GMI secures ₩900 billion investment from NVIDIA and others
October 3, 2026 · 2 min read
Also published in Svenska, Polski, Português (Brasil)
GMI, a provider of GPU-based cloud services, successfully raised a total of 900 billion KRW (approximately 668 million USD) in funding from investors including NVIDIA. The company announced this Series B funding on October 30, 2026.
Of the total amount, 223 billion KRW is equity investment led by ARCHIV, a venture capital firm focused on AI and robotics, with participation from NVIDIA. The remaining 445 billion KRW was secured as a credit facility led by CTBC Bank from Taiwan.
Domestic investors reportedly include KB Investment, DSC Investment, Kyobo Life, and KT. Although the company's valuation has not been disclosed, it has indicated that its annual recurring revenue (ARR) exceeds 810 billion KRW (approximately 600 million USD).
GMI is one of seven cloud providers building data centers according to NVIDIA's specifications, utilizing a full technology stack including not only GPUs but also networking and software. The company is also involved in NVIDIA's AI computing partnership, receiving credit support for facility construction in exchange for sharing some revenue generated from its cloud services.
The newly acquired funds will be used to expand AI computing infrastructure in the United States and Asia. GMI supports AI model training, large-scale inference, and deployment of AI agents on a single platform using the latest GPU clusters. CEO Alex Ye stated that the company's AI inference processing volume has surged to 1 trillion tokens per day, approximately 20 times higher than six months ago.
Key clients include OpenRouter, FireworksAI, and Hicksfield. Although Chinese clients currently account for just under 30% of total revenue, Ye predicts this will drop to around 10% next year. GMI is rapidly expanding its presence in the AI infrastructure market, especially by providing training and inference infrastructure using NVIDIA's latest GPUs to AI startups and companies that find it challenging to build large-scale data centers.