Nebius CEO Arkady Volozh explains why his blockbuster Meta deal isn't the endgame

Nebius stands out in the rapidly evolving AI landscape as a unique startup that emerged from the disbanding of Russian search giant Yandex in 2024. Armed with a substantial capital infusion of approximately $2.5 billion and a skilled workforce, Nebius is positioned far ahead of many new competitors in the AI infrastructure domain. With years of expertise in constructing large-scale data centers, the team is well-equipped to tackle the burgeoning demand for AI solutions. In a recent conversation with CEO Arkady Volozh during Nvidia's GTC conference in Silicon Valley, the company made headlines by announcing a monumental $27 billion AI computing agreement with Meta. This announcement propelled Nebius’s stock by 15%, elevating its valuation to around $30 billion. Following this event, Volozh shared insights on the current AI infrastructure landscape, highlighting what he refers to as the "four Cs"—capacity, capital, chips, and customers—which he believes are critical to understanding the industry's challenges. Volozh pointed out that the sector is grappling with significant bottlenecks. For instance, the demand for data centers has surged, but the physical capacity to build them is lagging behind. Supply chain issues have led to shortages in essential components, such as transformers and gas generators, with Nebius aiming for over 3 gigawatts of power contracted by the end of 2026. Capital requirements have also escalated dramatically; Volozh estimates that acquiring just 10% of the AI infrastructure market may necessitate around $400 billion in investments. The landscape for chips has shifted as well; the initial focus on GPU allocations has expanded to include memory chip shortages, further complicating production. Customer demand for AI services has become unmistakable, contrasting with the skepticism of the previous year. Volozh emphasized that Nebius’s approach to building its infrastructure differs from others that rely on leasing or external hardware. By constructing its own servers and data center components, Nebius not only enhances its profit margins but also ensures a more efficient allocation of resources tailored to customer needs. Volozh elaborated on Nebius's strategy, likening the company to a fourth hyperscaler in the market. This means controlling the entire computing stack, from physical infrastructure to software, allowing for significant cost savings and a more integrated service offering. He noted that most companies operate at one or two layers of this stack, paying a premium to third parties, while Nebius aims to capture value across all five layers of the AI ecosystem. Despite being a relatively new entity, Nebius benefits from its team's extensive experience in building infrastructure at scale. This expertise is crucial as the company navigates the complexities of securing land, permits, and energy contracts. Volozh stressed the importance of establishing trust within local communities, ensuring that Nebius's projects provide tangible benefits to residents. Addressing the issue of 'dark GPUs'—idle computing resources that customers cannot fully utilize—Volozh explained that Nebius has designed its cloud infrastructure specifically for AI engineers, ensuring optimal capacity management and reliable service as customers expand. Interestingly, Volozh clarified that while the Meta partnership is substantial, it does not define Nebius’s long-term business objectives. The core focus remains on building a multi-tenant AI cloud platform that meets the diverse needs of startups and enterprises. Collaborations with major players like Meta and Microsoft serve as vital support for Nebius’s growth, facilitating the development of infrastructure that will ultimately benefit a broader market. In summary, Nebius is strategically positioning itself for sustainable growth in the AI sector, leveraging its unique capabilities and partnerships to navigate the current challenges and capitalize on future opportunities.