CloudGPU Services
Enterprise GPU Infrastructure, Built on Sovereign Hydropower
Dedicated NVIDIA B300 capacity for Kathmandu — single-tenant, targeting low latency across South Asia, and powered entirely by renewable Himalayan hydroelectricity.
48x NVIDIA B300
Phase 1 on order
100% Hydropower
Renewable Himalayan hydroelectricity
Designed to align with DoIT Directive 2081
Sovereign data handling
Sub-15ms (target)
Regional latency across South Asia
Phase 1 build status
Company incorporated in Nepal
Complete, June 2026Hardware purchase order placed
CompleteColocation, TIA-942 Rated 3 facility, Kathmandu
In progressNetwork bandwidth
In progressPhase 1 GPU deployment
PlannedGeneral availability
Waitlist open
We publish our build status so prospective customers can plan around real timelines. Contact us for current detail.
Enterprise service lines
Four ways to consume dedicated GPU capacity, each backed by a dedicated account manager for allocation and configuration.
Bare-Metal GPU Clusters
Dedicated 8- and 16-GPU NVIDIA B300 nodes with high-bandwidth InfiniBand interconnect — single-tenant, no hypervisor overhead.
Virtual GPU Instances
Partitioned vGPU capacity for fine-tuning, training, and distributed inference — right-sized allocations without dedicated hardware.
Managed AI Inference Hosting
Managed orchestration for customer-owned clusters running vLLM, Ray, and TensorRT-LLM — we will operate the stack, you own the models.
Sovereign AI Appliances
Air-gapped, on-premise deployments for regulated BFSI and public-sector workloads that require full data isolation.
Model your costs
See current on-demand and committed rates across the full GPU lineup, with an interactive estimator.