How are neocloud companies like CoreWeave and Nebius, which frequently appear in the stock market these days, actually making money? As AI demand explodes, traditional hyperscalers like AWS, Google, and Microsoft have struggled to supply GPUs and computing resources on time, and neoclouds that quickly secure and rent out the latest NVIDIA GPUs are stepping in to fill that gap.

In particular, CoreWeave has rapidly expanded its business using a somewhat risky-looking model: raising funds from the financial sector based on its owned GPUs and long-term contracts with major clients like Microsoft, OpenAI, and Meta, and then using that money to buy the latest GPUs all over again. NVIDIA is also supplying GPUs to neoclouds while directly investing in them, creating a circular financing structure surrounding AI infrastructure. However, there is no guarantee that this growth formula will last. This is because the center of demand for AI computing is shifting from model training to inference, and traditional cloud companies are developing their own AI accelerators, which could potentially lower their reliance on NVIDIA GPUs. In today's O-Graph, we examined the stories of neoclouds and the cloud through five graphs.

Reported by An Hyemin | Filming: Hwang Se-hoe, Park Woo-jin | Editing: Lee Ki-eun | Design: Ahn Jun-seok | Intern: Shin Yeon-seong | Produced by Knowledge Contents IP Team