Reserved GPUs

Capacity that's yours alone.

Reserve NVIDIA H200 NVL GPUs for a term you choose. They are held for your organisation the whole time, and nobody else's work runs on them.

Coming soonEnterprise early access is open

How it works

Three steps.

01
Tell us the work
The models, the data and the deadline. A few sentences is enough.
02
Get a capacity plan
GPUs, storage and term, with a firm quote in rupees within one working day.
03
Run on capacity held for you
Your GPUs are reserved for you from the first day of the term to the last.

What you get

Held for the whole term
No queue and no contention: the GPUs are yours from the first day of the term to the last.
From one GPU to a whole server
Reserve a single H200 NVL, or several of them.
Nobody else on your GPUs
Reserved GPUs run only your organisation's work, which keeps tenancy questions simple.
The same workspace
Notebooks, pipelines, training and serving in your workspace, on GPUs that are always there.
Data that stays in Nepal
Datasets, checkpoints and models stay in our datacentre in the Kathmandu Valley.
Rates set by the term
Your quote shows the rate for the term up front, in rupees, with nothing added later.

Under the hood

Billing

Commitment: a term you choose. Rates are being finalised; every request gets a firm quote in rupees.

Questions

How long is a term?

You choose. Tell us how long you need the capacity and we quote the term in rupees.

Does anyone else use our GPUs?

No. Reserved GPUs run only your organisation's work for the whole term.

What do we work in?

The same AI workspace as every CoreValley product: JupyterLab, Kubeflow, Ray, MLflow, KServe and more, ready on day one.

Tell us what you want to run. It takes about a minute.