Russia AI Data Center Market Size, Share & Forecast to 2035

The Russia AI Data Center Market covers the infrastructure that Russian operators install to train and run artificial intelligence (AI) models: AI servers and graphics processing unit (GPU) accelerators, networking, storage, and the power, cooling and building work of halls built for AI. DC Market Insights values the Russia AI Data Center Market at USD 2.24 billion in 2025 and forecasts USD 9.65 billion by 2035, a CAGR of 15.71%. The 2025 value is built from the build itself: about 58 megawatts (MW) of new AI IT capacity at roughly USD 36.00 million of all-in spend per MW gives USD 2.09 billion, and the replacement of servers in older AI halls adds USD 155.52 million. Colocation is the fastest-growing segment at 20.37% a year because AI tenants are moving into new campuses built for racks of up to 44 kilowatts (kW), and the Siberian Federal District is the fastest-growing region at 23.28% a year because its power base can support single data centers of up to 200 MW.

Executive Summary

The Russia AI Data Center Market size was valued at USD 359.50 million in 2020, reached USD 2.24 billion in 2025, and is anticipated to reach USD 9.65 billion by 2035, at a CAGR of 15.71% during the forecast period.

Report attribute	Details
Historical Period	2020-2024
Base Year	2025
Forecast Period	2026-2035
Russia AI Data Center Market Size 2025	USD 2.24 Billion
Russia AI Data Center Market, CAGR	15.71%
Russia AI Data Center Market Size 2035	USD 9.65 Billion

The market grew at 44.23% a year between 2020 and 2025, from a low base set by the first GPU supercomputers of Yandex, Sber and MTS. The path was uneven: DC Market Insights models a 36.3% fall in 2022, when accelerator and processor supply was disrupted, followed by a recovery to USD 750.00 million in 2023 and USD 1.36 billion in 2024. Spending rose a further 64.5% in 2025 as Russian banks and internet companies moved generative AI from pilots into products.

Growth slows to 15.71% a year from 2025 to 2035 for three reasons that DC Market Insights builds into the model: spend per MW falls about 2.8% a year as each server generation packs more compute into each kilowatt, power in the Moscow region is scarce, and borrowing costs remain high. At a planning rate of RUB 82.50 per US dollar, a DC Market Insights assumption, the 2025 market is worth about RUB 185 billion.

Market Scope: Does Russia Have AI Data Centers, and What Does the Market Cover?

Yes, Russia has about 144 MW of AI IT capacity installed since 2020, DC Market Insights estimates. DC Market Insights sizes the Russia AI Data Center Market at USD 2.24 billion in 2025. The market is valued at first purchase in current US dollars, with 2025 as the base year and forecasts to 2035.

Included. AI servers and accelerators (GPU servers, AI-optimized servers and domestic AI server platforms), the networking and interconnect that links them, storage bought for AI clusters, power and cooling infrastructure for AI halls (uninterruptible power supplies, switchgear, liquid and high-density air cooling), and the construction of halls or buildings dedicated to AI.

Excluded. AI software and model development, staff costs, GPU cloud rental revenue (counted once, as the hardware the cloud provider buys), electricity bills, and general-purpose data center capacity that does not host AI workloads.

How the segments fit together. Component, data center type, workload, end user and federal district are five ways of cutting the same USD 2.24 billion. A GPU cluster that Sber builds for its own models sits in enterprise and bank-owned data centers by type and in banking, financial services and insurance (BFSI) by end user; a GPU server that a cloud provider rents out sits in hyperscale and cloud provider data centers.

Market Drivers: Is Russia Building AI Data Centers?

Yes, Russia added about 58 MW of new AI IT capacity in 2025, DC Market Insights estimates. DC Market Insights models additions rising to 173 MW a year by 2030, driven by bank and internet-company AI budgets, a national supercomputing target and data localization rules.

Sber Plans RUB 350 Billion of AI Investment in 2026, 2.3 Times Its 2025 Level

DC Market Insights estimates that Sber is the largest single buyer of AI infrastructure in Russia. The bank invested more than RUB 100 billion in AI in 2024 and RUB 150 billion in 2025, and plans to raise AI investment 2.3 times to RUB 350 billion in 2026. Those totals cover models, software and staff as well as infrastructure; DC Market Insights attributes about half of the 2025 figure to data center infrastructure. Sber chief executive Herman Gref said in November 2024 that the bank previously built 40 MW data centers but now needs sites of at least 200 MW, preferably 300 to 400 MW; Gref said in September 2024 that Sber runs more than 63,000 servers in four data centers.

Yandex Spent RUB 55 Billion on Alice AI in 2025

Yandex reported combined capital and operating spending of RUB 55 billion on its Alice AI assistant in 2025, on group revenue of RUB 1,441.1 billion (Yandex full-year 2025 results). The company says its capital expenditure relates mainly to servers and data center equipment. Yandex capital expenditure reached RUB 41.1 billion in the second quarter of 2026, 10.7% of revenue. Yandex Cloud plans a new availability zone in 2026 based on a new Yandex data center of more than 40 MW, and Yandex says its campuses allow capacity to grow to 180 MW, a record for Russia.

A National Target of 1 Exaflop of Supercomputer Capacity by 2030

The updated National AI Development Strategy, approved by presidential decree No. 124 of 15 February 2024, requires the combined peak capacity of all Russian supercomputers to reach at least 1 exaflop in 2030. A government roadmap signed in March 2026 aims to raise the combined capacity of Russian supercomputers at least tenfold. For comparison, Yandex’s largest system, Chervonenkis, delivers 21,530 teraflops from 1,592 NVIDIA A100 GPUs and ranked 19th on the TOP500 list in November 2021. The Russia Data Center Market report covers the wider capacity base these AI halls sit in.

Market Trends: How Are Russian AI Data Centers Changing?

New Moscow-region AI halls are designed for racks of up to 44 kW. Russian AI data centers are moving to denser racks, larger campuses and more inference work, against 4.7 to 7 kW per rack in many regional halls.

Rack Density Rises from 5 to 10 kW toward 40 kW

Selectel says a non-AI server rack draws 5 to 10 kW on average, while racks with AI equipment can draw up to 40 kW. New builds are designed around that gap. IXcellerate’s MOS11 facility in the Moscow region has 40 MW of electrical capacity with an average rack density of 12 kW and scalability up to 44 kW for high-performance computing. Cloud.ru’s new Moscow-region data center is designed for up to 890 racks at 20 kW per rack. Rostelecom’s Nizhny Novgorod data center, by contrast, provides 4.7 to 7 kW per rack. DC Market Insights forecasts power and cooling infrastructure to grow at 17.07% a year as operators retrofit and build for liquid and high-density air cooling.

Operators Move from Single Buildings to Campuses

Yandex is moving to a campus model, with campuses allowing up to 180 MW, and its data centers reach a power usage effectiveness (PUE) of 1.1. IXcellerate’s Veshki campus will deliver 7,500 rack spaces and more than 130 MW across two facilities. In the view of DC Market Insights, campus scale lowers the cost of each MW and is the most practical way to meet requests such as Sber’s 200 to 400 MW.

Inference Overtakes Training by the End of the Decade

Training accounts for USD 1.28 billion (57.0%) of 2025 spending and inference for USD 964.71 million (43.0%). DC Market Insights forecasts inference to grow at 19.32% a year and reach 58.5% of the market by 2035, as banks, telecom operators and online services run assistants and recommendation models for millions of users. MWS Cloud, the cloud arm of MTS, says about 60% of GPU demand in its market comes from AI and machine learning tasks.

Market Challenges: What Holds the Russia AI Data Center Market Back?

Three constraints slow growth to 15.71% a year, from 44.23% in 2020 to 2025. Export controls on accelerators, a shortage of power in Moscow and high interest rates are the constraints that DC Market Insights builds into the forecast.

Export Controls Raise the Cost of Every GPU Server

United States and European Union export controls widen the list of goods that cannot be shipped to Russia without a license. A June 2024 rule of the US Bureau of Industry and Security (BIS) added 522 Harmonized Tariff Schedule (HTS) six-digit code entries to the items that need a license for Russia and Belarus (US Federal Register, 18 June 2024). MWS Cloud links recent price increases to advanced GPU servers becoming 30% to 50% more expensive and to a tenfold increase in the price of memory. Reg.Cloud, a cloud provider, says the average monthly cost of dedicated GPU infrastructure rose 64% to RUB 2.3 million in the first five months of 2026, compared with the same period of 2025. DC Market Insights builds an all-in spend of USD 36.00 million per MW into 2025, a premium that reflects these costs.

Moscow Data Centers Are About 95% Full

Data centers in Moscow are about 95% full, and free capacity in the region does not exceed 1,400 rack spaces, according to an adviser to the governor of the Vladimir region in February 2026. A government forecast expects the power consumption of Russian data centers to grow at least 2.5 times to 2.5 gigawatts (GW) by 2030. Because the Central Federal District holds 64.8% of 2025 AI spending, lack of power near Moscow is the main reason DC Market Insights forecasts its share to fall to 55.4% by 2035.

High Borrowing Costs Have Paused Projects

The Bank of Russia key rate reached 21% in 2025, and the bank held it at 14% in September 2026. Some 38 data center construction projects worth about RUB 168.6 billion have been paused in the last three years, according to a Forbes Russia study reported by The Moscow Times. DC Market Insights reflects this in the low scenario, where new AI capacity runs 20% below the base case.

Market Opportunities: Where Are the New Revenue Pools?

AI colocation is the largest new revenue pool, growing to USD 1.80 billion by 2035. The second is AI capacity built away from Moscow near cheap generation; colocation starts from USD 282.68 million in 2025.

AI Colocation Grows Fastest at 20.37%

Few banks and enterprises can find power for their own 200 MW sites, so more of them will lease space in campuses built by colocation operators. IXcellerate commissions about 2,000 rack spaces a year, and its two Moscow campuses were expected to reach 10,329 rack spaces at the end of 2025. Cloud.ru, which used 58 MW of racks in the first half of 2026, is building its own 11 MW data center for completion in 2027. Reg.Cloud estimates that the bare-metal GPU server rental market in Russia reached RUB 17 billion and could double to RUB 34 billion in 2026; each rented GPU server first has to be bought and housed, which feeds this market.

Siberia and Nuclear Sites Offer Large Blocks of Power

The Ministry of Energy has proposed building large data centers in Siberia, where the energy base of the eastern part of the region allows data centers of up to 200 MW each. Rosatom’s Atomdata operates a 48 MW, 4,800-rack data center at the Kalinin nuclear power plant, where electricity cost about RUB 3 per kilowatt-hour in 2022, and in September 2026 Rosatom presented a concept for an AI data center powered by a floating nuclear power unit and cooled with seawater. DC Market Insights forecasts the Siberian Federal District to grow at 23.28% a year, the fastest of any federal district.

Pricing: How Much Does AI Data Center Capacity Cost in Russia?

An AI hall in Russia costs about USD 36.00 million per MW in 2025. That DC Market Insights estimate covers servers, networking, storage, power, cooling and construction. Building shell and engineering alone cost far less: Cloud.ru’s 11 MW data center is estimated to cost RUB 8 billion to RUB 12 billion for engineering infrastructure and construction.

DC Market Insights model input	2020	2025	2030	2035
New AI IT capacity (MW)	9	58	173	273
All-in spend per MW (USD million)	39.50	36.00	31.23	27.10
Replacement share of market	1.1%	6.9%	13.9%	23.3%

Rental prices show what buyers pay to use the capacity. Selectel lists an NVIDIA A100 80 GB GPU at RUB 292.02 per hour and an NVIDIA H200 141 GB GPU at RUB 587.67 per hour, or RUB 429,000 a month. Rack rental at Atomdata’s data center in Innopolis, Tatarstan, ranged from RUB 65,000 to RUB 170,000 a month in January 2025. DC Market Insights expects spend per MW to fall about 2.8% a year from 2026 as newer servers deliver more compute per kilowatt.

Standards and Regulation: Which Rules Shape AI Data Centers in Russia?

Decree No. 124 sets a 1 exaflop supercomputer target for 2030. Five sets of rules shape where and how AI data centers are built in Russia: the national AI strategy, the data localization law, foreign export controls, regional limits on cryptocurrency mining and the registry of Russian electronics.

National AI Development Strategy (decree No. 124, February 2024). Besides the 1 exaflop supercomputer target, the decree sets a target for AI to add at least RUB 11.2 trillion to gross domestic product in 2030 and for at least 80% of workers to have AI skills. The strategy also targets annual AI services of at least RUB 60 billion by 2030, up from RUB 12 billion in 2022.

Federal Law No. 242-FZ (data localization). In force since 1 September 2015, the law requires databases that record, store and process the personal data of Russian citizens to be located in Russia. In the view of DC Market Insights, the law keeps AI models trained on customer data on domestic infrastructure.

US and EU export controls. The BIS rule of June 2024 widened license requirements for Russia and Belarus, and the European Union and its partners expanded their Common High Priority list to 50 customs codes in February 2024.

Cryptocurrency mining limits. Mining is banned entirely in 10 Russian regions from 1 January 2025 until 15 March 2031, and restricted in parts of three Siberian regions during winter consumption peaks.

Registry of Russian electronics. The Ministry of Industry and Trade (Minpromtorg) keeps a register of Russian radio-electronic products, and Yadro’s G4208P G3 AI server, which supports up to eight GPU accelerators, has been added to it. A draft plan known as “Guillotine 2.0”, prepared by the Analytical Center under the Government, contains more than 20 initiatives for data centers, including simpler construction rules and the removal of excessive energy restrictions.

Market Segmentation: Which Segment Leads the Russia AI Data Center Market?

AI servers and accelerators lead with 63.4% of 2025 value. Hyperscale and cloud providers lead by data center type with 54.2%, and training leads by workload with 57.0%. Colocation is the fastest-growing segment at 20.37% a year, because operators that cannot secure their own power in the Moscow region lease AI-ready space instead.

By Component: AI Servers and Accelerators Lead with 63.4%

Component	2025 (USD billion)	Share 2025	2035 (USD billion)	CAGR 2025-2035
AI Servers and Accelerators	1.42	63.4%	5.80	15.09%
Networking and Interconnect	0.22	9.7%	1.10	17.59%
Storage	0.16	7.3%	0.76	16.62%
Power and Cooling Infrastructure	0.27	12.1%	1.31	17.07%
Facility Construction	0.17	7.5%	0.68	14.91%

Networking grows fastest among components at 17.59% a year because larger clusters need more switch ports and optical links per GPU.

By Data Center Type: Hyperscale and Cloud Providers Lead with 54.2%

Data center type	2025 (USD billion)	Share 2025	2035 (USD billion)	CAGR 2025-2035
Hyperscale and Cloud Provider	1.22	54.2%	4.92	15.00%
Enterprise and Bank-Owned	0.53	23.8%	1.99	14.05%
Colocation	0.28	12.6%	1.80	20.37%
Government and Research Supercomputing	0.21	9.4%	0.94	16.07%

Hyperscale and cloud providers such as Yandex and Cloud.ru buy most AI infrastructure, DC Market Insights estimates. Colocation gains 6.1 points of share by 2035.

By Workload: Training Leads with 57.0%

Training clusters account for USD 1.28 billion in 2025 and reach USD 4.00 billion by 2035, growing 12.09% a year. Inference infrastructure grows from USD 964.71 million to USD 5.64 billion, at 19.32% a year, and becomes the larger workload before 2030.

By End User: Internet and IT Services Lead with 41.6%

End user	2025 (USD billion)	Share 2025	2035 (USD billion)	CAGR 2025-2035
Internet and IT Services	0.93	41.6%	3.59	14.42%
BFSI	0.61	27.3%	2.42	14.74%
Telecom	0.24	10.8%	1.12	16.54%
Government and Research	0.22	9.6%	0.99	16.52%
Industrial and Energy	0.14	6.2%	0.93	20.88%
Others	0.10	4.5%	0.60	19.47%

BFSI is unusually large in Russia because Sber invests more in AI than any other company with a disclosed budget, DC Market Insights estimates. Industrial and energy users grow fastest at 20.88% a year as oil, gas, metals and power companies deploy AI on their own sites.

DC Exclusive: The AI Capacity Cost Stack by Data Center Type and Component

Hyperscale AI servers are the largest cell at USD 805.19 million in 2025. That cell holds 35.9% of all AI data center spending in Russia.

2025, USD million	Servers and Accelerators	Networking	Storage	Power and Cooling	Construction	Total
Hyperscale and Cloud Provider	805.19	125.89	80.67	127.34	76.90	1,215.99
Enterprise and Bank-Owned	329.69	44.67	47.79	67.79	44.02	533.96
Colocation	156.34	24.82	17.09	50.22	34.21	282.68
Government and Research Supercomputing	131.17	22.24	18.23	26.12	13.13	210.89
Total	1,422.39	217.62	163.78	271.47	168.26	2,243.52

Hyperscale and cloud providers spend 66.2% of their AI budget on servers, while colocation operators spend 29.9% on power, cooling and construction, because in a colocation hall the tenant buys the servers and the operator builds the shell.

The matrix is a DC Market Insights model, built from 54 inputs: 16 annual AI capacity cohorts (2020 to 2035), 16 points on the spend-per-MW curve, 2 replacement parameters and 20 type-by-component cost mixes. The model and its assumptions are explained on our research methodology page.

Regional Insights: Where Are the Largest AI Data Centers in Russia?

The Central Federal District leads with 64.8% of 2025 value. The district includes Moscow and the Moscow region. The Siberian Federal District is the fastest-growing region at 23.28% a year because it can offer large blocks of power that Moscow cannot.

Central Federal District Leads with 64.8% Share

The Central Federal District accounts for USD 1.45 billion in 2025 and USD 5.35 billion by 2035, a CAGR of 13.91%. The largest AI projects sit here: IXcellerate’s 130 MW Veshki campus, Yandex’s planned RUB 16 billion data center in Mozhaisk and Cloud.ru’s 11 MW build in the Moscow region. The district loses 9.4 points of share by 2035 because the Moscow grid is close to its limit.

Northwestern Federal District Holds 13.7% Share

The Northwestern Federal District, centered on St. Petersburg, is worth USD 307.36 million in 2025 and USD 1.34 billion by 2035, growing 15.87% a year. Two new St. Petersburg data centers are planned to offer 22 MW between them, including 15 MW at a Vyborgsky site that could go live in early 2028.

Volga Federal District Holds 9.1% Share

The Volga Federal District is worth USD 204.16 million in 2025 and USD 1.08 billion by 2035, growing 18.13% a year. Rostelecom’s Nizhny Novgorod data center, with 401 racks and 5 MW, was 93% contracted within a year, and the Innopolis data center in Tatarstan opened with 1,000 racks and 8 MW in its first phase.

Siberian Federal District Holds 5.2% Share

The Siberian Federal District is worth USD 116.66 million in 2025 and grows fastest to USD 945.56 million by 2035. The Ministry of Energy’s proposal for data centers of up to 200 MW in eastern Siberia and the winter limits on mining show both the power available and the need to plan for it.

Ural Federal District Holds 4.6% Share

The Ural Federal District is worth USD 103.20 million in 2025 and USD 588.56 million by 2035, growing 19.02% a year. Yekaterinburg is gaining capacity: Key Point’s data center there has a design capacity of 11 MW and 880 IT racks, Rostelecom’s Yekaterinburg capacity reached 6 MW and 644 rack spaces, and MegaFon opened a 1 MW data center in the city in July 2025.

Southern, North Caucasian and Far Eastern Districts Hold 2.6% Share

The remaining districts together are worth USD 58.33 million in 2025 and USD 347.35 million by 2035, growing 19.53% a year from a small base of regional edge and telecom sites.

Forecast Scenarios: How Could the 2035 Forecast Change?

The base case reaches USD 9.65 billion by 2035. The range runs from USD 6.82 billion in the low scenario to USD 12.11 billion in the high scenario.

Scenario	Assumption from 2026	2035 (USD billion)	CAGR 2025-2035
Low	New AI capacity 20% below base; spend per MW falls 4.0% a year	6.82	11.76%
Base	New AI capacity rises to 273 MW a year; spend per MW falls 2.8% a year	9.65	15.71%
High	New AI capacity 18% above base; spend per MW falls 2.2% a year	12.11	18.36%

The low case reflects tighter supply of accelerators, a slower cut in interest rates and further delays to grid connections in the Moscow region. The high case assumes large AI campuses in Siberia and next to power plants come online on time and that Sber’s 2026 investment step is sustained.

Competitive Insights: Who Owns the Largest AI Data Centers in Russia?

• Sber
• Yandex
• Cloud.ru
• MTS Web Services (MWS)
• Rostelecom Data Centers (RTK-DPC)
• Selectel
• IXcellerate
• Key Point
• MegaFon
• Atomdata (Rosatom)
• Yadro
• T-Bank

Sber is the largest buyer with about 40% of the Russia AI Data Center Market in 2025, and Sber and Yandex together account for about 62% (DC Market Insights estimate: half of Sber’s RUB 150 billion of 2025 AI investment and about RUB 40 billion of Yandex AI infrastructure spending, set against a market of about RUB 185 billion). Sber’s original Christofari and Christofari Neo supercomputers, built on NVIDIA V100 and A100 GPUs, still appear on the TOP500 list. Yandex runs the Chervonenkis, Galushkin and Lyapunov supercomputers with 3,776 NVIDIA A100 GPUs between them. Cloud.ru earned RUB 76.5 billion of revenue in 2025, up 50%, and its AI services and infrastructure made up 51.3% of revenue in the first half of 2026. MTS Web Services committed more than RUB 1 billion to a GPU-based AI cloud by the end of 2025, and MTS’s GROM supercomputer delivers 2.26 petaflops. Selectel spent RUB 7.9 billion of capital expenditure in 2025, including RUB 4.7 billion on servers. Rostelecom’s data center arm has invested more than RUB 38 billion over five years and runs 27,637 racks. IXcellerate, Key Point, MegaFon and Atomdata supply the colocation space, and Yadro supplies domestic AI servers. T-Bank has said that even building two 50 MW data centers is a difficult task, which shows how scarce large power blocks are.

Recent Developments

• In April 2025, local media reported that Yandex would invest about RUB 16 billion in a data center in Mozhaisk, with launch targeted for 2027 (Source: DatacenterDynamics news report).
• In September 2025, Yandex Cloud said it would launch a new availability zone in 2026 based on a new Yandex data center of more than 40 MW (Source: Yandex Cloud blog).
• In November 2025, IXcellerate announced its Veshki campus in the Moscow region with 7,500 rack spaces and more than 130 MW, including a 40 MW first facility with racks scalable to 44 kW (Source: IXcellerate news release).
• In November 2025, Sber said it intends to raise AI investment 2.3 times in 2026 to RUB 350 billion (Source: SberBusiness news report).
• In March 2026, the government approved a roadmap to raise the combined capacity of Russian supercomputers at least tenfold (Source: ComNews report).
• In June 2026, Cloud.ru began building an 11 MW data center in the Moscow region for up to 890 racks at 20 kW per rack, with commissioning expected in 2027 (Source: DatacenterDynamics news report).

We follow these projects, investments and policy changes as they happen in our data center industry updates. For the global picture, see the AI Data Center Market report.

Methodology: How We Built the Russia AI Data Center Market Model

DC Market Insights built the model from 54 inputs. The market is built bottom-up from 16 annual cohorts of new AI IT capacity (2020 to 2035), from 9 MW in 2020 and 58 MW in 2025 to 273 MW in 2035, each priced with an all-in spend per MW that falls from USD 39.50 million in 2020 to USD 36.00 million in 2025 and USD 27.10 million in 2035. Replacement spending was added for each cohort five years after installation, at 60% of its hardware value (80% of spend per MW). The total was split by 5 components, 4 data center types, 2 workloads, 6 end users and 6 federal district groups, and every split sums to the market total. The result was checked top-down against the AI spending disclosed by Sber (RUB 150 billion in 2025, of which about half is attributed to infrastructure) and Yandex (RUB 55 billion for Alice AI and RUB 18 billion for autonomous technology), the capital expenditure of Selectel and Cloud.ru, and the estimate by MWS Cloud that the market for GPU-based solutions deployed on customers’ own premises exceeds RUB 112 billion. All rouble values are converted at a planning rate of RUB 82.50 per US dollar. The analyst named on this page built the model, and the reviewer checked the arithmetic, the sources and the dates before publication.