Alexander Vedyakhin, Sberbank’s first deputy CEO, believes the global market is shifting away from massive, resource-heavy models. While acknowledging that Russian systems like GigaChat currently trail competitors such as Anthropic or DeepSeek in overall complexity, he argues that specialized, compressed AI offers a more pragmatic path for emerging economies. The goal is to deliver cost-effective solutions for specific tasks, such as credit scoring, rather than general-purpose tools that consume vast amounts of infrastructure.

The push for sovereign AI comes as Russia attempts to accelerate its own technological development despite Western sanctions limiting access to high-end hardware. Sberbank is exploring partnerships, including potential Chinese chip acquisitions, to sustain its research. Vedyakhin emphasizes that the industry is hitting a saturation point regarding parameter counts; instead, the future lies in efficiency and local adaptation. To demonstrate this focus on utility, the bank is currently testing an AI-powered tiling robot, shifting the corporate narrative from speculative automation to tangible, sector-specific application.