This throughput assumes 100% utilizations. A bunch of things raise the cost at scale:
- There are no on-demand GPUs at this scale. You have to rent them for multi-year contracts. So you have to lock in some number of GPUs for your maximum throughput (or some sufficiently high percentile), not your average throughput. Your peak throughput at west coast business hours is probably 2-3x higher than the throughput at tail hours (east coast morning, west coast evenings)
- GPUs are often regionally locked due to data processing issues + latency issues. Thus, it's difficult to utilize these GPUs overnight because Asia doesn't want their data sent to the US and the US doesn't want their data sent to Asia.
These two factors mean that GPU utilization comes in at 10-20%. Now, if you're a massive company that spends a lot of money on training new models, you could conceivably slot in RL inference or model training to happen in these off-peak hours, maximizing utilization.
But for those companies purely specializing in inference, I would _not_ assume that these 90% margins are real. I would guess that even when it seems "10x cheaper", you're only seeing margins of 50%.
jerrygenser · 19m ago
Re the overnight that's why some providers are offering there are batch tier jobs that are 50% off which return over up to 12 or 24 hours for non-interactive use cases.
caminanteblanco · 3h ago
There was some tangentially related discussion in this post: https://news.ycombinator.com/item?id=45050415, but this cost analysis answers so many questions, and gives me a better idea of how huge the margin on inference a lot of these providers could be taking. Plus I'm sure that Google or OpenAI can get more favorable data center rates than the average Joe Scmoe.
A node of 8 H100s will run you $31.40/hr on AWS, so for all 96 you're looking at $376.80/hr. With 188 million input tokens/hr and 80 million output tokens/hr, that comes out to around $2/million input tokens, and $4.70/million output tokens.
This is actually a lot more than Deepseek r1's rates of $0.10-$0.60/million input and $2/million output, but I'm sure major providers are not paying AWS p5 on-demand pricing.
Edit: those figures were per node, so the actual input and output prices would be divided by 12.$0.17/million input tokens, and $0.39/million output
bluedino · 20s ago
> A node of 8 H100s will run you $31.40/hr on AWS, so for all 96 you're looking at $376.80/hr
And what stinks is that you can't even build a Dell/HPE server like this online. You have to 'request a quote' for an 'AI Server'
Going through SuperMicro, you're looking at about $60k for the server, plus 8 GPU's at $25,000 each, so you're close to $300,000 for an 8 GPU node.
Now, that doesn't include networking, storage, racks, electricity, cooling, someone to set that all up for you, $1,000 DAC cables, NVIDIA middleware, downtime as the H100's are the flakiest pieces of junk ever and will need to be replaced every so often...
Setting up a 96 H100 cluster (12 of those puppies) in this case is probably going to cost you $4-5 million.
zipy124 · 2h ago
AWS is absolutely not cheap, and never has been. You want to look for the hetzner of the GPU world like runpod.io where they are $2 an hour, so $16/hr for 8, that's already half of aws. You can also get a volume discount if you're looking for 96 almost certainly.
An H100 costs about $32k, amortized over 3-5 years gives $1.21 to $0.7 per hour, so adding in electricity costs and cpu/ram etc... runpod.io is running much closer to the actual cost compared to AWS.
caminanteblanco · 2h ago
Ok, so the authors apparently used atlas cloud hosting, which charges $1.80 per h100/hr, which would change the overall cost to around $0.08/ million input and $0.18/million output, which seems much more in line with massive inference margins for major providers.
matt-p · 2h ago
188M input / 80M output tokens per hour was per node I thought?
Reversing out these numbers tells us that they're paying about $2/H100/Hour (or $16/hour for a 8xH100 node).
Disclaimer (one of my sites) https://www.serversearcher.com/servers/gpu - says that a one month commit on a 8XH100 node goes for $12.91/hour. The "I'm buying the servers and putting them in COLO rate" usually works out at around $10/Hour, so there's scope here to reduce the cost by ~30% just by doing better/more committed purchasing.
caminanteblanco · 2h ago
You were definitely right, I updated the original comment. Thanks for your correction!
paxys · 2h ago
According to the post their costs were $0.20/1M output tokens (on cloud GPUs), so your numbers are off somewhere.
"By deploying this implementation locally, it translates to a cost of $0.20/1M output tokens"
Is that just the cost of electricity, or does it include the cost of the GPUs spread out over their predicted lifetime?
zipy124 · 2h ago
This is all costs included. Thats 22k tokens per second per node, so per 8 h100's. With 12 nodes they get 264k tokens per second, or 950 million an hour. This get's you to roughly $0.2021 per million at $2 an hour for an h100, which is what they go for on services such as runpod.io . (cheaper if not paying spot-price + volume discounts).
dragonslayer56 · 3h ago
” Our implementation, shown in the figure above, runs on 12 nodes in the Atlas Cloud, each equipped with 8 H100 GPUs.”
Maybe the cost of renting?
34679 · 3h ago
I'm confused because I wouldn't consider a cloud implementation to be local.
randomjoe2 · 2h ago
Local doesn't refer to "on metal" anymore to many people
mwcz · 2h ago
"On metal" is muddied too. I've heard people refer to web apps running in an OCI container as being "bare metal" deployment, as opposed to AWS or whatever hosting platform.
That's silly, but the idea that "local" is not the opposite of remote is even sillier.
ffsm8 · 2h ago
You can run an OCI container on bare metal though. It doesn't stop being run on bare metal just because you're running in kernel namespaces, aka docker container
Lots of people were advocating for running their k8s on bare metal servers to maximize the performance of their containers
Now wherever that's applied to your conversation... I've no clue, too little context ( ｡ ŏ ﹏ ŏ )
okasaki · 2h ago
In my opinion, if you're running k8s on bare metal, that's "k8s on bare metal" but still "<your app> on kubernetes", not "<your app> on bare metal".
ffsm8 · 52m ago
Sorry, but then your opinion is just plain wrong
Bare metal in the context of running software is a technical term with a clear meaning that hasn't become contested like "AI" or "Crypto" - and that meaning is that the software is running directly on the hardware.
As k8s isn't virtualization, processes spawned by its orchestrator are still running on bare metal. It's the whole reason why containers are more efficient compared to virtual machines
bee_rider · 46m ago
Bare metal as in, no operating system? Does Linux really get in the way of these LLM inference engines?
ffsm8 · 40m ago
No, as I said in my previous comment: bare metal as in not a virtual machine
If you do bare metal as not being under a VM it fits. OCI on linux is cgroup so that counts as not a VM I'd say. Or at least it's a layer closer to the metal than a typical VM running OCI images.
I a Java app running on Linux bare metal?
bee_rider · 51m ago
Local doesn’t need to be “on metal,” but I’m still confused as to what they are saying. Are they running some local cloud system?
monsieurbanana · 2h ago
I missed that train
vFunct · 2h ago
My basement server really confused by all this...
DSingularity · 2h ago
I guess local for him is independent/private.
ollybee · 2h ago
H100's can be $2 and hour, so $192 an hour for the full cluster. They report 22k tokens per second, so ~ 80 million an hour, thats $16 an hour at $0.2 per million. Maybe a bit more for input tokens, but it seems a long way off.
zipy124 · 2h ago
I think you mis-read. Thats 22k tokens per second per node, so per 8 h100's. With 12 nodes they get 264k tokens per second, or 950 million an hour. This get's you to roughly $0.2021 per million at $2 an hour.
s46dxc5r7tv8 · 3h ago
Separation of the prefill and decoding layers with sglang is quite nifty! Normally 8xH100 would barely be able to hold the 4bit quantization of the model without even considering the KV cache. One prefill node for 3 decode nodes is also fascinating, nice writeup.
Just in case you have $3-4M lying around somewhere for some high quality inference. :)
SGLang quotes a 2.5-3.4x speedup as compared to the H100s. They also note that more optimizations are coming, but they haven't yet published a part 2 on the blog post.
abdellah123 · 3h ago
Wow, please edit the title to include Open-source !
numpad0 · 2h ago
These open models are just commercial binary distributions made available at zero cost with intention to cripple opportunities for Western LLM providers to capitalize on investments.
These are more like really gorgeous corporate swags than FOSS.
badsectoracula · 19m ago
> intention to cripple opportunities for Western LLM providers to capitalize on investments.
Western LLM providers release open weight models too (e.g. Mistral).
Blahah · 3h ago
Why? Open source isn't in the original title
SV_BubbleTime · 3h ago
Also “open source” I feel covers for “open weights” which is not the same thing.
This throughput assumes 100% utilizations. A bunch of things raise the cost at scale:
- There are no on-demand GPUs at this scale. You have to rent them for multi-year contracts. So you have to lock in some number of GPUs for your maximum throughput (or some sufficiently high percentile), not your average throughput. Your peak throughput at west coast business hours is probably 2-3x higher than the throughput at tail hours (east coast morning, west coast evenings)
- GPUs are often regionally locked due to data processing issues + latency issues. Thus, it's difficult to utilize these GPUs overnight because Asia doesn't want their data sent to the US and the US doesn't want their data sent to Asia.
These two factors mean that GPU utilization comes in at 10-20%. Now, if you're a massive company that spends a lot of money on training new models, you could conceivably slot in RL inference or model training to happen in these off-peak hours, maximizing utilization.
But for those companies purely specializing in inference, I would _not_ assume that these 90% margins are real. I would guess that even when it seems "10x cheaper", you're only seeing margins of 50%.
A node of 8 H100s will run you $31.40/hr on AWS, so for all 96 you're looking at $376.80/hr. With 188 million input tokens/hr and 80 million output tokens/hr, that comes out to around $2/million input tokens, and $4.70/million output tokens.
This is actually a lot more than Deepseek r1's rates of $0.10-$0.60/million input and $2/million output, but I'm sure major providers are not paying AWS p5 on-demand pricing.
Edit: those figures were per node, so the actual input and output prices would be divided by 12.$0.17/million input tokens, and $0.39/million output
And what stinks is that you can't even build a Dell/HPE server like this online. You have to 'request a quote' for an 'AI Server'
Going through SuperMicro, you're looking at about $60k for the server, plus 8 GPU's at $25,000 each, so you're close to $300,000 for an 8 GPU node.
Now, that doesn't include networking, storage, racks, electricity, cooling, someone to set that all up for you, $1,000 DAC cables, NVIDIA middleware, downtime as the H100's are the flakiest pieces of junk ever and will need to be replaced every so often...
Setting up a 96 H100 cluster (12 of those puppies) in this case is probably going to cost you $4-5 million.
An H100 costs about $32k, amortized over 3-5 years gives $1.21 to $0.7 per hour, so adding in electricity costs and cpu/ram etc... runpod.io is running much closer to the actual cost compared to AWS.
Reversing out these numbers tells us that they're paying about $2/H100/Hour (or $16/hour for a 8xH100 node).
Disclaimer (one of my sites) https://www.serversearcher.com/servers/gpu - says that a one month commit on a 8XH100 node goes for $12.91/hour. The "I'm buying the servers and putting them in COLO rate" usually works out at around $10/Hour, so there's scope here to reduce the cost by ~30% just by doing better/more committed purchasing.
Inference is more profitable than I thought.
Is that just the cost of electricity, or does it include the cost of the GPUs spread out over their predicted lifetime?
Maybe the cost of renting?
That's silly, but the idea that "local" is not the opposite of remote is even sillier.
Lots of people were advocating for running their k8s on bare metal servers to maximize the performance of their containers
Now wherever that's applied to your conversation... I've no clue, too little context ( ｡ ŏ ﹏ ŏ )
Bare metal in the context of running software is a technical term with a clear meaning that hasn't become contested like "AI" or "Crypto" - and that meaning is that the software is running directly on the hardware.
As k8s isn't virtualization, processes spawned by its orchestrator are still running on bare metal. It's the whole reason why containers are more efficient compared to virtual machines
https://en.m.wikipedia.org/wiki/Bare-metal_server
I a Java app running on Linux bare metal?
Just in case you have $3-4M lying around somewhere for some high quality inference. :)
SGLang quotes a 2.5-3.4x speedup as compared to the H100s. They also note that more optimizations are coming, but they haven't yet published a part 2 on the blog post.
These are more like really gorgeous corporate swags than FOSS.
Western LLM providers release open weight models too (e.g. Mistral).