Meta Just Handed Nebius $27 Billion for AI Compute

Nebius Group dropped a bombshell this morning that has Wall Street buzzing and the AI supply chain shifting under everyone's feet.

The Dutch AI cloud outfit just sealed a five-year deal worth up to $27 billion with Meta to supply dedicated infrastructure, including $12 billion worth of capacity built around one of the very first big rollouts of NVIDIA's Vera Rubin platform.

Credit: Nebius Group NV

Delivery kicks off in early 2027 across multiple sites, with Meta also locking in options for another $15 billion in extra capacity from Nebius clusters that will otherwise go to third-party customers.

According to Nebius' official announcement released today, the structure is straightforward but massive in scale: Meta gets priority access to the Vera Rubin racks while Nebius keeps building out its AI cloud business at full throttle.

This is no small add-on either.

It builds directly on the $3 billion deal the two companies struck late last year, turning what started as a test run into a full-blown long-term commitment.

Arkady Volozh, founder and CEO of Nebius, put it plainly in the release:

“We are pleased to expand our significant partnership with Meta as part of securing more large, long-term capacity contracts to accelerate the build-out and growth of our core AI cloud business. We will continue to deliver.”

Meta, for its part, is in full-on spending mode to chase frontier-model leadership.

Bloomberg noted the hyperscaler will pay as much as $27 billion over the next five years for this slice of capacity as it pours money into competing with the biggest players in the space.

Everyone loves to talk about Meta's sky-high capex numbers and Zuckerberg's obsession with owning the stack.

Yet here we are, with Meta turning to a relatively young European neocloud for next-gen NVIDIA silicon it apparently can't secure or deploy fast enough on its own.

Nebius didn't stumble into this.

It positioned itself early with NVIDIA, sold out existing capacity to Microsoft and others, and now gets to ride the Vera Rubin wave before most hyperscalers even line up.

This isn't Meta innovating harder. This is Meta admitting the build timeline for its own data centers can't keep pace with the training demands of whatever comes after Llama 4.

Nebius, headquartered in Amsterdam and listed on Nasdaq, just proved that independent AI clouds with sharp NVIDIA relationships can grab serious chunks of the pie that used to stay locked inside the big tech fortresses.

Users will feel the upside soon enough. Faster access to beefed-up Llama models and Meta AI features that actually work at scale without the usual rollout hiccups.

The infrastructure side, though, looks messier. Meta is essentially outsourcing critical training muscle, which could bite if power delays or supply hiccups hit Nebius' expansion plans.

Meanwhile Nebius guidance for 2026 stays unchanged, meaning the real revenue pop from this lands in 2027 and beyond, exactly when the industry expects the next compute crunch.

The bigger picture is brutal for anyone still pretending vertical integration wins everything.

Even Meta, with its endless cash and engineering talent, needs backup from specialists like Nebius to stay competitive.

This Nebius Meta partnership deal doesn't just pad Nebius' numbers. It rewrites the power map in AI infrastructure, handing leverage to anyone who can deliver the latest NVIDIA gear on time.

And that leaves the industry with one clear takeaway: the solo hyperscaler era is over.

Meta just proved it by writing a $27 billion check to someone else.