Russian Businesses Increase Consumption of Chinese AI Models Several-Fold

Kommersant has found that clients of major Russian cloud platforms have several-fold increased their consumption of foreign large language models this year. Yandex AI Studio, MWS GPT, and Cloud.ru confirmed the trend. On Yandex AI Studio, AI resource consumption in the first half of 2026 reached 594 billion tokens, 18 times more than in the same period last year. Yandex's own models accounted for 54% of consumption, Chinese Alibaba Qwen and DeepSeek for 21.2% and 13.5%, and American OpenAI GPT-OSS for 7.5%. On MWS GPT, consumption of Chinese AI models grew 11-fold to 400 billion tokens in the first half of 2026, with Alibaba Qwen the most popular (261.1 billion tokens), followed by Zhipu AI GLM (91.2 billion) and Moonshot AI Kimi (81.4 billion), which displaced DeepSeek. Cloud.ru reported an 18-fold growth in AI model consumption over eight months of 2026 but did not provide absolute figures. According to MWS Cloud, Russian companies are increasing consumption of Chinese AI models faster than American ones, as Chinese models have caught up with American capabilities and are now used for the most complex tasks. On Yandex AI Studio, consumption of open Chinese models grew five-fold year-on-year. In June, Sberbank's head German Gref stated that Russia lags behind China in AI by six to eight months, and the US ahead of China by another six months, but this gap is not critical as Russia is among the top five countries in AI development.