Developer

Build on the World's Pioneering Platform with Composable Memory-to-Compute Ratio.

up to

0GPUs

per instance

more than

0TB

memory capacity36 memory nodes × 4 TB each

per node

0GB/s

memory bandwidth

Invitation Only

Augmenting GPU with Potentially Infinite Memory

Expand GPU memory capacity beyond HBM limits — without adding compute nodes. X-Mem™ pilot lets you configure your ideal setup and reserve a VM in minutes. Choose how to get started below.

Already have an access token?

Reserve Now

Don't have one yet?

Get Early Access

Number of GPUs1

12345678

MTier Memory Capacity0 GB

02 TB4 TB6 TB8 TB

Instance Layout

GPU$2.501× · $2.50/GPU·h

Memory$0.000 GB · $0.01/GB·h

Total / hour$2.50credits

* The Instance Layout shown is for illustrative purposes only. Hardware specifications and pricing will vary by configuration — refer to the X-Mem™ pilot product page for the latest details, or contact us with any questions.

SDK/API Documentation

Learn how X-Mem™ integrates with your AI infrastructure. Architecture guides, API references, and integration examples.

Early Access

Performance Models

Simulate and reason about memory-to-compute ratios and its impact on performance. Explore design space for your generative AI systems.