Case Study: Linum achieves 25% lower AI video model training costs with Crusoe Cloud

Linum cuts AI video training costs by 25% with Crusoe

Linum, a small AI lab founded by a two-person team, faced a significant challenge in training a large-scale text-to-video model from scratch. The sheer density of video data and the associated costs of transferring petabytes of information were major constraints. Their strategy of streaming data live during training was hampered by poor internet uplink speeds at other cloud providers, which created bottlenecks and made the process prohibitively expensive and slow.

Crusoe provided its cloud computing platform, which featured a high-performance AI-first backbone network with superior internet uplink speeds. This solution eliminated the data-loading bottlenecks Linum experienced elsewhere. By implementing Crusoe Cloud, Linum achieved 25% lower training costs due to faster networking and data throughput. The Crusoe team also provided crucial hands-on engineering support throughout the process, enabling the successful training of their 2 billion parameter model.

Linum