Nebius Pursues $3.75B Raise After Meta and NVIDIA Partnerships

The AI infrastructure gold rush has a new contender making an audacious play. Nebius Group, the Amsterdam-headquartered neocloud provider that emerged from the ashes of Yandex's Russian exit less than two years ago, is raising $3.75 billion in convertible debt to finance what may be the most ambitious data center expansion in the sector. The announcement, made on March 17, sent shares tumbling more than 10% — even as the company simultaneously revealed two of the largest partnerships in cloud computing history [1].

The capital raise arrived one day after Nebius disclosed a five-year infrastructure agreement with Meta Platforms valued at up to $27 billion, and less than a week after NVIDIA announced a $2 billion strategic investment in the company [2][3]. Together, these deals catapult Nebius from a niche GPU cloud startup into a central node in the global AI supply chain — and raise fundamental questions about whether the neocloud model can scale fast enough to meet Silicon Valley's insatiable appetite for compute.

From Yandex Castoff to AI Powerhouse

Nebius's origin story is unlike anything else in tech. The company traces its roots to Yandex, Russia's dominant search engine and technology conglomerate. In July 2024, Yandex N.V. sold all of its Russian assets to a consortium of local investors for $5.2 billion — one of the largest corporate exits from Russia following the invasion of Ukraine — and retained a portfolio of international businesses focused on AI [4].

Under CEO Arkady Volozh, the Yandex co-founder who had stepped aside amid Western sanctions before returning to lead the restructured entity, the company rebranded as Nebius Group and resumed trading on the NASDAQ in October 2024. At the time, the stock hovered around $18-20 per share, giving the company a modest market capitalization. By March 2026, shares had climbed above $140 at their peak, valuing the company at nearly $29 billion [5].

The transformation has been powered by a singular bet: that the world's largest AI companies need dedicated, purpose-built cloud infrastructure that traditional hyperscalers like Amazon Web Services and Microsoft Azure are not optimized to provide. Nebius operates what the industry calls a "neocloud" — a cloud platform that specializes exclusively in the AI lifecycle, from training massive language models to running inference workloads, built around high-performance GPU clusters connected by InfiniBand networking [6].

Source: Nebius Group Earnings Reports

Data as of Mar 17, 2026CSV

The Meta Deal: $27 Billion in Five Years

The centerpiece of Nebius's transformation is its agreement with Meta Platforms, announced March 16 [2]. Under the five-year deal, Nebius will provide $12 billion of dedicated AI compute capacity across multiple locations, starting in early 2027, built around one of the first large-scale deployments of NVIDIA's next-generation Vera Rubin accelerator platform. Meta has also committed to purchasing up to $15 billion of additional compute capacity from Nebius clusters as new infrastructure comes online.

The deal represents a ninefold expansion of an initial $3 billion agreement Meta signed with Nebius in November 2025, taking the combined value of Meta's contracted spend to $30 billion [7]. For context, Meta's total capital expenditure guidance for 2026 is $115 billion to $135 billion — meaning its Nebius commitment alone could represent roughly 5% of annual spending over the contract's life [8].

The scale of the agreement places Nebius alongside CoreWeave, the other major neocloud that has secured massive hyperscaler contracts. CoreWeave, which went public in March 2025 at a $23 billion valuation and has since seen its market cap grow substantially, reported a $66 billion revenue backlog and projected $12-13 billion in 2026 revenue [9]. Meta has deals with both companies, suggesting the social media giant is deliberately diversifying its AI compute supply chain beyond the traditional Big Three cloud providers.

NVIDIA's $2 Billion Vote of Confidence

Five days before the Meta announcement, NVIDIA disclosed a $2 billion strategic investment in Nebius as part of a sweeping partnership agreement covering four areas: AI factory design, inference optimization, multi-generation infrastructure deployment, and fleet management [3].

"AI is at another inflection point — agentic AI, driving incredible compute demand and accelerating infrastructure buildout," NVIDIA CEO Jensen Huang said in the announcement. Volozh responded that the partnership would extend Nebius's capabilities "from gigawatt-scale AI factories to inference and software" [3].

The partnership calls for Nebius to deploy more than 5 gigawatts of NVIDIA capacity by the end of 2030 — a staggering figure that would make Nebius one of the largest single consumers of NVIDIA's data center hardware globally. The investment also gives NVIDIA a financial stake in Nebius's success, aligning incentives as both companies navigate the massive build-out ahead.

The $3.75 Billion Gamble

Against this backdrop, Nebius's decision to raise $3.75 billion in convertible notes is both logical and sobering [1]. The offering, structured as a private placement to qualified institutional buyers under Rule 144A, comprises two tranches: $2 billion of notes due March 2031 and $1.75 billion due March 2033. The company also granted initial purchasers an over-allotment option for an additional $562.5 million.

The notes will accrete to 120% of their original principal amount at maturity — meaning investors are effectively getting built-in returns even before any conversion to equity. The notes are redeemable beginning in 2029 (for the 2031 series) and 2030 (for the 2033 series), but only if Nebius's stock price reaches 130% of the conversion price multiplied by the applicable accretion ratio [10].

Nebius has been explicit about where the money is going: data center construction, expansion of its physical footprint, GPU procurement, and further development of its AI cloud platform. The company guided for $16 billion to $20 billion in capital expenditure for 2026 alone — a figure that dwarfs both its 2025 revenue of $529.8 million and its $3.68 billion cash position at year-end [5][11].

This is the core tension: Nebius must spend tens of billions to fulfill contracts worth tens of billions, in a business where the infrastructure must be built before the revenue arrives. The $3.75 billion raise, combined with the September 2025 convertible offering that raised approximately $2.75 billion and NVIDIA's $2 billion investment, gives Nebius a substantial war chest — but the capex guidance suggests it will need even more [12].

Source: Company earnings guidance / Yahoo Finance

Data as of Mar 17, 2026CSV

The Neocloud Boom — and Its Risks

Nebius is riding a wave that extends far beyond any single company. The five largest U.S. cloud and AI infrastructure providers — Microsoft, Alphabet, Amazon, Meta, and Oracle — have collectively committed to spending between $660 billion and $690 billion on capital expenditure in 2026, nearly doubling 2025 levels [8]. Goldman Sachs has projected that AI companies may invest more than $500 billion this year alone [13].

This spending supercycle has created enormous demand for the services that neoclouds provide. Unlike traditional cloud platforms, which offer a broad range of general-purpose services, neoclouds like Nebius, CoreWeave, and Lambda Labs focus exclusively on high-performance AI compute. Their advantage lies in purpose-built infrastructure: dense GPU clusters, InfiniBand networking, and software stacks optimized for training and inference workloads [6].

But the model carries significant risks. Neoclouds are essentially building expensive physical infrastructure against long-term contracts, in an industry where technological generations turn over every 18-24 months. If AI demand plateaus, if a new chip architecture disrupts NVIDIA's dominance, or if hyperscalers decide to bring more compute in-house, companies like Nebius could find themselves holding depreciating assets financed by billions in debt.

The market's reaction to Nebius's offering reflected these concerns. Shares closed at $116.25 on March 17, down 10.47% — erasing most of the 14% gain from the Meta deal announcement the day before [1]. Investors welcomed the revenue visibility that the Meta and NVIDIA partnerships provide but balked at the dilution and leverage implied by the convertible offering.

Financial Trajectory: Growth at Breakneck Speed

Nebius's financial results tell a story of hypergrowth from a small base. The company reported full-year 2025 revenue of $529.8 million, a roughly 480% increase year-over-year [11]. Quarterly revenue accelerated throughout the year: from $55.3 million in Q1 to $228 million in Q4, when the company achieved positive adjusted EBITDA for the first time. Exit annualized recurring revenue reached $1.25 billion, exceeding guidance of $900 million to $1.1 billion.

These are impressive numbers, but they must be viewed against the scale of investment required. With $16-20 billion in planned 2026 capex against roughly $530 million in trailing revenue, Nebius is spending more than 30 times its annual revenue on infrastructure. The company is betting that the Meta deal's $12 billion in committed capacity, plus the $15 billion in potential additional spend, will generate returns that justify this extraordinary investment.

The Bigger Picture: AI Infrastructure as the New Oil

Nebius's story illuminates a broader dynamic reshaping the technology industry. As AI models grow larger and more compute-intensive, and as applications like agentic AI create new waves of inference demand, the physical infrastructure underpinning AI has become a strategic asset comparable to energy infrastructure.

The parallels to the energy sector are not merely rhetorical. Just as oil companies must invest billions in exploration and drilling years before revenue flows, AI infrastructure companies must build data centers, procure GPUs, and deploy networking equipment well in advance of customer workloads. The convertible debt structures that Nebius and its peers use mirror the project finance techniques long employed in energy and infrastructure.

The Iran war and its disruption to energy markets — already a major storyline in 2026, with oil prices surging past $100 per barrel and the Strait of Hormuz effectively closed — adds another layer of complexity. Rising energy costs directly impact data center economics, as electricity represents a major operating expense. Nebius's plan to deploy 5 gigawatts of capacity by 2030 would consume more electricity than many mid-sized cities [3].

For investors, the question is whether Nebius can execute at the scale its contracts demand while managing the financial risks of massive capital deployment. The company has secured blue-chip partners in Meta and NVIDIA, demonstrated explosive revenue growth, and carved out a defensible position in the neocloud market. But it is also a company with barely $500 million in annual revenue that has committed to spending $16-20 billion in a single year — a ratio that would make even the most aggressive Silicon Valley growth investors pause.

What Comes Next

The next twelve months will be critical for Nebius. The company must begin deploying infrastructure for the Meta deal, with dedicated capacity expected to come online in early 2027. It must also integrate NVIDIA's next-generation Vera Rubin platform at scale — a technology that has not yet been widely deployed. And it must continue raising capital; at current spending rates, even the $3.75 billion from this offering will be consumed in months.

Nebius also announced alongside the convertible offering a concurrent registered direct offering of 12.5 million Class A shares at $92.50 each, raising an additional $1.16 billion [1]. Combined with the convertible notes, the company is tapping the capital markets for roughly $5 billion in a single week — a remarkable figure for a company that was a suspended NASDAQ stock less than 18 months ago.

The neocloud race is far from decided. CoreWeave, with its larger revenue base and deeper backlog, remains the market leader. Traditional hyperscalers continue to invest hundreds of billions in their own infrastructure. And new entrants are constantly emerging in a market where demand currently outstrips supply.

But Nebius has achieved something rare: in less than two years, it has transformed from a post-Yandex shell company into a credible challenger for a central role in the AI infrastructure stack, backed by two of the most powerful companies in technology. Whether that transformation can sustain itself through the capital-intensive gauntlet ahead is the $3.75 billion question.