GPU shortages block AI projects at 91% of Russian firms

GPU capacity is the main obstacle to AI projects for 91% of companies surveyed in Russia, according to research by OrionSoft. A shortage of AI specialists is also slowing adoption: 79% of large organizations said they lack the necessary personnel.

The findings cover the first half of 2026, when Russia’s AI infrastructure readiness index reached just 38 points out of 100. OrionSoft’s assessment suggests that the market has not yet reached full-scale industrial deployment. Companies want results from neural networks faster than they can build the teams and hardware needed to support them.

How far Russian companies have progressed

The survey included 100 representatives of large companies. Interest in AI is widespread, but most organizations have not moved beyond planning or early pilots:

Recommended reading

330 MW data center sues for California farm water

• 28% are running one or two AI projects.
• 21% are working on more than two projects.
• The remaining organizations are still planning launches.

That split points to a gap between experimentation and broader deployment. Companies may have identified use cases, but expanding them requires both trained machine-learning staff and access to GPU-equipped servers.

OrionSoft’s findings place the Russian market within a broader pattern also seen in other countries. Businesses are competing for machine-learning engineers, facing higher accelerator prices and dealing with limited cloud capacity. For Russian companies in 2026, planning an AI initiative therefore involves not only selecting models, but also securing physical GPU infrastructure.

The source does not provide a breakdown of the GPUs or servers respondents need, nor does it specify how the readiness index was calculated. It also does not identify when the surveyed companies expect to move from pilots to production.

Ilya Ignatov is a technical journalist and news writer. He graduated from the Moscow Technical University of Communications and Informatics with a degree in information security and has covered hardware, software and consumer electronics since 2018.