The Russian high-performance computing market is on the verge of significant changes. The government is actively discussing a set of state support measures for companies purchasing equipment for artificial intelligence systems. This is not just about subsidies, but about a systemic solution to the shortage of graphics processing units (GPUs), which has worsened amid sanctions restrictions.
The key task facing the working group of the relevant subcommittee is to create transparent and effective incentive mechanisms. There is no domestic GPU production, and import supplies are severely restricted. Therefore, the priority is to develop criteria that will clearly distinguish AI equipment (GPUs, NPUs, high-speed networks) from mining equipment. This is fundamentally important so that benefits do not go toward cryptocurrency mining but instead work to develop technology.
Timelines and Mechanisms
According to available data, by the end of September, the authorities intend to identify regulatory barriers preventing modern AI equipment from being classified as domestic products and prepare the groundwork for their revision. By the end of the year, proposals should be ready on preferential taxation for purchases and other mechanisms to reduce the cost of owning such equipment. The Ministry of Industry and Trade has been appointed responsible for implementation. However, specific rates and support parameters remain undisclosed — final decisions have yet to be approved.
The market situation is critical. According to expert estimates, over the past year, commercial prices for servers with high-performance GPUs have more than doubled, and in some cases tripled. A survey conducted among market participants showed that 47% of respondents expect further increases in GPU resource costs over the next 12 months.
Opinions Are Divided
There is no consensus in the expert community on how to properly administer the benefits. Some propose dividing equipment by power and class, arguing that expensive accelerators like the Nvidia H200 are unprofitable for mining. Others insist that the same chip is suitable for both machine learning and cryptocurrency mining, meaning the criterion should be the system as a whole, not the chip itself.
Classification is a separate issue. Currently, mining and AI equipment fall under the same HS code, creating a formal barrier. There are proposals to tie benefits not to the fact of purchase, but to confirmed intended use. Otherwise, analysts warn, support will leak into crypto mining and gray imports.
There are also those who advocate for phased localization of production, starting with assembly under lenient requirements and gradually raising the bar over ten years. Too strict requirements now, it is generally agreed, would only slow down an already difficult market. At the same time, a number of experts insist that benefits should apply only to equipment from the domestic products registry, and for this, a separate category in the OKPD 2 classifier for AI chips and accelerators must be introduced.
My view: the initiative is timely, but its success directly depends on how well officials manage to avoid a formal approach. Without clear criteria for "intended use" and measures against gray schemes, the benefits risk becoming another channel for subsidizing shadow imports, which would only worsen market imbalances.