Meta has signed a new long-term AI infrastructure agreement with Dutch neocloud operator Nebius Group worth up to $27 billion over five years, in one of the most significant AI compute deals announced to date.

The deal, disclosed on 16 March 2026, dramatically expands an existing commercial relationship between the two companies, and arrives just five days after Nvidia announced a $2 billion strategic investment in Nebius as part of a separate partnership.

The agreement is structured in two parts. Under the first, Nebius will provide $12 billion of dedicated compute capacity across multiple locations, starting in early 2027.

That infrastructure will be built around what Nebius describes as one of the first large-scale deployments of Nvidia's Vera Rubin platform, the chip maker's latest generation of AI-specialist accelerators.

Under the second part of the deal, Meta has committed to purchase additional available capacity from certain upcoming Nebius clusters up to a total of $15 billion over the same five-year period. Nebius says it intends to sell that capacity to third-party customers first, with any remaining capacity going to Meta.

“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,”  said Arkady Volozh, founder and CEO, Nebius

Nebius shares jumped 14% in pre-market trading on the news. The company said its financial guidance for 2026 remains unchanged, noting that the delivery of capacity under the new agreement does not begin until early 2027, meaning the financial impact will materialise primarily beyond the current fiscal year.

Expanding an existing relationship

The two companies have history. Meta signed an initial $3 billion five-year agreement with Nebius in November 2025, the existence of which Volozh disclosed to shareholders alongside the company's third-quarter earnings. Today's announcement represents a ninefold expansion of that commitment, taking the combined value of Meta's contracted spend with Nebius to $30 billion.

Meta has made AI its defining strategic priority. The company has indicated it plans to direct as much as $135 billion in capital expenditure toward AI-related projects in 2026 alone, funded largely by profits from its advertising business. Mark Zuckerberg has also previously stated Meta plans to invest $600 billion in US infrastructure by 2028.

The scale of AI spending among the major technology companies means firms like Nebius, which can provide specialised, large-scale compute capacity without the overhead of a full hyperscaler, are becoming increasingly central to how the industry is being built out. Major technology companies collectively are expected to invest around $650 billion in data centre and AI equipment in 2026.

Nebius is an unusual company. Nasdaq-listed and headquartered in Amsterdam, it is the international successor to Yandex NV — once known as the Google of Russia, after Volozh sold all Russian assets to a Kremlin-linked consortium in July 2024 for approximately $2.4 billion, retaining the cloud and data centre operations outside Russia under the new Nebius name.

Trading on Nasdaq, which had been suspended since February 2022 following Russia's invasion of Ukraine, resumed in October 2024.

Volozh, who co-founded Yandex in the 1990s and ran it as CEO for more than two decades, was placed on EU sanctions in 2022 but removed from the list in early 2024 after publicly condemning Russia's invasion.

The company he has built since is explicitly positioned as a neocloud, a newer class of AI-native cloud provider building data centres designed from the ground up for GPU-intensive AI workloads, rather than adapting general-purpose infrastructure after the fact. Nebius has data centres in Finland, France, and the United States, with a first gigawatt-scale AI factory recently approved and further expansion across multiple locations under way.

The company has been accumulating major clients at speed. In September 2025, it announced a deal with Microsoft worth an initial $17.4 billion with potential to reach $19.4 billion.

The Nvidia strategic investment brought in $2 billion and included a partnership agreement covering AI factory design, inference stack development, and early access to successive generations of Nvidia compute architecture.

Nebius says it intends to deploy more than five gigawatts of Nvidia systems by the end of 2030, though analysts have noted that figure represents a target rather than a contractual commitment. Nebius has also previously signed infrastructure agreements with AWS and other cloud customers, alongside serving AI startups and enterprises through its standard cloud business.

The Vera Rubin dimension

The involvement of Nvidia's Vera Rubin platform adds a further dimension to the announcement. Vera Rubin is Nvidia's next-generation AI chip architecture, succeeding the Blackwell platform. Its inclusion in a large-scale commercial deployment with Meta is a signal both of Nebius's priority access to cutting-edge Nvidia hardware, a competitive advantage rooted partly in Yandex's historic status as one of Nvidia's largest customers outside the US and China, and of Meta's desire to be operating at the frontier of available compute, rather than working with the previous generation of chips.

Volozh has previously described Nebius's access to Nvidia's latest hardware as a structural advantage for the company: an AI cloud built specifically for the technology rather than one adapted from a legacy platform. Whether that advantage persists as more hyperscalers and neoclouds compete for the same chips is a question investors will be watching closely.