Detailed Narrative
Aggressive Capacity Expansion for AI Compute
Nebius is aggressively ramping up its data center capacity, expecting to secure 220 megawatts of connected power by the end of 2025, with 100 megawatts active. This includes expansions in New Jersey and Finland. The company is also in the process of securing over 1 gigawatt of power by the end of 2026, driven by strong demand for AI compute. Greenfield sites are preferred for cost control and tailored phasing📎, aiming for a 20% lower total cost of ownership than market average.
Strong Financial Performance and Profitability Ahead of Schedule
The company reported $105.1 million in Q2 revenue, a 625% YoY and 106% QoQ increase. AI cloud infrastructure revenue grew more than 9x YoY. The core AI business achieved positive adjusted EBITDA this quarter, ahead of initial guidance, and is expected to remain positive. Group-level adjusted EBITDA is anticipated to turn positive by year-end 2025 and remain positive starting in 2026.
Software Platform Enhancements and Customer Traction
Nebius made significant enhancements to its software cloud platform to support expanding capacity and large-scale clusters. Key improvements include doubling network speed, increasing reliability through auto-healing and health check software, and improving flexibility with S3 storage migration. The company is gaining traction with large enterprise customers like Cloudflare, Prosus, and Shopify, while continuing to serve AI tech startups such as HeyGen and Lightning.AI.
Strategic Partnerships and Global Market Expansion
Nebius expanded its ecosystem through partnerships with Mistral, Baseten, SkyPilot, Lightning AI, and Anyscale, enhancing ease of use and developer support. The company strengthened its relationship with Nvidia, becoming a launch NCP partner for Nvidia DGX Cloud Lepton and expanding its AI cloud portfolio with Nvidia AI Enterprise. Nebius is also expanding its physical presence with new GPU clusters in the U.K. and Israel, expected to come online in early Q4, targeting vibrant local AI markets.
Inference as a Service and Go-to-Market Strategy
The company is developing an enterprise-grade Inference as a Service platform, leveraging its full-stack capabilities to deploy and scale open-weight AI models on dedicated infrastructure. This initiative aims to address pain points in production AI like unpredictable latency and GPU bottlenecks. Nebius is also building out its go-to-market leadership team, including GMs for various regions and a VP of Sales Strategy, to proactively pursue market opportunities and expand its enterprise footprint.