The Dutch AI cloud provider Nebul is expanding into Belgium. A SuperCluster featuring more than 1,000 Nvidia Blackwell Ultra GPUs housed in rack-scale GB300 NVL72 systems will be built in Machelen.
The new cluster is being built on the Datacenter United campus in Machelen, a short distance from Brussels. Lenovo is the server supplier, while Datacenter United is providing power, connectivity, and the substantial liquid cooling required. The procurement of the hardware has already begun. Nebul has not yet disclosed when the cluster will actually go live.
The GPUs will be housed in Nvidia GB300 NVL72 systems. These are rack-scale systems that function as a single unit; they represent the standard on which modern AI models are trained and run, offering great flexibility for scaling per rack and optimal utilization of GPU computing power.
Nebul, based in Leiden, positions itself as a sovereign AI provider, with computing power that remains fully compliant with European legislation. So far, the company has relied on Dutch data center partners such as Eurofiber, NorthC, and Switch Datacenters. Earlier this year, we described how Nebul built its AI factory, with NetApp serving as the storage layer. Even then, the focus was on private AI inference rather than model training, an area few companies pursue due to the high costs. The company also began using a Supermicro HGX B200 SuperCluster.
Machelen offers a ready-to-use location
Within Europe, Nebul considers Machelen, Belgium, an attractive location for the new computing power. The campus is ISO-certified, is located close to the Brussels internet hub, and already runs mission-critical workloads for European institutions. In addition, the location offers a low-latency connection to the Netherlands. That works out well, since that’s where the rest of Nebul’s capacity is located.
According to CEO Arnold Juffer, there is no shortage of customers. “The demand is there, and it exceeds the capacity that Europe is currently building,” he says.
More regions on the way
Machelen is therefore just one location where Nebul plans to expand. The company states that additional European regions and capacity are already in the works.
Nebul wants to do more than just host AI hardware. Last week, the company announced a partnership with DDN, in addition to its existing storage partners NetApp and VAST Data. DDN Infinia makes the KV cache available across the entire GPU fleet. According to both parties, this results in up to five times faster token responses and nearly 60 percent lower costs for reloading context.