GPU cloud providers
You sell GPU-as-a-Service: bare metal, GPU VMs, and inference. You need the bill to match the meter.
GPU billing software for neoclouds and data centers. Meter GPUs per second, price them on your own rate card, and send one invoice per customer. Inference billing per token and CPU billing for VMs and bare metal are included.
01 · Metering
Billing is only as good as the meter under it. RunOS records how long each workload holds a GPU, at every tier.
usage · example
| workload | tier | GPU-seconds |
|---|---|---|
| train-01 | on-demand | 1,209,600 |
| batch-07 | spot | 302,400 |
| chat-pool | inference | 518,400 |
02 · Pricing
An H200 doesn't cost what an L40S costs, and spot doesn't cost what on-demand costs. Put it all on one rate card and let the software do the sums.
rate card · example prices
| GPU | on-demand | spot | inference |
|---|---|---|---|
| H200 | $3.20/h | $1.60/h | $2.80/h |
| L40S | $1.10/h | $0.55/h | $0.95/h |
03 · Inference billing
Customers who call a model don't think in GPU-hours. They think in tokens. Sell inference the way they buy it.
inference rate card · example prices
| model | input / 1M | output / 1M |
|---|---|---|
| large-chat | $0.60 | $1.80 |
| small-chat | $0.10 | $0.30 |
| embeddings | $0.02 | n/a |
04 · Invoicing
Nobody likes a GPU bill they can't explain. Every line traces back to a workload and a time.
invoice · example
| line | hours | amount |
|---|---|---|
| H200 · on-demand | 336 | $1,075.20 |
| H200 · spot | 84 | $134.40 |
| H200 · inference | 144 | $403.20 |
| Total | $1,612.80 |
05 · CPU billing
Most data centers sell more than GPUs. RunOS GPU Billing also does CPU billing: virtual machines and dedicated servers go on the same invoice, so you run one data center billing system, not three.
one invoice · example
| line | type | amount |
|---|---|---|
| H200 · on-demand | GPU | $1,075.20 |
| 8 vCPU · 32 GB | VM | $86.40 |
| Dedicated server | bare metal | $420.00 |
| Total | $1,581.60 |
06 · Who it's for
You sell GPU-as-a-Service: bare metal, GPU VMs, and inference. You need the bill to match the meter.
You rent out VMs and servers, maybe GPUs next. You don't want to build billing from scratch.
You share one fleet across teams and want to charge each team for what it used.
Already have a billing system? Keep it. RunOS can export the usage records, and you price them yourself. Read more in neocloud software for GPU cloud operators.
07 · FAQ
GPU billing software turns GPU usage into charges. It records how long each customer held each GPU, applies your price for that GPU model and tier, and produces an invoice or draws down a prepaid balance. RunOS GPU Billing does this for neoclouds and data centers that run on RunOS.
Neocloud billing is billing for a GPU cloud provider. A neocloud rents out GPU capacity as bare metal, GPU VMs, and inference, so its billing software has to meter each of those per customer and put them on one invoice. It is also called GPUaaS billing, or GPU-as-a-Service billing.
Yes. RunOS GPU Billing is data center billing software as well as GPU billing software. It supports CPU billing: virtual machines are billed by vCPU and memory size, and dedicated bare-metal servers are billed for the time a customer holds them. They sit next to GPU lines on one invoice, so a hosting company or a data center with no GPUs can use it too.
Meter the time each workload holds a GPU, in seconds. Multiply that time by a rate that depends on the GPU model and on the tier: on-demand, spot, or inference. Add the lines up per customer for the billing period. RunOS does the metering, and RunOS GPU Billing does the rating and the invoice.
Yes. A rate card holds one price per GPU model and tier. On-demand, spot, and inference capacity are metered separately, so each one gets its own rate.
Yes. RunOS GPU Billing supports inference billing. It meters input and output tokens per API key and prices them per million tokens, per model. You can also bill an inference endpoint by GPU time, for a customer who wants dedicated capacity.
Yes. A customer can hold a prepaid balance that usage draws down, or you can invoice the customer at the end of the period.
Yes. RunOS GPU Billing reads metered usage from RunOS, so it is built for neoclouds and data centers that run on RunOS. If you already have a billing system, RunOS can also export the usage records to it.
Talk to us. Tell us what you run and how you charge for it today, and we will show you how it fits.
Tell us what you run and how you charge today. We'll show you how it fits.
Talk to us