GMO, NTT and QTnet complete remote distributed AI infrastructure trial between Tokyo and Fukuoka using IOWN APN

GMO Internet, NTT East, NTT West and QTnet completed a proof-of-concept for a remote distributed AI infrastructure between Tokyo and Fukuoka using IOWN APN.

They placed storage in Tokyo and GPUs in Fukuoka, connecting sites with a 100GbE IOWN APN link and tested ResNet and Llama2 70B training tasks.

Remote LLM training took 24.99 minutes versus 24.87 minutes locally (about 0.5% difference), and image classification times were similarly close.

The partners propose hybrid operations combining on-premises data and remote cloud GPUs and aim to promote IOWN APN as an AI backbone.