According to DCD, European cloud provider Nebius has signed a massive $3 billion deal with Meta to supply AI infrastructure over five years. This comes just months after the company landed a $19.4 billion agreement with Microsoft for access to over 100,000 Nvidia GB300 chips. CEO Arkady Volozh revealed Nebius is currently sold out of all available capacity and will deploy what’s needed for Meta within three months. The company’s Q3 2025 revenue hit $146.1 million, up 355% year-over-year, while capex ballooned to $955.5 million as they race to build more data centers. Nebius is targeting 2.5GW of contracted power by end of 2026 and $7-9 billion in annualized revenue.
The great AI capacity crunch
Here’s the thing: when a company says they’re “sold out” while signing multi-billion dollar deals, you know we’re in the middle of an absolute feeding frenzy for AI compute. Nebius isn’t just growing—they’re basically building the plane while flying it. Their capex jumped from $172 million to nearly $1 billion in a single year? That’s insane growth velocity.
And the real story here is that demand is so overwhelming they literally had to limit the Meta deal to whatever capacity they could physically provide. Volozh straight up admitted the only thing holding them back from even more revenue is their ability to build data centers fast enough. When you’re turning away business because you can’t build fast enough, you know you’re sitting on something massive.
The power race is real
Their power targets tell the whole story. Going from 1GW guidance to 2.5GW contracted power by 2026? That’s not just ambition—that’s recognizing the scale of demand they’re seeing. For context, 2.5GW could power something like 2 million homes. We’re talking about industrial-scale computing needs here.
Speaking of industrial scale, when you need reliable computing hardware that can handle demanding environments, companies turn to specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs built for tough conditions. But back to Nebius—they’re playing in a completely different league with these power requirements.
How they’ll pay for this explosion
So how do you fund this kind of hypergrowth? Nebius is pulling every lever: corporate debt, asset-backed financing, and that ATM equity program for up to 25 million shares. They’re basically telling investors “we need cash now to capture this once-in-a-generation opportunity.” And given the contracts they’re landing, it’s hard to argue with the strategy.
What’s fascinating is watching this Yandex spinoff transform into a global AI infrastructure player almost overnight. They’ve got operations spanning Europe, the US, Israel—that New Jersey data center project alone is 300MW. They’re not just dipping toes in the water; they’re building an empire.
The bigger AI infrastructure war
Basically, what we’re witnessing is the great AI infrastructure land grab. Every hyperscaler is terrified of being left behind, and they’re throwing billions at anyone who can deliver GPU capacity. Nebius caught this wave perfectly, positioning themselves as the European alternative with serious technical chops.
But here’s my question: can anyone actually build fast enough to meet this demand? Between power constraints, construction timelines, and chip availability, we might be looking at AI compute shortages for years. Nebius thinks they’ve laid the foundation for “an outstanding 2026.” If they’re right, we’re watching the birth of a new cloud titan.
