Artificial intelligence (AI) infrastructure company Nebius has launched its Nebius AI cloud offering.

The cloud is built using Nvidia GPUs, including the Nvidia H100 and H200 Tensor Core GPUs, and L40S GPUs. The company expects to add the Nvidia GB200 NVL72 platform soon.

Nebius' cloud has storage speeds of up to 100 GBps and 1m IOPS (input/output operations per second), and can handle large data sets, perform model training and share data across distributed nodes.

The platform also features managed Apache Spark through which Nebius provides fully managed services.

The platform's virtual machines come pre-configured with AI libraries and drivers meaning users can immediately deploy their workloads.

Roman Chernin, co-founder and chief business officer at Nebius, said: "The AI industry is changing incredibly fast, and so are the needs of AI practitioners. We spent months listening to our customers, and what they told us is that they need flexibility on capacity, they want real self-service access, and they need more than just basic infrastructure. This is what we have built, all in one place."

Andrey Korolenko, co-founder and chief product and infrastructure officer at Nebius, added: "Over the past year we have written a new code base to create a fully owned cloud offering specifically for AI. This is a true full-stack AI platform: a fully owned network of large-scale Nvidia InfiniBand-interconnected GPU clusters built to the Nvidia reference architecture, with a proprietary cloud platform on top including a suite of managed services, developer tools, and applications."

Nebius' origins can be found in tech giant Yandex. While the holding company is based in the Netherlands, it provided several services in Russia including a search engine, ride-hailing, and food delivery services, and was seen as a national strategic asset by the Kremlin.

After the war broke out in Ukraine, Yandex began looking for an exit strategy from Russia and in February 2024 sold off its Russian assets to a consortium of investors, while retaining Nebius and a data center in Finland.

In September of this year, Nebius deployed a Nvidia-based AI cluster in an Equinix data center in Paris, France. Earlier this month, the company announced it would be tripling the capacity of its Finland data center, adding more than 60,000 GPUs at the site.