Meta's $27B Bet: Who Really Wins the AI Infrastructure Race?

Meta is spending more on AI infrastructure this year than NASA's entire budget since its founding. So when the company quietly signs a $27 billion deal with a Dutch cloud firm most people have never heard of, it's worth asking: what exactly is being built—and who profits?

The Deal: What Just Happened

On Monday, Meta announced a long-term agreement with Nebius, a Netherlands-based AI cloud provider, to purchase up to $27 billion in computing infrastructure over five years. The contract breaks into two parts: $12 billion in dedicated capacity across multiple locations, and up to $15 billion in additional on-demand compute that Meta can draw on as needed.

The dedicated capacity includes what Nebius describes as one of the first large-scale deployments of Nvidia's newest AI chip, the Vera Rubin. That detail matters—access to cutting-edge silicon, at scale, before most competitors can get it, is increasingly the moat in the AI arms race.

Nebius shares jumped 14% in pre-market trading following the announcement. That reaction is telling, but it's also the latest in a string of surges: the stock has climbed more than 400% since the company listed in New York in 2024, rising over 200% in 2025 alone and another 35% year-to-date in 2026.

Who Is Nebius, Exactly?

Nebius isn't a household name, but its backstory is unusual. The company was carved out of Yandex—Russia's dominant search and tech giant—in 2022, after the war in Ukraine prompted a restructuring of Yandex's international operations. The entity that emerged, led by founder and CEO Arkady Volozh, rebranded and listed on Nasdaq in 2024, positioning itself as a European-headquartered AI cloud specialist.

Sponsored

Advertise with Us

[email protected]

Sponsored

Since then, it has moved fast. Last September, Nebius signed a deal with Microsoft worth up to $19.4 billion over five years. Last week, Nvidia announced a $2 billion investment in the company—sending shares up another 16%. Now Meta. The pattern suggests Nebius has successfully positioned itself as a neutral, high-capacity AI infrastructure partner that the biggest players in the world are willing to commit to at scale.

The Bigger Spending Context

Meta's deal doesn't exist in a vacuum. The company has guided for AI-related capital expenditure of between $115 billion and $135 billion in 2026 alone. Across the major hyperscalers—Amazon, Alphabet, Microsoft, and Meta—total AI infrastructure spending this year is projected to reach roughly $700 billion.

To put that in perspective: $700 billion is larger than the GDP of Switzerland. It's being deployed in a single year, into a single technology category.

The investment thesis driving this is straightforward: whoever controls the compute infrastructure controls the pace of AI development. Long-term, large-scale contracts like the Meta-Nebius deal are a way to lock in capacity before it becomes scarce—and before prices rise further. Last week, UK-based AI data center startup Nscale raised $2 billion at a $14.6 billion valuation, backed partly by Nvidia, signaling that investor appetite for AI infrastructure extends well beyond the hyperscalers themselves.

Winners, Losers, and the Unanswered Questions

The obvious winners are Nebius shareholders, Nvidia (whose chips underpin nearly all of this infrastructure), and the broader AI hardware supply chain. For investors, the question is how much of this upside is already priced in—Nebius at +400% since IPO is not a discovery play.

For enterprise customers and smaller AI developers, the picture is more complicated. As hyperscalers lock up long-term capacity through exclusive or semi-exclusive deals, available compute for everyone else could tighten. Cloud pricing for AI workloads has been volatile; sustained demand from the biggest buyers could push costs higher for those without the leverage to negotiate Meta-scale contracts.

Regulators in the EU and US have begun scrutinizing the concentration of AI infrastructure in a handful of companies. The Nebius deal—a European-headquartered firm supplying American hyperscalers—adds an interesting geopolitical wrinkle. Is this the kind of transatlantic tech partnership Brussels wants to encourage? Or does it raise questions about where strategic AI capacity ultimately resides?