While I knew that one off the top of my head... one of the neat "Show HN" that I recall from a bit ago: Show HN: Find the relevant Xkcd comic for your post using RAG https://news.ycombinator.com/item?id=44799291
> How it works - Simply paste your entire message or post into the search box to get the most relevant xkcd for it. No need to search by keywords, etc.
Consider your own computer... how often does it get hot under a regular load and the fans kick on? That "fans kick on" is transferring the heat to air and jettisoning it into the room... and you're dealing with 100 watts there. Scale that up to kilowatts that are always running.
There is a lot of energy that is being consumed for computation and being converted into heat.
The other part if that is... its a lot easier to do that transfer heat into some other material and jettison it on earth, without having to ship the rack into space and also deal with the additional mechanics of getting rid of hot things. You've got advantages of things like "cold things sink in gravity" and "you can push heat around and sink it into other things (like phase change of water)" and "you don't need to be sitting on top of a power plant in order to use the power."
There's very little on JWST that makes it hot. Furthermore, the spacecraft bus where the non-scientific instruments exist) is on the other side of the sunshade so that the waste heat from the RAD750 (a single processor clocked at 118 MHz - PowerPC 750 architecture (Macintosh computers in '97) that uses 5 watts of power designed to operate -55°C to 125°C) doesn't interfere with the scientific instruments.
Putting an AI rack there is a completely different scale of power and cooling than what JWST uses.
Yes. Radiating heat away is a big problem - I totally agree! But stopping heat conduction is a solved problem! Consider the temperature of your pot handles. That‘s all I was saying.
What is the power budget for that DGX. The power budget for ISS is 75–90 kW. If your DGX fridge needs more power, it will need similar capacity of solar and radiators.
> In 2027, the unified approach will almost certainly be taken to new levels as NVIDIA's “Kyber” system will launch, with 576 GPUs in a single rack requiring a whopping 600 kW, equivalent to delivering enough power for 500 US homes into the space of a filing cabinet.
That fridge needs 600 kW of power. That will require 6x more solar panel space than the ISS and 6x more radiators. It's not about the volume of the object but rather its power and heat budgets.
If that fridge can reject all of the heat from the rack in space, it will work much better on earth and all the data centers would be using that fridge instead of their cooling towers and AC.
DGX is only 10,2 kW. But also only 8 H200s. Kyber seems a bit more power efficient (per-GPU) but requires much more power for a single unit. With that power requirement it doesn’t seem like it will fly.
10 kW is something that could be household load (a backup generator for a home when there's a power outage).
If one could put a DGX in my basement without issue (there's an idea - would you trust me with a DGX rack in my basement for six months for winter heating? https://www.cnbc.com/2026/01/27/data-centers-ai-district-hea... ), what is the value of shipping it into space? Granted, I couldn't afford a DGX in my basement... but it's not one DGX that you're putting in a datacenter, but rather racks upon racks of aisles upon aisles.
You're describing cryogenic fuels there and dumping heat into them. Dumping heat (sparks, electricity) into liquid oxygen would not necessarily be the best of ideas.
Dumping heat into liquid hydrogen wouldn't be explosive, but rather exacerbate the problem of boil off that is already one of the "this isn't going to work well" problems that needs to be solved for space fuel depots.
> Large upper-stage rocket engines generally use a cryogenic fuel like liquid hydrogen and liquid oxygen (LOX) as an oxidizer because of the large specific impulse possible, but must carefully consider a problem called "boil off", or the evaporation of the cryogenic propellant. The boil off from only a few days of delay may not allow sufficient fuel for higher orbit injection, potentially resulting in a mission abort.
They've already got the problem of that the fuel is boiled off in a matter of days. This is not a long term solution for a place to dump waste heat. Furthermore, it needs to be at cryogenic temperatures for it to be used by the spacecraft that the fuel depot is going to refuel.
> In a 2010 NASA study, an additional flight of an Ares V heavy launch vehicle was required to stage a US government Mars reference mission due to 70 tons of boiloff, assuming 0.1% boiloff/day for hydrolox propellant. The study identified the need to decrease the design boiloff rate by an order of magnitude or more.
0.1% boiloff/day is considered an order of magnitude to large now. That's not a place to shunt waste heat.
Compare the cost of a RAD750 (the processor on the JWST) to its non rad hardened variant. Additionally, consider the processing power of that system to modern AI demands.
I just calculated the potential weight of solar cells in space. Can't say about cost. Idea is mot of the weight of panel is because of glass/plastic protection on top and frame, these are there to protect from rain, hail, wind and dust. In space the elements it will need protection from will be different. I could be completely off but have no claims on cost and feasibility of this.
A solar panel deployed to space isn't deployed in its open / unframed configuration. Rather, it's sent in a way that is folded up into a compact volume and then unfolds into the full size.
You'll note that there is still a frame that it gets unfolded with and that you've got the additional mechanical apparatus to do the unfurling (and the human there to fix it if there are problems.
> Starting with the Enhanced variant, the solar panels were also upgraded to the UltraFlex, an accordion fanfold array, and the fuel load was increased to 1,218 kilograms (2,685 lb).
> In 2011, Orbital replaced Dutch Space on the project and gave ATK’s space components division, which was already supplying the substrates for Dutch Space’s Orion solar panels, a $20 million deal to provide UltraFlex arrays for later Cygnus flights.
I'd suggest giving https://www.gutenberg.org/ebooks/2130 a read.
reply