Lium (SN51) Crosses $30 Million in Hardware With 225 B300 GPUs and 91% Utilization

Getting enough GPU power for an AI project can be frustrating when cloud providers have long waitlists, complicated onboarding, or expensive pricing.

Lium, a decentralized GPU rental marketplace, is trying to remove those friction points, and its latest numbers show the platform is growing fast.

Here are the key takeaways:

1) Over $30 million in hardware hosted on the platform.

2) 225 NVIDIA B300 GPUs included in its stack.

3) 91% of supply is rented out.

4) Flexible access through the web dashboard, CLI, SDK, and AI agent tools, with per-second billing and no minimum rental period.

5) 95% of rental fees go to providers, with additional rewards available for idle hardware. 

The B300 is one of NVIDIA’s newest and most powerful data-center cards. It carries 288 GB of high-bandwidth memory and is built for large AI models.

Lium lists them for rent in the roughly $7–$10 per GPU-hour range depending on the pod and location. Having 225 of them available on an open marketplace is notable because they remain scarce on many traditional clouds.

For AI developers who just need compute, the practical takeaway is straightforward. More B300 capacity is live and mostly rented, which signals real demand.

Hardware providers can list qualifying GPUs to earn from rentals, while users can check current prices and availability directly at lium.io.

Together, these facts point to a decentralized GPU market moving past the early experimental stage and into something people are actively using.

→ Here’s how to earn by contributing compute on Lium (SN51).