Nebius AI Cloud Growth Fuels Stock Surge

Nebius Group, the Amsterdam-based specialist in AI-tailored cloud infrastructure, has reported a second quarter of fiscal year 2026 that significantly outpaced market expectations. The company's revenue soared by an impressive 454% year-over-year, reaching $582.3 million. This robust growth was primarily fueled by its core AI cloud services segment, which saw a staggering 514% increase to $574.9 million, far exceeding the approximately $557 million anticipated by analysts.

The market has responded enthusiastically to these results, with Nebius's Nasdaq-listed stock (NBIS) climbing approximately 16.6% in pre-market trading. Year-to-date, the stock has appreciated by a remarkable 131%, underscoring investor confidence in the company's specialized "Neocloud" model. This model focuses on providing GPU capacity and a comprehensive software stack specifically for AI workloads, positioning Nebius as a key player against hyperscalers like AWS, Microsoft Azure, and Google Cloud, as well as fellow AI infrastructure providers such as CoreWeave, Crusoe, and Lambda.

Profitability within the AI cloud division also showed significant improvement. Adjusted EBITDA for the segment reached $285.7 million, translating to a margin of 49.7%. This marks a substantial increase from 45% in the preceding quarter and 24% in Q4 2025, indicating enhanced operational efficiency and pricing power. On a consolidated basis, adjusted EBITDA swung to a positive $236.2 million, a dramatic turnaround from a $21 million deficit in the prior year's comparable period. The company's annualized recurring revenue (ARR) now stands at $3.0 billion, up from $1.9 billion at the close of the first quarter.

Nebius is aggressively reinvesting in its expansion, allocating roughly $5.7 billion in the second quarter alone towards GPUs, associated hardware, and data center development. The company ended the quarter with a strong liquidity position of approximately $8 billion. Funding for this rapid scaling is being sourced through multiple avenues, including an at-the-market program that generated about $2.8 billion in gross proceeds by placing 12.7 million Class A shares. Additionally, a $775 million secured debt facility was secured in July, backed by existing GPU infrastructure and a contract with an investment-grade client.

The company's strategic focus on securing long-term commitments is evident in its deal pipeline. Nebius reported closing four major contracts in Q2, each valued at over $1 billion. The annual contract value per megawatt is between $20 million and $25 million, with older GPU generations commanding over 30% higher pricing than in the previous quarter. A significant portion of these new agreements, around 70%, include substantial upfront payments covering 50-60% of the investment costs. This has effectively reduced the expected payback period for new deployments to an average of one year and ten months, down from two to three years.

Key clients contributing to this growth include AI startup Reflection, Canadian AI firm Cohere, an unnamed major US AI laboratory, and a prominent US quantitative trading firm. Other notable customers include AMI, Basecamp Research, and Higgsfield. Nebius has also fulfilled its capacity commitments to Microsoft and is developing a second agreement with Meta, slated for activation in early 2027. The company has raised its target for contracted power capacity by year-end to 5 gigawatts and plans to bring over 1 gigawatt online annually from 2027 onwards, with new facilities established in the UK, Estonia, and Finland.

To mitigate capital expenditure demands, Nebius is introducing an "asset-light" partnership model. Under this structure, partners will own and finance the data center infrastructure and hardware, while Nebius provides its proprietary cloud platform, system architecture, and global sales organization. This division of labor allows Nebius to generate high-margin revenue from capacity it doesn't directly finance, addressing industry bottlenecks in both capital and available capacity.