The Russian government intends to resolve the issue of dividing data processing centers (DPCs) into two separate classes by the end of September 2026: facilities intended exclusively for artificial intelligence, and all other sites. This initiative, overseen by the Ministry of Digital Development under the leadership of Sberbank Senior Vice President Andrei Belevtsev, is designed to ensure targeted state support and protect budget subsidies from misuse.
The idea for the reform emerged back in 2025 on the instructions of Deputy Prime Minister Alexander Novak. The key task is to create a fair competitive environment and eliminate situations where ordinary data centers obtain AI-infrastructure status for the sake of benefits. Under the "Guillotine 2.0" project, the Ministry of Digital Development, the Ministry of Construction, and the Analytical Center under the government must officially clarify the very concept of a DPC by September 1. Notably, the Ministry of Digital Development had previously given a negative review of such an initiative, and an article on computing infrastructure never made it into the relevant law signed by Vladimir Putin on July 26.
Why this matters: scale and cost
The argumentation of supporters of the changes is built on colossal investments. The architecture of AI sites is fundamentally different: servers with graphics processors consume enormous amounts of energy, and the facilities themselves are often built far from end users. According to estimates, the cost of a conventional DPC with a capacity of 10 MW is about 20 billion rubles, while a similar AI data center would cost twice as much — approximately 40 billion rubles. Representatives of the company "Atomdata" (a Rosatom structure) predict that the market's need for new capacity by 2030 will reach 2.5 GW — double the current volume of all commercial sites in the country.
Criticism: criteria are vague
The largest operator, "RTK-DPC" from the Rostelecom group, opposed the bill. Dmitry Panyshev confirms the presence of specific characteristics in specialized facilities, including high connected power and special engineering systems. However, he emphasizes: such high-tech infrastructure can easily be used for traditional computing as well, so directly tying subsidies to the wording "for AI" seems unjustified. The operator proposes evaluating facilities solely by their level of reliability.
Market analysts, including leading consultant at iKS-Consulting Stanislav Mirin, point to the complete absence of clear technical criteria. One cannot simply call racks at 15, 20, or 50 kW high-load, and liquid cooling cannot serve as the sole factor. Powerful equipment is often used for standard tasks without AI: three-dimensional graphics rendering, modeling complex physical processes, precise strength calculations.
In my opinion, objective criteria for a rigid division do not yet exist, and the real tasks in this area are handled by a limited circle of companies. The authorities could support such organizations manually, without creating a new class. The industry does not currently count on large-scale state assistance: last week, the ANO "Digital Economy" requested only 72 billion rubles per year from the government for tax and financial benefits to develop technologies and build sites. This is a drop in the ocean compared to the needs, but even it requires clear rules of the game, which do not yet exist.