Nebius Pulls $3.5 Billion from Citi to Accelerate GPU and Data-Center Buildout

On October 22, 2026, Nebius announced it had secured a $3.5 billion financing facility from Citi. The money is earmarked for data-center development and GPU purchases — the two inputs that define capacity in the AI cloud business right now. That use-of-proceeds statement is not boilerplate. It is a capital allocation decision made explicit, and it signals where Nebius believes the constraint on its growth actually sits.

The Structure Is the Story

The facility matures in 364 days — one day short of the threshold that would typically trigger additional regulatory and disclosure requirements for some debt instruments. That is a deliberate choice, not a round number. At an initial rate of 2.75 percentage points over the term SOFR benchmark, the cost of capital is meaningful but not punishing for a company moving fast in a market where idle GPU capacity is the real expense. Short-dated, floating-rate — this is bridge money, not a long-term balance-sheet commitment. Nebius is buying optionality, not locking in a capital structure.

The company was explicit about that framing: it said the facility increases flexibility for additional fundraising supporting growth in 2027 and beyond. Read that carefully. Nebius is not presenting this as the financing — it is presenting it as the financing that makes the next financing possible on better terms. Infrastructure deployed now generates the revenue history and utilization metrics that underpin equity or debt raises later. The $3.5 billion is an accelerant, not a destination.

Multiple Funding Channels Running in Parallel

Nebius also cited customer prepayments and operating cash flow as concurrent funding sources alongside the Citi facility. That detail matters. Prepayments from customers mean counterparties are willing to commit capital ahead of delivery — a signal of demand confidence and, practically, a form of zero-cost debt. Operating cash flow means the existing business is generating enough to contribute to expansion, not just sustain itself. Stacking these three channels — the facility, prepayments, and operating cash — reduces dependence on any single source and preserves negotiating leverage for the next round.

This is a capital strategy built around avoiding concentration risk in the funding stack, which is a relatively sophisticated posture for a company in scale-up mode. It also suggests Nebius's finance team is anticipating a tighter or more volatile fundraising environment in 2027 and wants multiple entry points rather than one large door.

What This Says About the Infrastructure Moment

The $3.5 billion number lands in a market where data-center and GPU financing has become a distinct asset class almost overnight. The fact that Citi is the sole named counterparty on a facility this size — and that Nebius is drawing on it inside a 364-day window — reflects how quickly large financial institutions have reoriented around AI infrastructure as a fundable category. A year ago, this kind of deal would have been a consortium. That it isn't suggests either that Nebius's credit profile and revenue visibility are strong enough to command a single lender, or that Citi moved aggressively to own the relationship. Likely both.

The deeper shift here is structural: AI cloud providers are increasingly being financed like utilities or telecoms — capital-intensive businesses with long asset lives and (ideally) contracted revenue — rather than like software companies. The short maturity on this facility is an exception to that pattern, not a contradiction of it. Nebius is using short-term debt to build long-term assets, betting that by the time the facility matures, the assets will support a more permanent capital structure.

If that bet is right, the $3.5 billion becomes the seed round for a much larger 2027 balance sheet. If the GPU buildout stalls or customer demand softens, the 364-day clock becomes a different kind of pressure entirely.