Artificial intelligence infrastructure provider Nebius Group NV today announced the launch of its first graphics processing unit clusters in the U.S. with a deployment in Kansas City, Missouri.

Nebius was previously known as Yandex NV, the Dutch holding company of the prominent Russian technology company of the same name. Yandex sold its Russian assets to a consortium of Russian investors for $5.4 billion in July, retaining its international businesses and rebranding as Nebius, allowing the company to focus on AI infrastructure and services independently of its former Russian operations.

The new Kansas City cluster is scheduled to go live in the first quarter of 2025 and will house thousands of Nvidia Corp. GPUs, primarily H200 Tensor Core GPUs, in the initial phase, with Nvidia Blackwell chips arriving later in 2025. The colocation has also been designed to be expanded from an initial 5 megawatts up to 40 megawatts, or about 35,000 GPUs at full potential capacity.

Nebius says it’s actively expanding its presence in the U.S. as part of its strategy to become a leading provider of AI infrastructure to AI builders globally. The company is also in advanced discussions for a second, larger-scale GPU cluster in the U.S., also slated to come online in 2025.

In addition, Nebius has also opened two new customer-facing hubs in San Francisco and Dallas, with a third office set to open in New York later this year.

“Our first GPU cluster and offices represent a pivotal step in our expansion in the U.S. market,” said Arkady Volozh, founder and chief executive officer of Nebius. “Serving American customers from American facilities means low latency and maximizes the advantages of our AI-native cloud.”

“We will be building more fully owned GPU clusters across the US to meet [the] exploding demand for high-quality AI infrastructure from U.S. AI developers and enterprises,” Volozh added.

Nebius’ full-stack AI infrastructure is being purpose-built to meet the demands of the global AI industry and leans on deep technical expertise across hardware and software, cloud engineering and machine learning. The company’s AI-native cloud is designed to manage the full machine learning lifecycle – from data processing and training to fine-tuning and inference – all in one place.

A recently launched product, Nebius AI Studio, expanded the company’s offering to app builders, with access to a range of state-of-the-art open-source models in a flexible, user-friendly environment at among the lowest price-per-token on the market.

Photo: Nebius