Nebius completed its first Madrid deployment in one month, bringing Nvidia RTX 6000 GPUs online as part of a European expansion. The facility includes B300 systems and two additional deployment phases scheduled for the coming weeks. The company also secured five first place results in the latest MLPerf AI inference benchmarks, including a full rack result using the Nvidia GB300 NVL72 platform.

The buildout supports a capacity roadmap starting at 0.2 gigawatts and 200,000 H100 equivalent GPUs, rising to a final target of 5 gigawatts and 4 million GPUs. As part of its early hardware access, the provider ran the DeepSeek R1 model at approximately 603,000 tokens per second across 72 GPUs using the next generation Vera Rubin NVL72 platform.