Nebius Group Adds $476M Market Cap on Meta AI Deal, Nvidia Tie-Up
European AI infrastructure play gains validation as Meta picks non-hyperscaler compute partner amid U.S. capacity constraints.
Nebius Group closed up 21.3% on Tuesday after disclosing a multi-year infrastructure agreement with Meta and an expanded partnership with Nvidia, adding roughly $476 million in market capitalization to the Amsterdam-listed AI compute provider. The stock, which trades under ticker NBIS, moved on volume 340% above its 30-day average as institutional desks repositioned around the first major European AI infrastructure contract with a Magnificent Seven incumbent.
Meta will deploy Nebius-managed GPU clusters for AI model training and inference workloads, marking the first time the social platform operator has contracted significant compute capacity outside its own data centers and the hyperscaler trio of AWS, Azure, and Google Cloud. Nebius, spun out of Yandex in late 2023, operates 12 data center facilities across Europe and North America with 47,000 Nvidia H100 and H200 GPUs currently deployed. The company did not disclose contract value but confirmed the initial deployment covers 8,000 H200 units scheduled for Q2 delivery, with option clauses extending through 2027. Nvidia separately formalized Nebius as a preferred infrastructure partner in its AI Enterprise program, granting early access to Blackwell architecture shipments expected in Q3.
The deal matters because it confirms three theses allocators have debated since Nebius listed six months ago. First, Meta's willingness to source outside its captive infrastructure signals that internal build-out is no longer keeping pace with model scaling needs—Llama 4 training runs are reportedly consuming 40% more compute per parameter than Llama 3, and Meta's capex guidance of $60-65 billion for 2025 still leaves a gap. Second, European data sovereignty concerns are creating structural demand for non-U.S. domiciled compute, particularly for inference workloads subject to GDPR and the EU AI Act. Third, Nvidia's formal endorsement resolves lingering questions about Nebius's supply chain access post-Yandex separation; the company now holds the same tier-one OEM status as CoreWeave and Lambda Labs. Credit Suisse equity research upgraded Nebius to Outperform on Wednesday morning, raising its price target from €28 to €38 on revised 2026 EBITDA estimates that now assume 62% GPU utilization versus the prior 48% base case.
The Meta contract also reshapes competitive dynamics in the AI infrastructure layer. CoreWeave, valued at $19 billion in its December Series C, has operated as the de facto non-hyperscaler compute partner for OpenAI and Microsoft. Nebius now offers a European regulatory arbitrage that CoreWeave's U.S. domicile cannot match, while its lower equity valuation—currently $2.8 billion—implies a 6.2x forward revenue multiple versus CoreWeave's estimated 11x. Lambda Labs, recently valued at $1.5 billion, remains subscale with fewer than 10,000 deployed GPUs. The gap between Nebius's current utilization rate of 71% and CoreWeave's reported 89% suggests room for margin expansion if the company can convert its remaining 13,600 idle GPUs into contracted capacity over the next two quarters.
Operators should track three developments. First, whether Nebius secures follow-on contracts with other Magnificent Seven players—Microsoft and Google are both building European sovereign cloud regions and face similar capacity constraints. Second, Blackwell delivery timing; Nvidia has pushed volume shipments from Q2 to Q3, and any further delay compresses Nebius's ability to deploy the 22,000 GB200 units it has on order before year-end revenue recognition deadlines. Third, whether the company accesses public debt markets to fund its $1.8 billion capex plan for 2025-2026—current liquidity of $420 million covers near-term GPU deposits but leaves minimal buffer for construction overruns.
Nebius reports Q1 earnings on May 12. Consensus expects $187 million in revenue, up 118% year-over-year, with the Meta contract contributing meaningfully starting in Q2. The company's October roadshow projected reaching EBITDA breakeven at 75% utilization; it is now 4 percentage points away.