A Partnership That Maps Policy to Geography
When Rostelecom and the Krasnoyarsk Territory administration signed a digital infrastructure partnership at the St. Petersburg International Economic Forum (SPIEF 2026, June 3–6), it looked like a routine intergovernmental handshake. It was more than that. The agreement — under which RTK-TSOD, Rostelecom's data centre subsidiary, will study phased construction of a new facility in the region — directly operationalised Deputy Prime Minister Alexander Novak's June 1 article in the journal Energy Policy, which named Krasnoyarsk Territory as one of Russia's most promising locations for future data centre development, precisely because of its hydroelectric infrastructure on the Yenisei River.
Russia's digital infrastructure problem is geographic before it is technical. Moscow and the Moscow Region host approximately 80% of the country's operating data centres. St. Petersburg accounts for a further 7%. The remaining fraction is spread across a country spanning eleven time zones. For a state trying to grow installed data centre capacity 2.4-fold — from 1.8 GW today to 4.3 GW by 2030, as Novak projected — building more in Moscow is neither economically sustainable nor strategically sound. The capital's power grid cannot absorb unlimited additional demand; land costs are high; and geographic concentration creates resilience risk for a state that has explicitly framed data infrastructure as a sovereign concern.
The solution Novak outlined is geographically elegant: take the compute to the power, not the other way around. New facilities are to be sited in regions where Russia has abundant, underutilised energy — nuclear plants in Leningrad, Tver, Smolensk, Kursk, Tomsk, and Sverdlovsk oblasts; major hydroelectric capacity in Krasnoyarsk, Khakassia, Irkutsk, and Amur regions. The Rostelecom-Krasnoyarsk agreement is the first publicly visible contract to give that directive a specific geography and a named operator.
The Energy Arbitrage Argument
Critics of top-down industrial policy should reckon honestly with the steelman case here: it is strong. Electricity is the single largest operating cost for a data centre, typically 40–60% of total OpEx. Siberia and Russia's Far East benefit from power prices well below Moscow rates, driven by surplus hydro and nuclear generation historically dedicated to energy-intensive industries like aluminium smelting. As those industries restructure, stranded capacity exists that can be redirected cheaply. A data centre anchored to a large hydro station on the Yenisei draws power that is simultaneously low-carbon, cost-predictable over decades, and not subject to gas price volatility. These are genuine competitive advantages, not bureaucratic fictions.
Rostelecom's commercial appetite for exactly this kind of deal is real and predates the Krasnoyarsk agreement. In April 2026, the company disclosed plans for a new facility of approximately 100 MW at a cost of around 100 billion rubles — its largest to date — and explicitly cited the impossibility of securing adequate grid capacity in the Moscow region as the primary driver for looking elsewhere. The state-linked company and the state directive are pulling in the same direction, which is a meaningful difference from policy environments where mandates run counter to commercial logic.
The scale of latent demand makes the ambition credible. Novak estimated that Sberbank, Yandex, Rostelecom, and Rosatom alone collectively require roughly 2 GW of new data centre capacity by 2030. Annual power consumption by Russian data centres is already growing at approximately 18% per year. At that compounding rate, the installed base doubles well before 2030 even without greenfield development in new regions. The question is not whether Russia builds more data centres; it is where. Novak's framework and the Krasnoyarsk partnership are attempts to answer that question before the Moscow grid answers it by default.
Where State-Led Build-Out Creates Friction
The concern is not the energy logic, which is sound. The concern is the execution model. Russia's approach is explicitly state-mediated: a Deputy PM's journal article functions as a de facto policy directive; regional governments negotiate with state-linked carriers; RTK-TSOD is positioned as the primary implementation vehicle. Three structural friction points are worth naming.
Equipment constraints. Russia's data centre expansion faces a hardware bottleneck that geography cannot solve. Sanctions since 2022 have restricted access to high-performance server components and advanced networking gear. Parallel import channels and Chinese hardware partnerships partially fill the gap, but the constraint compresses both speed and unit economics. A 4.3 GW national target requires equipment that cannot yet be domestically produced at scale.
MOU-to-shovel distance. The Krasnoyarsk agreement is a framework to 'work out the possibility' of a data centre — not a committed project with locked capital and a construction start date. The gap between a SPIEF partnership signing and an operational Siberian data centre campus is years of permitting, grid extension, fibre connectivity buildout, and cooling engineering. Russian regional digitalization agreements have a mixed record of converting intent into infrastructure on schedule.
Market pluralism. A policy environment in which Rostelecom is the designated national carrier for state-adjacent data centre development crowds out co-location operators, cloud-native providers, and potential foreign partners who might bring faster innovation cycles. Russia's compute infrastructure ecosystem benefits from competition — multiple operators with different incentive structures consistently outperform single-carrier models on cost and responsiveness. If the siting directive becomes synonymous with Rostelecom monopoly, that advantage is forfeited.
The Broader Trajectory
Russia is not alone in directing compute infrastructure toward cheap, clean power. The United States, Ireland, and the Nordic countries have all seen data centre investment cluster around abundant electricity — though primarily through market price signals rather than Deputy PM journal articles. The difference is institutional: in most markets, cheap power attracts capital; in Russia, cheap power attracts state-directed tenants. Whether that distinction matters for outcomes depends on whether RTK-TSOD can commission facilities that anchor government systems, attract commercial colocation, and ultimately prove the Siberian energy arbitrage to private operators who might follow.
Novak's data point — that global data centre power consumption may reach 110 GW by 2030, rising from 1.5% to 3% of global energy consumption — frames Russia's domestic challenge within a genuine worldwide constraint. The country has real energy assets for the compute era. The question is whether the state can organise its way to using them without crowding out the competitive dynamics that make compute infrastructure economically productive rather than just physically present.
The Rostelecom-Krasnoyarsk partnership is a credible first step. It is not yet evidence that the step will be taken.