> For the complete documentation index, see [llms.txt](https://docs.nebulanetwork.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nebulanetwork.ai/gpu-marketplace/overview/detailed-rental-guide/renting-a-gpu.md).

# Renting a GPU

The Nebula AI platform provides a **decentralized, peer-to-peer GPU rental system**, allowing users to lease high-performance compute power for AI training, deep learning, rendering, and more. Renting a GPU is a streamlined process, ensuring security, flexibility, and transparent pricing. This section will guide you through the complete process, from **connecting your wallet** to **deploying workloads on a rented GPU**.

***

### **Connecting Your Wallet & Funding Your Account**

To access the Nebula AI marketplace, you must connect a **Web3-compatible wallet**.

1. Click **"Connect Wallet"** on the platform.
2. Select your preferred wallet and approve the connection.
3. Ensure your wallet is funded with **$NAI** or **ETH**.

Once connected, your wallet serves as your **Nebula AI account.**

***

### **Browsing the GPU Marketplace**

After connecting your wallet, you can browse the **available GPUs** using the marketplace interface. Each listing provides detailed specifications, ensuring you select the best GPU for your workload.

The marketplace allows you to filter GPUs by:

* **Model & Performance**
* **VRAM Size**
* **Compute Power**
* **Availability & Rental Type** (Spot vs. On-Demand)
* **Pricing & Rental Duration**
* **Location**

Clicking on a listing provides more **detailed specifications**, including performance benchmarks, uptime history, and past rental reviews.

***

### **Selecting Rental Type: Spot vs. On-Demand**

Nebula AI offers two types of GPU rentals:

#### **On-Demand Rentals (Guaranteed Access)**

* Provides **exclusive access** to a GPU for the entire rental period.
* Ideal for **long-running AI training** or **continuous workloads** that cannot tolerate interruptions.
* **Higher pricing** due to guaranteed uptime.

#### **Spot Rentals (Dynamic Pricing, Interruptible)**

* Available at **a lower cost**, but another renter may outbid you and take over the rental.
* Suitable for **batch processing tasks** and **non-urgent workloads**.
* **Great for cost savings**, but not recommended for long-term, mission-critical workloads.

You can set **maximum bid limits** for **Spot Rentals**, ensuring you only compete within your budget.

***

### **Confirming & Deploying Your Rental**

Once you’ve selected a GPU and rental type, you will proceed to payment.

1. Click **"Rent Now"** and approve the **smart contract transaction** from your Web3 wallet.
2. Once the transaction is confirmed, your rental session will start.
3. You will receive **connection details** based on the selected **access method**:
   * **SSH Credentials** (For direct Linux terminal access)
   * **Jupyter Notebook** (For ML/AI workflows)
   * **API Endpoints** (For automated workload deployment)

Your rental session will be **immediately available**, and you can start executing your workloads on the rented GPU.

***

### **Managing Your Rental Session**

Once your rental is active, you can monitor its performance from the **Rental Dashboard**. The dashboard provides:

* **Live GPU usage metrics** (VRAM, Compute Load, Power Consumption).
* **Session Timer** (Time remaining for rental).
* **Extend Rental Option** (If additional time is needed).
* **End Rental Button** (To manually stop usage before the session expires).

Rentals **automatically expire** when the time limit is reached, and all **session data is wiped**. **Make sure to back up your work before the rental ends.**

***

### **Ending Rental & Auto-Expiration**

When your rental period **expires**, the GPU is immediately released back to the marketplace. If you need more time, **you must extend the rental before expiration**.

💡 **Important Notes:**

* **No refunds** are provided for unused rental time.
* **All storage is cleared** upon expiration, so ensure you save your work externally.
* If you end your session early, the remaining time **cannot be transferred**.

For users requiring **persistent storage** across rental sessions, Nebula AI will introduce **cloud-linked storage options** in future updates.