"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?
replies(3):
Maybe the cost of renting?
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
That's silly, but the idea that "local" is not the opposite of remote is even sillier.
Inference is more profitable than I thought.
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 )
These are more like really gorgeous corporate swags than FOSS.
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.
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%.
I a Java app running on Linux bare metal?
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.
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.
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
Western LLM providers release open weight models too (e.g. Mistral).
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. But it should cost less than AWS after a year and a half.
Of course, a process running inside Kubernetes Pod, on a baremetal node will show up in `top` if I run it on the node directly. In such terms, it is running directly on hardware.
But when I deploy this Pod, I'm not interacting with the OS in any way. I'm interacting with Kubernetes apiserver, telling it what to run, not really caring about the operating system underneath. In such terms, the application is running "in k8s".
I consult in this space and clients still don't fully understand how complex it can get to just "run your own LLM".
You can also use duty cycle metrics to scale down your gpu workloads to get rid of some of the slack.
> These two factors mean that GPU utilization comes in at 10-20%.
Why don't these two factors cancel out? Why wouldn't a company building a private GPU cluster for their own use, also sit a workload scheduler (e.g. Slurm) in front of it, enable credit accounting + usage-based-billing on it, and then let validated customer partners of theirs push batch jobs to their cluster — where each such job will receive huge spot resource allocations in what would otherwise be the cluster's low-duty point, to run to completion as quickly as possible?
Just a few such companies (and universities) deciding to rent their excess inference capacity out to local SMEs, would mean that there would then be "on-demand GPUs at this scale." (You'd have to go through a few meetings to get access to it, but no more than is required to e.g. get a mortgage on a house. Certainly nothing as bad as getting VC investment.)
This has always been precisely how the commercial market for HPC compute works: the validated customers of an HPC cluster sending off their flights of independent "wide but short" jobs, that get resource-packed + fair-scheduled between other clients' jobs into a 2D (nodes, time) matrix, with everything getting executed overnight, just a few wide jobs at a time.
So why don't we see a similar commercial "GPU HPC" market?
I can only assume that the companies building such clusters are either:
- investor-funded, and therefore not concerned with dedicating effort to invent ways to minimize the TCO of their GPUs, when they could instead put all their engineering+operational labor into grabbing market share
- bigcorps so big that they have contracts with one big overriding "customer" that can suck up 100% of their spare GPU-hours: their state's military / intelligence apparatus
...or, if not, then it must turn out that these clusters are being 100% utilized by their owners themselves — however unlikely that may seem.
Because if none of these statements are true, then there's just a proverbial $20 bill sitting on the ground here. (And the best kind of $20 bill, too, from a company's perspective: rent extraction.)
I assumed it was to slightly correct this problem and on the surface it seems like it'd be useful for big data places where process-eventually is enough, e.g. it could be a relatively big market. Is it?
A company standing up this infrastructure is presumably not in the business of selling time-shares of infrastructure, they're busy doing AI B2B pet food marketing or whatever. In order to make that sale, someone has to connect their underutilized assets with interested customers, which is outside of their core competency. Who's going to do that?
There's obviously an opportunity here for another company to be a market maker, but that's hard, and is its own speciality.
It also does not always "clearly" have this new meaning. Somebody who is used to running programs directly (with no intermediate OS) on hardware might not understand what you're saying, or might ask you to clarify, and you probably shouldn't feel put upon by a totally understandable misinterpretation.
edit: Especially when you keep repeating "directly on hardware" when you mean "not on a VM." VMs also run on hardware. You're saying that you're only running on one OS instead an OS in your OS.
A major issue we have right now is, we want the coding process to be more "Agentic", but we don't have an easy way for LLMs to determine what to pull into context to solve a problem. This is a problem that I am working on with my personal AI search assistant, which I talk about below:
https://github.com/gitsense/chat/blob/main/packages/chat/wid...
Analyzers are the "Brains" for my search, but generating the analysis is both tedious and can be costly. I'm working on the tedious part and with batch processing, you can probably process thousands of files for under 5 dollars with Gemini 2.5 Flash.
With batch processing and the ability to continuously analyze 10s of thousands of files, I can see companies wanting to make "Agentic" coding smarter, which should help with GPU utilization and drive down the cost of software development.
You don't know who owns the GPUs / if or when your job will complete and if the owner is sniffing what you are processing though
"Linux was a Finish conspiracy to cripple hard-working US operating systems makers. It's not really open because I don't understand it."
Leave nationalism to those who want to use you as mere pawns in their imaginary Chess game.
Their GPU availability is amazing though
The hot parts are/were on allocation to both vendors. They try to sus out your use case and redirect you to less constrained parts.
Depreciation and GPU failure rate over time must be considered, which I don't see mentioned in the article.
However, I don’t think these companies provision capacity for peak usage and let it idle during off peak. I think they provision it at something a bit above average, and aim at 100% utilization for the max number of hours in the day. When there is not enough capacity to meet demand they utilize various service degradation methods and/or load shedding.
This is why I use DS though. I think its the only ethical option due to its efficiency. I think that outweighs all other considerations at this point.
I validate the compute renters because ITAR. Lots of hostile foreign powers trying to access compute .
My main business is ITAR related , so I have incredibly high security in place already.
We are multi tenant from day zero and have slurm etc in place for accounting reasons for federal contracts etc. we actually are spinning up federal contracting as a service and will do a ShowHN when that launches.
Riches in the niches and the business of business :)
Almost absolutely no one releases their training data.