Hello everyone,
The JLL report has reached deep into the sci-fi box! Data centers in orbit are a fascinating vision, but I have to pull the plug on this timeline.
Here's a reality check on why we need to sort this topic in a completely different way:
1. the 2027 illusion (physics always wins)
The report is selling an absolute pipe dream here. Yes, transportation will be cheaper thanks to SpaceX and solar power in space will be virtually infinitely available. But they fail to mention the biggest physical problem: cooling. It may be freezing cold in space, but there is a vacuum. On Earth, we cool servers using air or water (convection). In a vacuum, however, there is no air to dissipate the heat. Currently, server farms in space would simply grill themselves because they can only radiate heat extremely inefficiently via huge radiators. Added to this is the unprotected cosmic radiation, which immediately destroys commercially available microchips without massive shielding (so-called bit flips).
My verdict: This is an exciting R&D playground for the next decade. But no one will be making any real commercial profit from it until 2027.
2. the investment case for big tech (AMZN, MSFT, GOOGL)
So how do we categorize these giants? Quite simply: as cash machines, they belong in the indestructible A-fundament of the portfolio.
When you buy Microsoft, Alphabet or Amazon, you are buying their gigantic cloud and AI monopoly. Whether these companies shoot a few server prototypes into space in five years' time as a feasibility study is not even a rounding error in their balance sheets, which weigh tons. These companies are bought because of their massive operating cash flows in the here and now - not because of space fantasies, which are mainly burning money at the moment.
My conclusion: Cool story for the distant future, but for real fundamental investment decisions over the next few years, this is just a castle in the air. Don't let reports like this dictate your portfolio!
Greetings, Raketentoni
The JLL report has reached deep into the sci-fi box! Data centers in orbit are a fascinating vision, but I have to pull the plug on this timeline.
Here's a reality check on why we need to sort this topic in a completely different way:
1. the 2027 illusion (physics always wins)
The report is selling an absolute pipe dream here. Yes, transportation will be cheaper thanks to SpaceX and solar power in space will be virtually infinitely available. But they fail to mention the biggest physical problem: cooling. It may be freezing cold in space, but there is a vacuum. On Earth, we cool servers using air or water (convection). In a vacuum, however, there is no air to dissipate the heat. Currently, server farms in space would simply grill themselves because they can only radiate heat extremely inefficiently via huge radiators. Added to this is the unprotected cosmic radiation, which immediately destroys commercially available microchips without massive shielding (so-called bit flips).
My verdict: This is an exciting R&D playground for the next decade. But no one will be making any real commercial profit from it until 2027.
2. the investment case for big tech (AMZN, MSFT, GOOGL)
So how do we categorize these giants? Quite simply: as cash machines, they belong in the indestructible A-fundament of the portfolio.
When you buy Microsoft, Alphabet or Amazon, you are buying their gigantic cloud and AI monopoly. Whether these companies shoot a few server prototypes into space in five years' time as a feasibility study is not even a rounding error in their balance sheets, which weigh tons. These companies are bought because of their massive operating cash flows in the here and now - not because of space fantasies, which are mainly burning money at the moment.
My conclusion: Cool story for the distant future, but for real fundamental investment decisions over the next few years, this is just a castle in the air. Don't let reports like this dictate your portfolio!
Greetings, Raketentoni
•
1515
•
4Mês
@Raketentoni I don't imagine cooling to be that difficult. Water cooling in a closed circuit with pumps and the ambient cold should be feasible with virtually zero energy costs, right?
The issue of radiation is a good point. I assume the chips would be protected by the shell of the data center in a similar way to people. But I'm not a physicist to be able to classify that. But the satellites and everything else up there also have chips in them, so that must already be solvable today?
I don't believe in 2027 either, but why not further in the future?
The issue of radiation is a good point. I assume the chips would be protected by the shell of the data center in a similar way to people. But I'm not a physicist to be able to classify that. But the satellites and everything else up there also have chips in them, so that must already be solvable today?
I don't believe in 2027 either, but why not further in the future?
•
22
•@Keineui
Great objections! It's precisely these technical discussions that make the forum here so strong. Your train of thought on cooling and radiation sounds absolutely logical from an earthly perspective, but the physics in orbit unfortunately puts a few massive stumbling blocks in our way.
Let's take a quick technical look at this:
**1. the cooling fallacy (the thermos flask principle)**
The big misconception is that space is not simply "cold" like a fridge, but a vacuum. And a vacuum insulates perfectly - just like the layer in a thermos flask! On Earth, water cooling only works because the radiator can release the heat at the end to the ambient air flowing past (convection).
There is no air in space. You can release the heat there *only* as infrared radiation. To radiate the gigantic waste heat from modern AI server farms into the void, you would need absurdly huge cooling panels weighing tons. Your closed water cooling system transports the heat away from the server, but you can't get it out of the space station efficiently afterwards. The box would simply cook itself.
**2. the radiation dilemma (Nvidia chips vs. Mars rover)**
You're absolutely right: satellites have chips on board. But these are extremely expensive, specially hardened processors ("radiation-hardened"). The catch: in order to make these chips resistant to the brutal cosmic radiation (so-called single event upsets, which swap ones and zeros), they usually lag decades behind terrestrial performance. A modern Mars rover operates with the computing power of a computer from the 1990s!
But if you want to operate data centers for AI models up there, you need ultra-modern, tiny 3nm chips (like those from Nvidia). If you wanted to shield an entire data center so thickly with shells that these earthly standard chips would survive, the thing would be so incredibly heavy that the rocket transport costs would immediately pulverize any business case.
**3. the distant future**
Sure, in 20 or 30 years there may be new material solutions. But we are investors here. We don't tie up our capital in 30-year fantasies and R&D projects when we can already buy companies down on earth today that are printing real money with tangible business models.
It remains an extremely exciting topic, but as an investment case it is still 100% science fiction for us!
Greetings, Raketentoni
Great objections! It's precisely these technical discussions that make the forum here so strong. Your train of thought on cooling and radiation sounds absolutely logical from an earthly perspective, but the physics in orbit unfortunately puts a few massive stumbling blocks in our way.
Let's take a quick technical look at this:
**1. the cooling fallacy (the thermos flask principle)**
The big misconception is that space is not simply "cold" like a fridge, but a vacuum. And a vacuum insulates perfectly - just like the layer in a thermos flask! On Earth, water cooling only works because the radiator can release the heat at the end to the ambient air flowing past (convection).
There is no air in space. You can release the heat there *only* as infrared radiation. To radiate the gigantic waste heat from modern AI server farms into the void, you would need absurdly huge cooling panels weighing tons. Your closed water cooling system transports the heat away from the server, but you can't get it out of the space station efficiently afterwards. The box would simply cook itself.
**2. the radiation dilemma (Nvidia chips vs. Mars rover)**
You're absolutely right: satellites have chips on board. But these are extremely expensive, specially hardened processors ("radiation-hardened"). The catch: in order to make these chips resistant to the brutal cosmic radiation (so-called single event upsets, which swap ones and zeros), they usually lag decades behind terrestrial performance. A modern Mars rover operates with the computing power of a computer from the 1990s!
But if you want to operate data centers for AI models up there, you need ultra-modern, tiny 3nm chips (like those from Nvidia). If you wanted to shield an entire data center so thickly with shells that these earthly standard chips would survive, the thing would be so incredibly heavy that the rocket transport costs would immediately pulverize any business case.
**3. the distant future**
Sure, in 20 or 30 years there may be new material solutions. But we are investors here. We don't tie up our capital in 30-year fantasies and R&D projects when we can already buy companies down on earth today that are printing real money with tangible business models.
It remains an extremely exciting topic, but as an investment case it is still 100% science fiction for us!
Greetings, Raketentoni
•
22
•
@Raketentoni Thank you for the clear and comprehensible summary.
•
22
•