The Nvidia-backed cloud operator has secured $3.5B in financing arranged with Citi, adding debt to the balance sheet as it races to convert demand for accelerated computing into working data halls.
Nebius sits in the neocloud tier, renting GPU capacity to AI developers that cannot get the allocations they want from the hyperscalers. Those contracts are typically signed years ahead, which makes them bankable collateral for lenders willing to underwrite construction and hardware.
Debt has become the preferred fuel for this layer of the market, because the capital need is enormous and equity is expensive at current valuations. Lenders are effectively financing servers and shells against signed customer commitments.
The financing also illustrates how the AI buildout is being underwritten by institutions that have no direct exposure to models or software. Banks are treating compute capacity more like infrastructure, with the characteristics of a utility financing.
For Nebius, the constraint has rarely been demand. The harder problems are power, grid connections and the lead time on chips, all of which stretch the gap between a booked contract and a billable cluster.
The company did not disclose the interest rate, covenant terms or how the financing is split between equipment and construction.
