OpenAI going to win with an inferior and expensive to build out tech stack?

https://one.google.com/about/#compare-plans

Edited December 13, 2025 by Spekulatius

[NnnnotSoSmart]

Gulp!

 

Edited December 14, 2025 by NnnnotSoSmart

[Spekulatius]

Pretty good comparison of the AI and the fiber optic boom in 1999/2000:

https://www.wsj.com/finance/stocks/the-eerie-parallels-between-ai-mania-and-the-dot-com-bubble-f99be6fe?mod=hp_lead_pos9

[Longnose]

On 12/14/2025 at 6:38 AM, Spekulatius said:

Pretty good comparison of the AI and the fiber optic boom in 1999/2000:

https://www.wsj.com/finance/stocks/the-eerie-parallels-between-ai-mania-and-the-dot-com-bubble-f99be6fe?mod=hp_lead_pos9

While there may be parallels.  I still strongly disagree that AI boom is comparable to the .com fiber optic boom. 

[Milu]

On 12/13/2025 at 12:46 AM, Spekulatius said:

@Milu you ignore physics. There is an easy way to do it and a hard way. Everything in space is hard and expensive and I don’t see a compelling reason to do this in the first place. Tell me one and I am ears.

 

Yes, I know heard that there is more sun radiation per unit area but did you know that you need to protect solar cells in space because or the hard radiation’s?How easy is it to install a solar cell ins pace versus on earth. What are launch cost to get a kg of any material in space versus some place on earth?

 

I think it makes more sense to put datacenters on the bottom of the ocean than in space , especially on the continental  shelf. Way cheaper and indeed cooling would not be an issue at all. Same with Antarctica.

Yes maybe, I'm not a physicist so can't speak with major authority on the topic, but I do know a lot of smart people seem to think it's a valid idea. I can somewhat understand why Musk would have a bias for space based data centres, but don't see what Sundar Pichai or Jensen Huang would be so positive on the approach. Wouldn't they rather go for the bottom of the ocean approach you outline if it is indeed cheaper and less complicated, I assume they are smart fellows who evaluate these type of things?

[Longnose]

33 minutes ago, Milu said:

Yes maybe, I'm not a physicist so can't speak with major authority on the topic, but I do know a lot of smart people seem to think it's a valid idea. I can somewhat understand why Musk would have a bias for space based data centres, but don't see what Sundar Pichai or Jensen Huang would be so positive on the approach. Wouldn't they rather go for the bottom of the ocean approach you outline if it is indeed cheaper and less complicated, I assume they are smart fellows who evaluate these type of things?

my immediate thought on bottom of ocean stuff is potential for ongoing maintenance.  The ocean is brutal with salt, animals, algae, and more...  I feel you'd have ongoing costs of something that the ocean will be constantly trying to erode/destroy. 

 

Send it to space... has lots of its own challenges but seems much more prone to up front costs that if you can navigate then it should be much more stable and no where near as much potential ongoing maint. 

 

I'm just a gomer so i could be wrong.  I don't want anything that lives on the bottom of the ocean. and I have personal love / infatuation with the ocean.

[Pelagic]

Nicely done piece on space based data centers that allows you to choose some of your own assumptions regarding cost with a good discussion of the engineering challenges involved.

https://andrewmccalip.com/space-datacenters