Nvidia announced this week that it is investing $2 billion in Amsterdam-based Nebius, taking an 8.3% stake in the company as part of a deeper push into the neocloud layer of the AI infrastructure stack. Nebius, in turn, has committed to deploying more than 5 gigawatts of data center capacity by 2030. The numbers are large enough to sound abstract. The strategic logic is not abstract at all.
The neocloud concept — cloud infrastructure built specifically for AI workloads, sitting between the hyperscalers and enterprise buyers — has been one of the more significant structural developments in the industry over the past 18 months. Companies like CoreWeave, Lambda Labs, and now Nebius are building GPU-dense compute clusters for customers who need more than AWS and Azure offer but cannot justify the capital expenditure of building their own infrastructure. The demand is real: AI model training and inference at scale requires a level of hardware access and cluster configuration that the general-purpose hyperscaler model was not designed to provide efficiently.
Comments