At the Digital Solutions 2026 forum, representatives from IKS Holding, YADRO, Rubytech, Avito, T-Bank, and RTK-TsOD discussed scaling AI workloads. The session focused on managing corporate infrastructure, balancing internal computing power with cloud services, and optimizing resource allocation.
Viktor Urusov from Rubytech emphasized the importance of integrating AI into core business processes. The company measures efficiency by tracking token costs and pipeline execution. Meanwhile, Avito has shifted to an AI-First management model, focusing on transparent monitoring of GPU consumption and algorithm performance.
T-Bank reported success in reducing the unit cost of AI experiments, allowing large teams to access computing resources affordably. By balancing internal GPU usage with external token purchases, the bank maintains low costs per engineer during the experimental phase.
RTK-TsOD introduced an AI assistant designed for secure customer environments. The tool processes logs, monitoring metrics, and diagnostic data within a protected perimeter. The company notes that the ROI from AI implementation can reach 250 percent.
Experts concluded that large companies will likely adopt a hybrid model for AI infrastructure. This approach combines internal computing resources, cloud infrastructure, and external services, with critical workloads remaining within the company's own secure perimeter.
