Table of Contents
Nebius’ CEO Roman Chernin said during a presentation at the Bank of America Global Technology Conference that inference is one of the fastest-growing segments of the market and provides infrastructure providers with greater flexibility because workloads can be optimized through software and distributed across different hardware platforms
In sum – what to know:
Inference growth – Nebius said inference is one of the fastest-growing segments of the AI market as usage expands beyond model training into recurring application workloads.
Software focus – The company views software, inference optimization and recent acquisitions as important tools for improving performance and efficiency.
Diversified strategy – Nebius aims to serve a broad range of customers while expanding infrastructure capacity through both large contracts and self-developed data centers.
Neocloud company Nebius Group used the Bank of America Global Technology Conference to outline how it believes the AI infrastructure market is evolving, arguing that future growth opportunities extend beyond providing GPU capacity and increasingly include managed infrastructure, inference platforms, and AI services.
Speaking at the conference, the firm’s chief business officer Roman Chernin described Nebius as an AI infrastructure platform focused on serving different layers of AI customers, from large model developers to application builders and, potentially, developers of AI agents. “We don’t sell data center capacity. We sell product that built on top,” Chernin said during the presentation.
His comments provide insight into how Nebius views the next phase of the AI infrastructure market as demand expands beyond model training toward large-scale deployment of AI applications and services.
Chernin said the market consists of several layers of AI infrastructure consumption. At the foundation are hyperscalers and large AI labs that primarily require large-scale compute capacity for model training. Above that are AI-native companies that prefer managed infrastructure rather than operating their own software stacks and clusters.
Nebius initially focused on this segment through what Chernin described as managed infrastructure services. The company later expanded into managed inference through its Token Factory platform, which serves developers building AI products and applications.
According to Chernin, many of these customers increasingly prefer consuming models and inference services rather than directly managing infrastructure.
A recurring theme throughout the discussion was Nebius’ view that inference workloads represent a growing opportunity.
While training large AI models requires significant compute resources, Chernin noted that training is typically a one-time activity, whereas inference generates recurring demand as users interact with AI systems.
According to Chernin, inference is one of the fastest-growing segments of the market and provides infrastructure providers with greater flexibility because workloads can be optimized through software and distributed across different hardware platforms.
He said this flexibility can help extend the useful life of AI infrastructure investments and create additional opportunities to improve efficiency.
Chernin argued that software capabilities are becoming an important differentiator in the AI infrastructure market.
He said Nebius’ strategy is to build services that allow customers to focus on research, products and applications while the company manages the underlying infrastructure complexity.
The company recently acquired Eigen AI and Clarifai, two moves Chernin said are intended to strengthen Nebius’ inference capabilities.
According to him, Eigen brings expertise in extracting more performance from individual GPUs through model optimization techniques, while Clarifai contributes experience operating inference systems at scale, including orchestration, caching and workload management.
Although Nebius aims to build a diversified customer base spanning AI-native companies, enterprises and developers, Chernin said large contracts with hyperscalers remain important. He explained that such agreements help finance infrastructure expansion and support investment in new products and services.
At the same time, Nebius is pursuing a broader mix of customers, workloads and contract structures to create greater flexibility. “We believe that our long-term business is in a diversified portfolio of the customers,” Chernin said.
The company also expects to bring more infrastructure development in-house. According to Chernin, a significant portion of future capacity additions will come from data centers Nebius develops itself, providing greater control over costs, deployment schedules, and operational flexibility.
