# Compute Reservations

### Compute Reservations

USDAI’s integration with GPU.NET’s marketplace allows users to reserve and manage GPU compute resources efficiently, addressing both short-term and long-term needs with flexibility and precision.

#### Pre-Booking: Securing Resources Ahead of Time

* **Purpose**: Users can lock USDAI to **pre-book GPU resources**, ensuring availability during high-demand periods like AI research deadlines or product launches.
* **How It Works**: Through Dapp.gpu.net, users commit USDAI to reserve compute capacity for a future date. The system locks the tokens, allocating resources from GPU.NET’s provider pool based on availability and pricing.
* **Example**: An AI startup anticipates a surge in model training needs during a hackathon. They lock 200 USDAI to secure 200 GPU-hours a month in advance, avoiding last-minute shortages or inflated costs.
* **Benefits**: Pre-booking with USDAI guarantees access, mitigates supply risks, and provides cost certainty, critical for time-sensitive projects.

#### Dynamic Scaling: Real-Time Compute Adjustments

* **Purpose**: USDAI credits allow users to **adjust compute allocations dynamically**, scaling resources up or down in response to real-time needs without delays.
* **How It Works**: Users burn USDAI to access additional GPU capacity instantly or release unused credits back to the network, with allocations managed via smart contracts on GANChain or Solana.
* **Example**: A gaming company rendering real-time graphics for a live event starts with 50 USDAI worth of compute. As player demand spikes, they add 20 USDAI to scale up GPU resources seamlessly, then scale down post-event.
* **Benefits**: Dynamic scaling ensures efficient resource use, minimizes waste, and adapts to fluctuating workloads—ideal for applications like live AI inference or on-demand HPC.