The cluster, located in Kansas City, Missouri, is expected to go live in Q1 2025 and will house “thousands” of Nvidia GPUs, primarily H200s, with additional support from Blackwell units expected later in the year.
Nebius said its initial phase will start at 5 MW before scaling up to 40 MW or around 35 thousand GPUs at full potential capacity.
Subscribe today for free
“Our first GPU cluster in the US and new offices represent a pivotal step in our expansion in the US market,” said Arkady Volozh, founder and CEO of Nebius. “Serving American customers from American facilities means lower latency and maximises the advantages of our AI-native cloud.”
Headquartered in Amsterdam, Nebius operates data centres across Europe but now has eyes for the US.
Beyond the Kansas City site, Nebius has opened two customer-facing hubs in San Francisco and Dallas, with a third office set to open in New York later this year.
Nebius’ US expansion follows its pledge to invest more than $1 billion in AI infrastructure by mid-2025 to support AI-native cloud offerings to its global customers.
“We will be building out more GPU clusters across the US to meet exploding demand for high-quality AI infrastructure from US AI developers and enterprises,” Volozh added.
The company has already invested $200 million since the beginning of the year and is currently constructing two new data centres in France while also expanding its Finnish site in Mäntsälä.
RELATED STORIES
Nebius to triple capacity at Finnish data centre to support AI workloads
Nebius continues $1bn AI infrastructure spend with new GPU cluster in Paris
