Powerful GPUs, rented by the second.
Rent a GPU from ₹11/hr, with JupyterLab open in your browser in under a minute. Or skip the GPU: one API call returns an image from ₹2 or a video clip from ₹42. Pay with UPI from a prepaid wallet that can never overspend.
RTX 4090
24 GB VRAM
₹47/hr
published rate
What a run actually costs
₹11/hr
GPUs from
44+
GPU types priced
₹0
forex markup, billed in INR
Per-second
billing, no minimum
One-click templates
Pick a template, start working
Each one boots with CUDA, drivers and the tools already installed, so the first minute goes on your work and not on setup.
PyTorch + Jupyter
Fine-tuning, research and coursework. JupyterLab in your browser with PyTorch and CUDA ready.
Launch PyTorch + JupyterStable Diffusion (ComfyUI)
Image and video generation with the ComfyUI node editor, opened straight in your browser.
Launch Stable DiffusionLLM (text-generation-webui)
Chat with and serve open models such as Llama and Qwen. Load a model and start talking.
Launch LLMDon't need a whole GPU? Just type a prompt.
Get an image or a short video, made on the same GPUs. Images from ₹2, video from ₹42. Nothing to rent, nothing to install, no subscription. A generation that fails costs you nothing.
Building an app? The same images and clips come from one API call, billed from the same wallet.
Images
Video
Generated with openly licensed AI models. Powered by Stability AI.
How it works
From signup to a running GPU
1. Add credits
Top up in a few clicks with UPI, card or netbanking.
2. Pick a GPU and a template
Every GPU online shows its model, VRAM and hourly rate before you launch.
3. Run, then stop
JupyterLab opens in under a minute, or connect over SSH. Billing is per second and ends the moment you stop.
Transparent pricing
Popular GPUs, transparent rates
Use cases
One marketplace, every GPU workload
Fine-tune LLMs
LoRA/QLoRA and full fine-tunes on 24GB–80GB cards without the cloud bill.
Serve inference
Spin up vLLM or Ollama and serve open models with low latency.
Image & video gen
Stable Diffusion, ComfyUI and diffusion pipelines on fast consumer GPUs.
Research & notebooks
Jupyter in under a minute for experiments, papers, and coursework.
Batch & training jobs
Per-second billing for checkpoint-friendly training runs — the meter stops when the run does.
Build AI products
Prototype to production on the same rupee-billed marketplace.
Earn with GHive
Your idle GPU is leaving money on the table
Got a gaming rig, an ex-mining setup, or spare datacenter capacity? Install the GHive Host app, click Start, and earn while you sleep. 80% of every rupee a renter pays goes to you.
- Taskbar app — one click to start or stop hosting
- GHive sets a fair fixed rate for your GPU
- Track hours & earnings in your dashboard
- Paid out to your UPI
Estimated monthly earnings
What you keep after electricity, from Quiet (2 h rented a day) to Busy (12 h). An estimate, not a promise — demand decides your hours. Try your own numbers
Trust, plainly
What we protect, and what we don't
Isolated from other renters
Each rental runs in its own unprivileged container with its own disk.
Prepaid, never overspend
Your instance stops when the wallet runs out. You can never owe more than you loaded.
Erased when you stop
Your workspace is wiped at the end of every rental. Download what you want to keep first.
Hosts are bound by contract
Hosts own the hardware, so they could technically see a rental's files. Our Host Agreement forbids it, and breaking it means removal and forfeited earnings. Don't upload secrets you can't rotate.
FAQ
Questions, answered
GPUs start from ₹11/hr. GHive publishes one transparent rate per GPU type, billed per second from your prepaid wallet — no setup fees, no egress traps, no minimums.
No. Sign up, then top up with UPI, card or netbanking.
Under a minute. Pick a GPU, click Rent, and JupyterLab opens in your browser; SSH works too. No application or sales call.
Each rental runs in its own isolated container, and its storage is wiped when you stop. The host owns the hardware, so they could technically see a rental's files; our Host Agreement forbids it, with removal and forfeited earnings for anyone who does. Don't upload secrets you can't rotate — use short-lived API keys.
Whatever hosts have online right now; Browse shows the live list. GHive publishes a rate for 44 GPU classes, from the RTX 3060 Laptop to the H100.
Yes. Install the GHive Host app, click Start, and earn a transparent per-GPU rate. You control when your GPU is available.
Rent your first GPU today.
No card, no waitlist, no minimum. Pay per second from a prepaid wallet.