On-demand GPUs
Whole GPUs, by the second.
Use an NVIDIA H200 NVL when the work needs one and stop when it's done. You pay for the seconds it runs, with a 60-second minimum, in rupees.
Coming soonEnterprise early access is open
How it works
Three steps.
- 01
Open your workspace
A notebook or a pipeline, ready when you sign in.
- 02
Start on an H200 NVL
One GPU, or several for a bigger job.
- 03
Stop, and the meter stops
You pay for the seconds the work ran, in rupees.
See it
Per second, in rupees. Nothing rounded up.
- Billing
- Per secondGPU time and MIG slices are metered by the second they run.
- Minimum
- 60 s60 seconds per run. Nothing is rounded up to the hour.
- Currency
- NPRInvoiced in Nepalese rupees, so there is no exchange-rate margin.
- Ways to pay
- eSewa · KhaltiOr bank transfer and invoice. VAT at the statutory rate.
Rates are being finalised. Every request gets a firm quote in rupees within one working day.
What you get
On-demand GPUs, in detail.
Billed by the second
No rounding up to the hour. Stop the job and the meter stops with it, after a 60-second minimum.
141 GB on every GPU
Room for large models and long contexts on a single card, at 4.8 TB/s.
Notebooks and pipelines
JupyterLab, Kubeflow pipelines and Ray jobs, ready in your workspace.
No commitment
Nothing to reserve and no contract. When the work turns steady, reserved capacity costs less.
Close to your data
Storage sits beside the GPUs in the same datacentre, so data never crosses a border to be read.
Under the hood
The key numbers.
- GPU memory
- 141 GB
- Memory bandwidth
- 4.8 TB/s
- Minimum per run
- 60 seconds
- GPU memory
- 141 GB
- Memory bandwidth
- 4.8 TB/s
- Minimum per run
- 60 seconds
- GPU
- NVIDIA H200 NVL
- Currency
- Nepalese rupees (NPR)
- Ways to pay
- eSewa, Khalti, bank transfer or invoice
Billing
Per second, 60-second minimum
Commitment: none. Rates are being finalised; every request gets a firm quote in rupees.
How billing worksQuestions
About on-demand gpus.
What is the smallest charge?
Sixty seconds. After that you pay for each second the work runs.
Can we use more than one GPU?
Yes. Ask for the number of GPUs a job needs when you request access, and we plan capacity for it.
Start with on-demand gpus.
Tell us what you want to run. It takes about a minute.