Products
GPU pods
Run your own containers on NVIDIA H200 GPUs. Take a whole card, a hardware slice or a cheaper shared slice, and pay by the second either way.
Early access is open to enterprises on NVIDIA H200. More GPUs, and access for more teams, are coming soon. We reply within one working day.
built for ml engineers, startups and research teams. Technology: mig · hami · per-second billing
Hardware slices (MIG)
Split an H200 into as many as seven isolated parts, each with its own memory and compute. If a neighbour's job crashes or runs out of memory, yours keeps running.
Shared slices (HAMi)
When you don't need hardware isolation, take a share of a card's memory and compute instead. It packs more work onto each GPU, so it costs less, and we label the difference rather than hide it.
Billed by the second
You pay for GPU-seconds, with a 60-second minimum. Stop a pod and the meter stops with it: no rounding up to the hour, no charges after it ends.
Ready-made images
CUDA, PyTorch, TensorFlow, vLLM, DeepSpeed and the usual fine-tuning tools are installed from first boot. Or bring your own image.
Storage that outlives the pod
Attach a replicated network volume that survives restarts, or use fast local NVMe for scratch. Mount the same volume in every pod in a project.
SSH and open ports
Root inside your container, SSH with your own key, and open ports for TensorBoard, Jupyter or your own service.
- GPU
- H200 · 141 GB · available now
- Coming soon
- RTX PRO 6000 · 96 GB
- Ways to rent
- whole card · MIG · HAMi
- Billing
- per second, 60 s minimum
- Region
- np-ktm-1 (Kathmandu)
Start with a quote.
Tell us what you want to run. We reply within one working day with a capacity plan and a firm rupee quote.
Other products