Russian authorities have gotten down to business in earnest to address the acute shortage of graphics processing units (GPUs) needed for the development of artificial intelligence systems. With no domestic chip production and imports severely restricted by sanctions, the government is seeking ways to stimulate purchases. The key task is not just to support businesses, but also to clearly separate AI equipment from mining hardware, so that benefits do not go toward cryptocurrency mining.
Action plan and timelines
According to the action plan of the working group of the government subcommittee, by the end of September, regulatory barriers preventing modern AI equipment from being classified as domestic products must be identified and eliminated. In parallel, technical criteria are being developed that will allow GPUs, NPUs, and high-speed networks to be unambiguously classified as AI solutions, excluding their use for mining.
By the end of the year, the group must present specific proposals on preferential taxation and other mechanisms that reduce the cost of owning such equipment. The Ministry of Industry and Trade has been appointed as the responsible body. However, specific support parameters have not yet been disclosed — neither rates nor mechanisms for providing benefits have been publicly announced.
Expert opinions: how to separate AI and mining
There is no unified approach in the professional community to separating equipment. Nikita Kazaryan, Head of the AI Platform direction at MWS Cloud, notes that over the past year, commercial GPU prices have more than doubled, and in some cases tripled. According to an MWS Cloud study, 47% of respondents expect further increases in GPU resource costs over the next 12 months.
Kazaryan proposes dividing equipment by power and class: purchasing expensive accelerators like the Nvidia H200 for mining is economically unfeasible. In turn, Yuri Tyurin, Technical Director at MD Audit, insists that the same chip is suitable for both AI and mining, so the purpose of the system should be assessed, not the component itself.
Sofia Smirnova, Advisor at O2 Consulting, points to a formal barrier: in the TN VED classifiers, mining and AI equipment fall under the same code. She recommends tying benefits not to the fact of purchase, but to confirmed intended use, otherwise support will inevitably leak into crypto mining and gray imports.
Dmitry Panyshev, a representative of the Data Center, emphasizes the localization problem: GPUs are not produced in Russia, and localization of production is required for Ministry of Industry and Trade support. He proposes a phased approach — starting with soft assembly requirements and gradually raising the bar over approximately 10 years. In his opinion, overly strict conditions would only slow down the market.
Andrey Pankov, CEO of HiTech, takes a stricter stance, insisting that benefits should apply only to equipment from the registry under Government Resolution No. 719. The main problem, in his words, is the absence of a separate category for AI chips and accelerators in OKPD 2, which makes it impossible to recognize the products as domestic.
My analysis: It is obvious that without clear criteria and a functioning system for monitoring intended use, any tax relief risks turning into subsidizing the crypto industry rather than technological sovereignty. The key intrigue remains not the size of the benefits, but how effectively the state can build administration — this determines whether the Russian AI sector will receive a real stimulus or whether we will witness another round of gray schemes.