Everything to get you running in 5 minutes — renting a GPU, or earning from one. Stuck anywhere? support@gridshare.in or the 🛟 Get help button on any running instance.
From zero to a running GPU notebook:
Sign up at gridshare.in/marketplace with your email and verify it (we send a one-time code). No credit card needed.
Add money from ₹100 via UPI/Razorpay — no forex card, no USD. Every top-up gets a proper GST tax invoice (in your company's name if you set a GSTIN — useful for expensing).
The GPUs page shows live machines with the exact ₹/hr rate, VRAM, and a measured compute score. Filter by VRAM or price. What you see is what you pay — one honest rate, no hidden tiers.
One click: PyTorch + JupyterLab, Stable Diffusion, Ollama/LLM chat, Whisper, VS Code, and 20+ more. The instance boots and shows Open (browser) and SSH buttons when ready — first launch on a node can take a few minutes while the image downloads.
Everything under /workspace is yours and auto-saved (~every 30 seconds, up to 4 GB). Billing is per second — the dashboard shows the live meter.
Hit Stop — billing stops that second. Your files in /workspace are saved and kept for 30 days (first 1 GB free, then ₹6/GB per month), so pressing ▶ Resume later puts them back. Balance running low? We warn you by email, and the instance auto-stops before you can go negative.
Sign up, copy your one-line command, paste it. Pause your machine from the dashboard any time and it takes no new rentals.
Sign up or sign in → and your dashboard shows a personalised command for your OS. It carries a single-use setup token, so the machine registers to your account and the terminal asks you nothing.
Ubuntu/Debian, driver 520+. Docker and the NVIDIA toolkit are installed automatically.
RX 6000/7000 series. ROCm set up by the installer.
Windows 10/11 with WSL2 — the portal also offers a double-click installer. The script handles WSL setup.
Any M-series Mac. Native agent — no Docker needed.
Add and verify your bank details in the dashboard (never in the terminal). Choose daily payouts (₹10 flat fee, from ₹150), weekly (₹5, from ₹300) or monthly (free, from ₹500) — sent by IMPS/NEFT to any Indian bank.
Instances: per second, only while your instance is running. On GridShare app templates the meter waits until your app is actually ready — you don't pay for the boot wait (up to a 15-minute window). Stopping ends billing immediately. Your saved /workspace is kept for 30 days (first 1 GB free, then ₹6/GB per month) after you Stop (Resume or ▶ Resume restarts the clock; deleted after 30 days if untouched). Every payment gets a GST invoice.
Serverless endpoints bill differently. You pay per second of compute actually used, plus 20% of that rate while a worker is held warm between requests so your next call skips the cold start. A warm worker shuts down after 10 idle minutes and the charge stops; it also stops if you pause, delete or scale the endpoint to zero. Keep-warm shows as a separate daily line per endpoint in your billing history. See the keep-warm section of the Terms.
The files under /workspace (checkpoint cap 4 GB). They are autosaved every ~30 seconds while running and again when you press Stop, restored when you Resume on any available GPU, and kept for 30 days (first 1 GB free, then ₹6/GB per month) after you Stop. Not included, by design: anything outside /workspace (for example software installed into the home directory over SSH), Python environments (venv, site-packages) and model files (.safetensors, .ckpt, .pt, .bin, .gguf) — re-install or re-download those after resuming, or switch on Keep my large files on the rental card to keep model files and anything over 256 MB in GridShare cloud storage (up to 500 GB, ₹6/GB per month). On SSH instances you start in /workspace.
Billing halts the moment a node is unreachable. Any seconds you paid for but didn't get are refunded automatically when the workload relocates. You never pay for GPU time a failed node did not deliver. (Serverless keep-warm is a separate, disclosed charge for holding capacity ready — see "How exactly am I billed?" above.)
Un-served time (a node dying mid-run) is refunded automatically. Your unused wallet balance — funds you loaded but haven't spent — stays yours and never expires. A cash refund of it is at our discretion, for a genuine reason, on request by email (less the ~2% gateway fee), and it is returned if you close your account. Delivered compute and bonus credits are non-refundable. Details in the refund policy, or write to support@gridshare.in.
Your code, data and outputs in /workspace survive — auto-saved every ~30 s, up to 4 GB in total. What is skipped, whatever the file size: anything inside a folder named venv, .venv, site-packages, node_modules, models, .cache, .git, wandb, .ollama or __pycache__. So keep downloaded weights in /workspace/models/ — they re-download on the new machine and never eat your 4 GB. A small fine-tuned adapter or training checkpoint saved anywhere else under /workspace does come with you; download anything critical before you stop.
Payouts go to your verified bank account (IMPS/NEFT) on your chosen schedule once your balance crosses the minimum — ₹150 on the daily plan, ₹300 weekly, ₹500 monthly. Below the minimum, earnings simply roll over; nothing is lost. Add your PAN in the dashboard — payouts without a PAN attract higher TDS as required by law.
Workloads run isolated (container/sandbox) with no access to the host's files — and the host has no access to yours inside the container. Workspace auto-saves are stored encrypted at rest in our object storage (Singapore region — disclosed in the privacy policy). For confidential-compute needs, talk to us about enterprise options.
Every running instance has a Get help button that attaches your instance context to a support ticket automatically — fastest route for instance issues.
support@gridshare.in — support hours are 11 AM – 8 PM IST. Messages that arrive outside these hours are answered from 11 AM. Abuse reports: abuse@gridshare.in.