Nebius is moving part of its AI cloud expansion to partner-owned data centres, a shift meant to add compute capacity without funding every new facility itself.

AI Business reported that the Dutch neocloud plans to run its platform inside infrastructure owned and operated by partners under an asset-light data centre model.

Nebius Asset-Light Data Centre Model Moves Facility Funding To Partners

Under the new strategy, partners finance and own the infrastructure and hardware and operate the data centre facility.

The cloud provider keeps responsibility for system architecture, hardware design and the software stack, and it also pledged to find buyers for the resulting capacity.

In a press release, Arkady Volozh, chief executive, said the company was prepared to work with "data centre investors, regional partners and others with capacity or capital."

Revenue Shares And Committed Capacity Are Under Discussion

The model is not limited to one commercial structure.

Implementation talks have started with several partners, and the options under consideration include revenue-sharing agreements, licensing fees, commissions and committed-capacity deals.

Committed-capacity arrangements would give the company access to specified amounts of compute that it could sell on to customers.

The same service-standard promise applies across company-owned infrastructure and partner sites.

Volozh described the model as a way for infrastructure partners to benefit from AI growth.

He also said the software would let partners reach a wider customer base with better margins than conventional wholesale bare-metal contracts.

Meta And Microsoft Contracts Set The Capacity Context

AI Business listed Meta's $27 billion March capacity deal and Microsoft's $19.4 billion New Jersey capacity agreement as the company's biggest contracts to date.

Those contracts provide capacity demand evidence, but the partner model still leaves commercial details outside the public record.

The company has not named the infrastructure partners, first partner sites, committed-capacity volumes, facility timelines or customer allocations for the asset-light model.