GPU cards
Exact model, VRAM, card count and form factor for the node.
Dedicated GPU servers
A complete GPU server node, defined before order.
Dedicated GPU server capacity for teams that need control of the host, accelerator topology, storage, network and software stack.
The complete node
Four parts, quoted as one system.
A dedicated GPU server is more than the card count. The quote states each part, because any one of them can bottleneck the workload.
CPU and memory
Host compute and RAM sized so the cards are actually fed.
Storage
NVMe capacity and throughput for datasets, weights and checkpoints.
Network
Port, transfer and any interconnect the topology requires.
Per-card or per-node
Public reference prices are per GPU card per month in INR. A complete-node quote states the full system, current availability and the acceptance checks that close delivery.
See per-card reference pricesCommon questions.
Is the price for one GPU card or the complete server?
This page is for complete GPU server nodes. The written quote states the card count, server configuration and complete monthly price. Catalogue prices on the GPU hub are per GPU card unless their row says otherwise.
Can ZenoCloud operate the GPU server?
Yes. Add deployment, monitoring, maintenance, incident response and recovery support as separate Infrastructure Management services.
Which GPU architectures are available?
Current models, form factors, card counts and locations are confirmed during qualification. ZenoCloud does not treat a model name alone as proof of an orderable node or workload fit.
Is dedicated GPU server pricing published per card?
No. A dedicated-node quote covers the complete server, including every GPU, CPU, RAM, storage, network and location. Do not multiply a per-card catalogue price to estimate a node.