GPU Inventory — India Nodes (Private Beta)

Rent GPU compute.
Pay via UPI.

Browse NVIDIA GPUs from Indian providers. RTX 4090 from ₹44/hr. No dollar conversion. GST invoice. Email alerts when your instance is ready.

₹12/hr starting price · 5–20ms latency (India nodes) · UPI / Card payment · GST invoice included · 15% bonus for students
All GPUs RTX 4090 ₹44/hr RTX 3090 ₹34/hr RTX 3080 ₹25/hr RTX 4080 ₹38/hr RTX 3070 ₹20/hr
🚀 Launch with a template — click to filter compatible GPUs ✕ Clear filter
🔥PyTorch 2.3Training
🧠TensorFlow 2.16Training
📓JupyterLabNotebook
🎨Stable Diffusion XLImage Gen
🖼️ComfyUIImage Gen
🦙OllamaLLM · Local
vLLMLLM · API
🎙️WhisperAudio / STT
🤖DeepSeekCode · LLM
👁️LLaVAVision LLM
🐳Custom DockerBring your own
🐧Ubuntu BareRaw CUDA
Showing 12 GPUs across India

📊 GPU Pricing — All 9 tiers All prices ex-GST (18% GST added at checkout, claimable as ITC for businesses)

Tier GPU Models VRAM Price (on-demand) Avg Latency (India)
StarterRTX 2060, RTX 3060, GTX 1660 Ti6–12 GB₹12/hr5–15ms
PlusRTX 2070, RTX 3060 Ti, 2070 Super8 GB₹15/hr5–15ms
StandardRTX 2080, RTX 3070, RTX 40708 GB₹20/hr5–20ms
Standard+RTX 2080 Ti, RTX 3070 Ti11 GB₹22/hr5–20ms
ProRTX 3080 (10GB), RTX 4070 Ti10–12 GB₹25/hr5–20ms
Pro+RTX 3080 Ti, RTX 3080 12GB12 GB₹28/hr5–20ms
Pro MaxRTX 309024 GB₹34/hr5–20ms
Pro Max+RTX 4080, RTX 4080 SUPER16 GB₹38/hr5–20ms
UltraRTX 4090 24GB24 GB₹44/hr5–20ms
ExtremeRTX 509032 GB₹46/hr5–20ms

Simple on-demand pricing — pay per second after the first minute. No bidding, no lock-in.

How it works
Rent GPU compute in seconds. No setup required — use our pre-built templates or bring your Docker image.
1
Top up wallet
Add ₹100–₹10,000 via UPI (GPay, PhonePe, Paytm) or debit/credit card. Students get 15% bonus on every top-up (up to ₹3,000 total).
2
Pick a GPU
Browse by GPU model, VRAM, city, and price. See live availability and reliability scores.
3
Choose template
One-click: PyTorch, TensorFlow, JupyterLab, Stable Diffusion, vLLM, Whisper. Or bring your own Docker image.
4
SSH in & run
Instance ready in seconds. SSH directly or use Jupyter. Live GPU/memory metrics. Email alert when ready. Per-second billing.
GridShare vs. alternatives
Why Indian ML teams choose GridShare over US clouds and Indian alternatives.
Feature GridShare Jarvislabs Vast.ai RunPod E2E Networks
RTX 4090 price ₹44/hr 🏆 ₹62/hr $0.42–0.69 (≈₹35–57) $0.74 (≈₹61) Not available
UPI / INR payment UPI + Card INR card USD only USD only INR card
India-hosted nodes All India India Global P2P US/EU India
GST invoice Always
Email alerts (WhatsApp coming)
Student 15% bonus .ac.in / .edu.in
Latency (Bangalore) 5–20ms 5–20ms 150–250ms 150–250ms 8–25ms
Serverless inference API Scale-to-zero · India US only
OpenAI-compatible API vLLM drop-in
AMD ROCm GPU support Private Beta Limited
SSH public key management Auto-injected
Secrets / encrypted env vars Fernet AES
Webhooks (HTTP callbacks) HMAC signed
Team / Org accounts Shared wallet
Persistent storage volumes Beta · ₹6/GB/mo
Budget / spend alerts Email (WhatsApp coming)
Referral programme ₹200 both ways

Prices as of June 2026. Vast.ai/RunPod prices converted at ₹83/$. ✓ = available. ✗ = not available.

Beta · Scale to Zero · India Nodes

Serverless GPU Inference API

Deploy any AI model as an HTTP endpoint. GridShare auto-scales GPU workers per request and charges only for active compute — zero cost when idle. Per-second active-compute billing. In beta — running on a young network, so expect rough edges.

₹0.0125/sec
Active compute (RTX 4090)
₹0
Cost when idle (scale-to-zero)
100ms
Min billing quantum
<2s
Cold start (warm worker pool)
Create Endpoint → View API Docs
Always-on GPU instance
₹44/hr
Billed even when no requests come in
vs
Serverless endpoint (same GPU)
₹4.50/hr
If inference runs 10% of the time — 90% saved
=
90% cost reduction
For batch workloads, internal tools, and dev environments that see burst traffic — serverless is the only rational choice.
Pre-built Serverless Endpoints
🦙
LLM Inference
vLLM · Llama 3 8B / 70B · Mistral · Gemma
Speed
~2s / request
Cost
₹0.025/req
GPU
RTX 4090 24GB
Context
128K tokens
OpenAI-compatible API. Drop-in replacement for OpenAI — just change the base_url. Supports streaming, function calling, and JSON mode.
🎙️
Speech-to-Text
Whisper Large v3 · FasterWhisper · Hindi + 99 languages
Speed
5s per min audio
Cost
₹0.06/min audio
GPU
RTX 3080 10GB
Languages
Hindi, EN, 98+
Transcribe audio files or live streams. Native Hindi, Tamil, Bengali support. Returns timestamped segments and speaker diarization.
🎨
Image Generation
Stable Diffusion XL · FLUX · ControlNet · IP-Adapter
Speed
~15s / image
Cost
₹0.19/image
GPU
RTX 4090 24GB
Resolution
Up to 2048×2048
Generate images via REST API. Pass a prompt, get a URL back. LoRA fine-tunes, ControlNet poses, and IP-Adapter style transfer all supported.
🐳
Custom Model
Bring your own Docker image · Any framework · Any model
GPU VRAM
6–24 GB
Framework
Any
Scale
0 → 20 workers
Billing
Per-second
Push your Docker image, define a /run endpoint, and GridShare handles auto-scaling, billing, and monitoring. Works with PyTorch, TensorFlow, JAX, TensorRT.

🧑‍💻 Call your endpoint — 5 lines of code

# Install: pip install gridshare import gridshare # Authenticate with your API key (from /app → API Keys) client = gridshare.Client(api_key="gs_live_...") # Call a serverless endpoint (async, scales to zero when idle) result = client.serverless.run( endpoint="my-llama-3-8b", input={"prompt": "Explain gradient descent in Hindi", "max_tokens": 512} ) print(result.output) # ₹0.025 billed (2 seconds compute)
How billing works

⚡ Active billing

You pay ₹0.0125/sec only while a GPU worker is actively processing your request. A 2-second LLM call costs ₹0.025. Minimum: 100ms.

💤 Idle = ₹0

When no requests are running, workers scale to zero. You pay nothing. No monthly minimums, no reservation fees, no standing charges.

🔥 Keep-warm option

Set min_workers=1 to keep one GPU warm at all times. Billed at ₹0.0025/sec idle (20% of active rate) — eliminates cold starts.

📈 Auto-scale up to 20×

GridShare auto-spawns additional GPU workers as queue depth grows. Scale from 0 → 20 workers in seconds. Each worker bills independently.

💳 Same wallet

Serverless usage draws from your GridShare wallet — the same one used for GPU instances. Top up once via UPI, use across all compute types.

🧾 GST invoice

All serverless billing is included in your monthly GST invoice. Startups with GSTIN can claim Input Tax Credit on all compute spend.

Ready to go serverless?
Create a free account → top up wallet → deploy your first endpoint in under 5 minutes.
Get Started Free →