Roman Chernin, Nebius ｜ Raise Summit 2025

Nebius Business Model & Scope

• Operates as a full-stack, AI-specialized cloud provider spanning data center design, construction, hardware, and sophisticated software platforms.
• Differentiates through an integrated "computer" approach designed to evolve from dedicated clusters to a scalable, pay-per-consumption cloud experience for diverse users.
• Founded one year prior to the interview and recently celebrated its first anniversary.
• Currently a publicly traded company with a capital strategy that allows for advanced capacity deployment (raising capital before sales) to offer flexibility in on-demand, spot, and reservation models.

Infrastructure Footprint & Scale

• Expanded its infrastructure footprint from one data center in Finland to seven locations across Europe, the US, and the Middle East.
• Targets high tens of thousands of GPUs in operation by the end of the current year.
• Designs entire data centers and rack systems with specific software efficiency and cooling requirements in mind, leveraging proprietary IP in engineering and cooling systems.
• Transitioned from air-cooled systems to liquid-cooled infrastructure for most new builds to support high-density computing at gigawatt-scale campuses.

Market Dynamics & Customer Base

• Market shift observed where x86 market share dropped from 73% to 13% in one year, driven by GPUs, XPUs, and the integration of software into hardware.
• Customer base is diversifying from AI-native startups and technical labs to established enterprises, specifically in digital companies, life sciences, and FinTech (hedge funds).
• Service portfolio supports varied workloads including real-time inference (low latency), batch inference (cost-efficient using compute leftovers), and large-scale training jobs (dedicated interconnected clusters).

Financial Targets & Strategic Direction

• Committed to shareholders to achieve revenue of 1 billion Russian Rubles (RR) by year-end, representing a 10x growth trajectory.
• Growth strategy for the second half of the year and into 2026 focuses on two dimensions: securing sufficient data center and power capacity for "super labs" and hyperscalers, and expanding the software stack for enterprise adoption.
• Anticipates a significant acceleration in enterprise adoption over the next two years as use cases for RAG and automation prove viable.
• Plans to continue moving the market from domain-specific HPC to "AI factories" and supercomputing capabilities available to the masses.