RESERVE THE COMPUTE YOU NEED.
Hourly GPU compute for training, inference, and simulation. Reserve what you need.
Products
Choose the configuration that best fits your workload.
RTX 3090
NVIDIA GeForce RTX 3090 · 24 GB VRAM
| Size | GPU | VRAM | CPU | RAM | Disk | Hourly rate | Weekly | Monthly |
|---|---|---|---|---|---|---|---|---|
| S | 1× GPU | 24 GB | 4 cores | 16 GB | 100 GB | NPR 37.80/hr | NPR 6,350 | NPR 19,278 |
| M | 1× GPU | 24 GB | 8 cores | 24 GB | 200 GB | NPR 52.92/hr | NPR 8,891 | NPR 26,989 |
Need a second host instead of a bigger one?
How to reserve an instance in four steps
Choose GPU
GPU options: RTX 3090, RTX 5080, and RTX 5090 (available), DGX Spark and RTX PRO 6000 (coming soon).
Choose term
Hourly, daily (24×), weekly (168×), monthly (510× Pro).
Fund wallet
Top up via Stripe (USD), Fonepay, or GetPay (NPR).
Run workload
Connect via SSH and start computing.
Need custom GPU provisioning?
We can build a custom instance around your workload. Tell us how much GPU, CPU, memory, or storage you need, and we'll suggest a configuration.
Need configuration not listed in our catalog? Contact us and we'll help you set it up!
Talk to ANKSUPPORTED PAYMENT METHODS.
Fund your wallet in USD or NPR via Stripe, Fonepay, or GetPay.
Three steps to fund
Choose method, complete payment, use wallet once confirmed.
Stripe
Pay in USD with an international card.
- 1Choose Stripe
- 2Open secure checkout
- 3Confirm payment
Fonepay
Pay in NPR via Fonepay.
- 1Choose Fonepay
- 2Scan payment request
- 3Confirm payment
GetPay
Card banking for NPR payments.
- 1Choose GetPay
- 2Complete request
- 3Confirm payment
What we offer
ANK gives Nepal's AI builders capable GPUs, published hourly rates, and help choosing a configuration for the work at hand.
Clear product catalog
Compare RTX 3090, RTX 5080, and RTX 5090 now. Plan around GPUs marked Coming Soon.
Hourly pricing
Clear hourly rate before you begin. Top up in USD or NPR.
GPU guidance
Describe your workload. ANK helps identify a suitable GPU.
Built for AI work
Training, experimentation, or inference — we match the right compute.
Security in progress
Hardening platform with firewalls and network isolation.
Persistent storage
Separate plans (20–500 GB+) in NPR/USD. No egress fees.
What can I do with instances?
Training, inference, simulation, multimedia — any compute-intensive task.
Training runs
Train or fine-tune models on your own data.
Inference workloads
Run trained models for predictions and classifications.
Simulation & modeling
Run CFD, FEA, and other compute-intensive simulations.
Data processing
Process large datasets, run batch jobs, automate workloads.
Multimedia workloads
Process, edit, transform high-res video, images, media.
Research & development
Run computational experiments without owning GPU hardware.
Trusted by teams building with AI.
Hear from teams using AnK infrastructure to support their AI development and compute-intensive workloads.
“From understanding our requirements to helping us get started, the AnK team has been incredibly supportive. The compute has been promising and has accelerated our training workloads.”
“AnK's dedicated resources effortlessly handle our heaviest workloads, backed by an incredibly hardworking, technically sharp support team. AnK is the go-to infrastructure partner for any AI firm or researcher serious about scaling.”