Country Reports · Forecast 2026-2035

South Korea AI Data Center Market Size, Share & Forecast to 2035

Market at a glance

The South Korea AI Data Center market was worth USD 1,005.8 million in 2025 and is forecast to reach USD 5,823.6 million by 2035, growing at a 19.2% CAGR.

$1.01BMarket size, 2025
$5.82BForecast, 2035
19.2%CAGR, 2025-2035
$35.4MMarket size, 2020
Market size, USD million
35.4 2020 1,005.8 2025 5,823.6 2035F

What is inside the report

  1. 01Introduction
  2. 02Research Methodology
  3. 03Executive Summary
  4. 04Market Dynamics
  5. 05Pricing Analysis
  6. 06Standards and Regulation
  7. 07Market Segmentation
  8. 08DC Exclusive: Korea GPU-to-Megawatt Tracker
  9. 09Regional Analysis
  10. 10Province and City Analysis
  • 13chapters
  • 2020-2024Historical period
  • 2026-2035Forecast period
  • —Companies profiled

Full table of contents

Segmentation

Component
Electrical InfrastructureCooling InfrastructureBuilding Shell Fit-Out and RacksAI Network FabricOperations Services and Software
Workload
AI TrainingAI Inference
Data Center Type
Hyperscale and Cloud Self-BuildTelecom Operator AI Data CentersColocation and GPU Cloud CampusesGovernment Research and Enterprise AI

Full analysis

The South Korea AI Data Center Market is valued at USD 1.74 billion in 2026 and is projected to reach USD 5.82 billion by 2035, up from USD 1.01 billion in 2025, a CAGR of 19.20% from 2025 to 2035 (DC Market Insights estimate). The money in Korea follows graphics processing unit (GPU) commitments: at the Asia-Pacific Economic Cooperation (APEC) summit in Gyeongju in October 2025, NVIDIA Corporation (NVIDIA) said Korean government and industry partners would add more than 260,000 of its GPUs, including over 50,000 for the National AI Computing Center that the Ministry of Science and ICT (MSIT) is setting up outside the Seoul capital area (Source: NVIDIA and MSIT). Every one of those GPUs needs a powered, liquid-cooled hall, and DC Market Insights converts the 260,000 GPUs into about 468 MW of IT load at 1.8 kW per GPU, close to the 507 MW of AI capacity the model completes in 2026-2028. The South Korea AI Data Center Market covers the facility infrastructure built and run for artificial intelligence (AI) workloads in South Korea: electrical systems, cooling, the building shell, fit-out and racks, the AI network fabric, and operations services and software; GPU servers, chips and electricity bills are excluded. The 2025 build receipt is 50% of 57.5 MW of new AI IT capacity completed in 2025 × USD 11.00 million per AI MW (USD 316.25 million) plus 50% of 114.0 MW due in 2026 × USD 11.22 million (USD 639.54 million), plus USD 0.40 million of services on each of 125.0 MW of installed AI capacity (USD 50.00 million). Operations Services and Software is the fastest-growing segment at 38.64% a year as the installed AI base climbs from 125 MW to 2,624 MW, and the Honam Region is the fastest-growing region at 31.48% a year because SK Telecom’s southwest AI data center hub and the non-capital siting rule for the national center pull new halls toward Jeollanam-do and Gwangju.

Key Market Statistics: How Big Is the South Korea AI Data Center Market in 2026?

DC Market Insights values the South Korea AI Data Center Market at USD 1.74 billion in 2026, 72.72% above 2025. The jump comes as new AI IT capacity doubles from 57.5 MW to 114.0 MW while the GPU halls announced in 2025 move into construction.

  • Market size 2025: USD 1.01 billion
  • Market size 2026: USD 1.74 billion
  • Market size 2030: USD 4.04 billion
  • Market size 2035: USD 5.82 billion
  • CAGR 2025-2035: 19.20% (low case 16.27%, high case 21.29%)
  • Largest component: Electrical Infrastructure, 38.01% of 2025 spend (USD 382.32 million)
  • Fastest component: Operations Services and Software, 38.64% a year; fastest workload: AI Inference, 25.02% a year
  • Leading region: Capital Region (Seoul, Gyeonggi and Incheon), 60% of 2025 spend; fastest-growing region: Honam Region, 31.48% a year
  • Largest province: Gyeonggi Province, USD 301.74 million of 2025 spend
  • Market leader: SK Group (SK Telecom and SK Broadband), about 24% of 2025 AI facility spend (DC Market Insights estimate)

Key Takeaways: What Do 260,000 GPUs and a 5 GW Telecom Plan Mean for Korean AI Hall Buyers?

DC Market Insights draws four takeaways for buyers in the South Korea AI Data Center Market, which grows from USD 1.01 billion in 2025 to USD 5.82 billion in 2035.

  • 36% of 2035 spend stays in the Capital Region, down from 60%, because power system impact assessments under the Distributed Energy Act slow new large loads near Seoul.
  • USD 11.00 million buys one AI megawatt of facility before servers, because 100 kW racks need direct-to-chip liquid loops and heavier power trains.
  • 58% of 2035 spend goes to AI Inference because Korean banks, carmakers and portals run models after the 2025-2028 training clusters are built.
  • 23.28% a year is the growth rate of Colocation and GPU Cloud Campuses because investors such as Kohlberg Kravis Roberts (KKR) and IMM Investment (IMM) now fund telecom data center spin-offs.

South Korea AI Data Center Market Summary: What Did Korean AI-Ready Halls Cost in 2025?

The South Korea AI Data Center Market size was valued at USD 35.38 million in 2020, reached USD 1.01 billion in 2025, and is anticipated to reach USD 5.82 billion by 2035, at a CAGR of 19.20% during the forecast period. DC Market Insights estimates that South Korea spent USD 1.01 billion on AI data center facilities in 2025, up from USD 515.39 million in 2024.

Report attribute Details
Historical period 2020-2024
Base year 2025
Forecast period 2026-2035
Market size 2025 USD 1,005.80 million
Market size 2026 USD 1,737.26 million
CAGR 2025-2035 19.20%
Market size 2035 USD 5,823.60 million
Segments covered Component, Workload, Data Center Type
Regions covered Capital Region, Chungcheong Region, Yeongnam Region, Honam Region, Gangwon and Jeju; 12 provinces and cities

Growth runs at 72.72% in 2026 and 40.87% in 2027, then slows to 6.24% by 2035 because yearly AI completions level off near 330 MW once the GPU commitments of 2025 are housed, while the AI share of new Korean data center capacity rises only from 60% in 2030 to 65% in 2035. South Korea’s installed data center IT load climbs from 1,547 MW at the end of 2025 to 3,294 MW in 2030 and 5,269 MW in 2035 in the DC Market Insights model; the AI part rises from 125 MW to 2,624 MW, or from 8% to 50% of the base. The AI facility spend modelled here is a slice of the South Korea Data Center Market, which also counts conventional halls and IT equipment.

South Korea AI Data Center Market size chart: USD 35.38 million in 2020, USD 1.01 billion in 2025 and USD 5.82 billion by 2035, a 19.20% CAGR from 2025 to 2035 (DC Market Insights)

What Does the South Korea AI Data Center Market Cover, and Where Does Colocation Rent Fit?

The South Korea AI Data Center Market is USD 1.01 billion of facility spend in 2025 on halls that host AI training and inference clusters, plus the services that keep those halls running.

Included

Medium-voltage substations on campus, switchgear, uninterruptible power supply (UPS) systems, busway and rack power distribution; chillers, dry coolers, coolant distribution units (CDUs), rear-door heat exchangers and direct-to-chip liquid loops; the building shell, slab floors, containment and racks rated above 30 kW; the AI network fabric (spine-leaf cabling, optical trunks and in-hall switching enclosures); and operations, monitoring and lifecycle services for AI halls. Cooling for all Korean halls, AI and conventional, is sized in the South Korea Data Center Cooling Market.

Excluded

GPU servers, accelerators, high-bandwidth memory and storage; land; grid electricity bills; transmission works paid by Korea Electric Power Corporation (KEPCO); GPU-as-a-service and cloud AI revenue; and colocation rent. Rent is the boundary with the South Korea Data Center Colocation Market: that report counts what tenants pay operators each year, while the AI data center figure counts what owners spend to build and run AI halls, so the two figures must not be added.

How the segments fit together

Each USD of the USD 1.01 billion 2025 total sits once in each dimension: five components, two workloads and four data center types, each adding up to the total. Of the 2025 spend, 95.03% is build capital (USD 955.80 million) and 4.97% is operations services and software (USD 50.01 million); by 2035 the services share rises to 22.53% as the installed AI base grows 21 times.

Why Do South Korea AI Data Center Market Estimates Differ?

Estimates differ by up to five times because some definitions count the GPUs, while DC Market Insights counts only the facility, which comes to USD 1.74 billion in 2026. SK Telecom puts the full bill for a 1 GW AI data center at about KRW 70 trillion, or about USD 50 million per MW at KRW 1,400 per USD (Source: SK Telecom, July 2026); the facility layer modelled here is USD 11.00-11.22 million per MW, so servers, chips and memory make up roughly 78% of the full bill and sit outside the DC Market Insights scope.

Broader definitions also add electricity, GPU-as-a-service revenue and AI cloud subscriptions. SK Telecom alone reported AI data center revenue of KRW 136.2 billion in the second quarter of 2026, up 92.5% a year (Source: SK Telecom), which is service revenue, not facility spend. Some studies also count every hyperscale hall as “AI-ready” whatever it hosts; DC Market Insights counts only halls built for AI training and inference racks.

Element In this report Why
Power, cooling, shell, racks, AI network fabric Yes, USD 955.80 million build capital in 2025 Facility spend for AI halls
Operations services and software for AI halls Yes, USD 50.01 million in 2025 Recurring facility spend
GPU servers, accelerators, memory No IT equipment, about 78% of a full AI data center bill
Electricity bills No Utility cost, set by KEPCO tariffs
GPU-as-a-service and colocation rent No Revenue of operators, sized in other reports

The out-of-scope terms in prose: GPU servers, AI accelerators, high-bandwidth memory, electricity, GPU-as-a-service, AI cloud services and colocation rent are all excluded. In-scope buyers include hyperscale clouds, the three telecom operators, colocation and GPU cloud developers, government research institutes and universities, and small and medium-sized enterprises (SMEs) that rent AI racks through GPU cloud campuses, whose halls are counted through the developer.

Why Are Telecom Operators and Chaebol Building South Korea AI Data Centers So Fast?

Three drivers lift South Korea AI data center facility spend from USD 1.01 billion in 2025 to USD 4.04 billion in 2030, each tied to a dated commitment.

The 260,000-GPU APEC commitment needs about 468 MW of halls

In October 2025 NVIDIA said Korean partners would add more than 260,000 of its GPUs: over 50,000 each for the National AI Computing Center, Samsung’s semiconductor AI factory, SK Group’s AI factory and Hyundai Motor Group, and over 60,000 for NAVER Corporation’s (NAVER) cloud arm, NAVER Cloud (Source: NVIDIA). At 1.8 kW per GPU, including networking and storage in the rack, DC Market Insights estimates that these orders need about 468 MW of AI IT load, worth about USD 5.25 billion of facility spend at USD 11.22 million per MW. The model completes 507.2 MW of AI capacity in 2026-2028, so the GPU orders alone fill 92% of it.

MSIT’s National AI Computing Center adds a public anchor tenant outside Seoul

MSIT plans to secure more than 50,000 GPUs through public-private collaboration and to add over 15,000 GPUs at the National AI Computing Center by 2028, with total investment above KRW 2 trillion and a government contribution of KRW 80 billion by 2030 (Source: Ministry of Science and ICT, September 2025). The center must sit outside the Seoul metropolitan area, which is one reason DC Market Insights gives government research and enterprise AI halls 14% of 2025 spend and shifts their growth toward Honam and Chungcheong.

SK Telecom’s 5 GW domestic plan turns a telecom into a campus builder

SK Telecom said in July 2026 that it is pursuing up to 15 GW of AI data center capacity, with 5 GW in Korea opening in stages from 2029, and plans to expand its Ulsan AI data center, built with Amazon Web Services (AWS) from 2027, to 1 GW (Source: SK Telecom). DC Market Insights does not assume all 5 GW is built by 2035: the model adds 2,608 MW of AI capacity in 2026-2035 for the whole country, about half of SK Telecom’s stated domestic ambition, because grid assessments and tenant contracts pace the build.

Two shifts change what a Korean AI hall contains, and both raise spend per megawatt from USD 11.00 million in 2025 to USD 13.41 million in 2035.

Direct-to-chip cooling becomes standard for Blackwell-class racks

Cooling Infrastructure grows 18.94% a year, faster than the 15.88% of electrical work, because racks for 50,000-GPU clusters run at 100 kW or more and need CDUs and liquid loops. DC Market Insights lifts cooling from 25% of 2025 build capital to 30% of 2035 build capital.

Telecom data center spin-offs bring private equity into AI campuses

In August 2026 a KKR and IMM consortium agreed a KRW 3.08 trillion equity investment in SK Horizon, the data center business carved out of SK Broadband, which will expand to 318 MW of total capacity (Source: SK Telecom). Outside capital lets operators build ahead of contracts, which is why Colocation and GPU Cloud Campuses rise from 20% to 28% of spend between 2025 and 2035.

Inference halls follow the 2025-2028 training wave

AI Training carries 64% of 2025 spend, but AI Inference grows 25.02% a year to 58% of 2035 spend as NAVER, Samsung, Hyundai and banks put models into production. Inference halls sit closer to users, which keeps the Capital Region at USD 2.10 billion of 2035 spend despite its falling share.

What Holds Back South Korea AI Data Center Construction: Grid Rules or Power Prices?

Grid access and power prices are the two brakes, and DC Market Insights puts their combined drag at the gap between the USD 5.82 billion base case and the USD 6.93 billion high case in 2035.

Power system impact assessments under the Distributed Energy Act

The Special Act on the Activation of Distributed Energy (Act No. 19437, June 2023, in force from June 2024) requires power system impact assessments for large electricity users in designated areas (Source: Korea Legislation Research Institute). A 100 MW AI hall near Seoul now waits on that review, so the model cuts the Capital Region’s share of AI spend from 60% in 2025 to 36% in 2035.

Industrial electricity prices rose 42% in two years

KEPCO’s average industrial electricity price rose from 118.66 KRW/kWh in 2022 to 168.17 KRW/kWh in 2024, a 41.7% increase (Source: KEPCO). At 168.17 KRW/kWh, a 100 MW AI hall running at 80% load and a power usage effectiveness (PUE) of 1.3 pays about KRW 153.21 billion (USD 109.4 million) a year for power, which pushes owners toward liquid cooling and lower PUE designs rather than cheaper air-cooled halls.

Where Will South Korea AI Data Center Suppliers Earn the Most After 2027?

Two supplier pools grow fastest, together worth USD 1.85 billion of 2035 spend.

Operations and lifecycle services for a 2,624 MW AI base

Operations Services and Software grows from USD 50.01 million of 2025 spend to USD 1.31 billion in 2035, 38.64% a year, as services of USD 0.40-0.50 million per installed AI MW apply to a base that rises from 125 MW to 2,624 MW. Facility managers, liquid-cooling maintenance firms and data center infrastructure management (DCIM) software vendors gain most.

AI network fabric inside 50,000-GPU halls

AI Network Fabric rises from USD 105.14 million of 2025 spend to USD 541.41 million in 2035, 17.81% a year, because each training cluster needs dense optical trunks and spine-leaf cabling. Korean cable makers and optical module suppliers can sell into halls that NVIDIA’s October 2025 list places with Samsung, SK, Hyundai and NAVER.

How Much Does a Megawatt of AI Data Center Capacity Cost in South Korea?

Each megawatt of AI facility capacity cost about USD 11.00 million before servers for halls finished in 2025, and DC Market Insights lifts that price 2% a year to USD 13.41 million per MW for 2035 completions.

Year Facility cost per AI MW (USD million) Services per installed AI MW per year (USD million) New AI IT capacity (MW)
2020 9.96 0.35 2.4
2025 11.00 0.40 57.5
2030 12.14 0.45 282.0
2035 13.41 0.50 331.5

Public price anchors back the band. SK Telecom’s KRW 70 trillion per GW for a complete AI data center equals about USD 50 million per MW at KRW 1,400 per USD, of which the facility layer modelled here is about 22%. Power is the main running cost: KEPCO’s industrial price of 168.17 KRW/kWh in 2024 makes electricity worth about USD 1.09 million per MW of AI IT load a year at 80% load and a PUE of 1.3.

Which Korean Laws and Rules Shape AI Data Center Siting and Design?

Four rules shape where Korea’s USD 1.74 billion of 2026 AI facility spend lands.

  • The Framework Act on the Development of Artificial Intelligence and Establishment of Trust (AI Basic Act), passed on 26 December 2024, came into force on 22 January 2026 with a grace period of at least one year for penalties; its support measures include the establishment of AI data centers (Source: Ministry of Science and ICT).
  • The Special Act on the Activation of Distributed Energy requires power system impact assessments for large loads from June 2024 (Source: Korea Legislation Research Institute).
  • MSIT’s National AI Computing Center rules require a site outside the Seoul metropolitan area and target a 50% share of domestic AI chips in the center by 2030 (Source: MSIT, February and September 2025).
  • Public-sector AI workloads must run in halls certified under the Cloud Security Assurance Program (CSAP) run by the Korea Internet and Security Agency (KISA), and operators hold Information Security Management System (ISMS-P) and ISO/IEC 27001 certification, which adds compliance work to operations services.

Which South Korea AI Data Center Segment Leads by Component, Workload and Data Center Type?

Electrical Infrastructure leads the South Korea AI Data Center Market with 38.01% of 2025 spend, AI Training leads workloads with 64%, and Hyperscale and Cloud Self-Build leads data center types with 36%. Operations Services and Software is the fastest-growing segment at 38.64% a year because services scale with the installed AI base, not with yearly completions.

South Korea AI Data Center Market by Component

Component 2025 (USD million) Share 2025 2035 (USD million) CAGR 2025-2035
Electrical Infrastructure 382.32 38.01% 1,669.34 15.88%
Cooling Infrastructure 238.95 23.76% 1,353.52 18.94%
Building Shell Fit-Out and Racks 229.39 22.81% 947.46 15.24%
AI Network Fabric 105.14 10.45% 541.41 17.81%
Operations Services and Software 50.01 4.97% 1,311.87 38.64%
Total 1,005.80 100% 5,823.60 19.20%

South Korea AI Data Center Market by Workload

Workload 2025 (USD million) Share 2025 2035 (USD million) CAGR 2025-2035
AI Training 643.71 64.00% 2,445.91 14.28%
AI Inference 362.09 36.00% 3,377.69 25.02%
Total 1,005.80 100% 5,823.60 19.20%

South Korea AI Data Center Market by Data Center Type

Data center type 2025 (USD million) Share 2025 2035 (USD million) CAGR 2025-2035
Hyperscale and Cloud Self-Build 362.09 36.00% 2,445.91 21.05%
Telecom Operator AI Data Centers 301.74 30.00% 1,048.25 13.26%
Colocation and GPU Cloud Campuses 201.16 20.00% 1,630.61 23.28%
Government Research and Enterprise AI 140.81 14.00% 698.83 17.37%
Total 1,005.80 100% 5,823.60 19.20%

Telecom Operator AI Data Centers hold 30% of 2025 spend because SK Telecom, KT Corporation (KT) and LG Uplus own most of Korea’s early GPU halls, but their share falls to 18% by 2035 as hyperscale clouds and investor-funded campuses build larger sites.

DC Exclusive: The Korea GPU-to-Megawatt Tracker, 260,000 GPUs Turned Into 468 MW of AI Halls

For the South Korea AI Data Center Market, DC Market Insights built this tracker from 14 inputs: five GPU commitments announced in October 2025 total 260,000 GPUs, which convert to about 468 MW of AI IT load and about USD 5.25 billion of facility spend at 1.8 kW per GPU and USD 11.22 million per MW; the conversion factors are DC Market Insights estimates.

GPU commitment GPUs IT load (MW) Facility value (USD million) Named site or timing
National AI Computing Center (MSIT) 50,000 90 1,009.80 Outside the Seoul capital area; 15,000+ GPUs by 2028
Samsung semiconductor AI factory 50,000 90 1,009.80 Samsung fab sites
SK Group AI factory 50,000 90 1,009.80 Ulsan AI data center with AWS, operations from 2027
Hyundai Motor Group 50,000 90 1,009.80 Not named
NAVER Cloud 60,000 108 1,211.76 NAVER’s Gak Sejong data center (GAK Sejong): 55 MW in H1 2027, 200 MW by 2028
Total 260,000 468 5,250.96 Against 507.2 MW of AI completions modelled for 2026-2028

The tracker explains the 2026-2028 curve: the GPU orders fill 92% of the AI capacity the model completes in those three years, so the forecast does not rely on unannounced demand until 2029, when SK Telecom’s staged 5 GW domestic plan starts. NAVER’s GAK Sejong plan of 55 MW in the first half of 2027 and 200 MW by 2028 (Source: NAVER, June 2026) is the largest single named AI hall; it explains why Sejong City carries 6.30% of 2025 spend.

A second cut splits 2025 spend by data center type and workload, both of which sum to the yearly total.

Data center type (2025, USD million) AI Training AI Inference Total
Hyperscale and Cloud Self-Build 224.49 137.60 362.09
Telecom Operator AI Data Centers 199.15 102.59 301.74
Colocation and GPU Cloud Campuses 140.81 60.35 201.16
Government Research and Enterprise AI 79.26 61.55 140.81
Total 643.71 362.09 1,005.80

Seoul or Jeollanam-do: Where Is South Korea AI Data Center Spending Moving by Region?

The Capital Region leads with 60% of 2025 spend (USD 603.48 million), and the Honam Region is the fastest-growing region at 31.48% a year because SK Telecom’s southwest hub and the non-capital rule for the national AI center send new campuses to Jeollanam-do and Gwangju.

Region 2025 (USD million) Share 2025 2035 (USD million) CAGR 2025-2035
Capital Region 603.48 60.00% 2,096.50 13.26%
Chungcheong Region 140.81 14.00% 757.07 18.32%
Yeongnam Region 120.70 12.00% 1,222.96 26.06%
Honam Region 90.52 9.00% 1,397.66 31.48%
Gangwon and Jeju 50.29 5.00% 349.42 21.39%
Total 1,005.80 100% 5,823.60 19.20%

Capital Region (Seoul, Gyeonggi and Incheon)

SK Telecom launched its H100-based GPU-as-a-service at its Gasan AI data center in Seoul in January 2025 (Source: SK Telecom), and most early telecom and cloud GPU halls sit in Gyeonggi and Seoul. The region still adds USD 1.49 billion of yearly spend by 2035 but its share falls to 36% as grid reviews slow new 100 MW sites.

Chungcheong Region (Sejong, Daejeon and Chungcheongbuk-do)

NAVER opened GAK Sejong in November 2023 with a design of up to 270 MW, and its NVIDIA AI factory there adds 200 MW by 2028 (Source: NAVER). Chungcheong grows 18.32% a year to USD 757.07 million in 2035.

Yeongnam Region (Ulsan, Busan and Gyeongsangbuk-do)

The SK Group and AWS Ulsan AI data center begins operations by 2027, and SK Telecom plans to take it to 1 GW (Source: SK Group and SK Telecom). Yeongnam grows 26.06% a year, and Ulsan rises from USD 36.21 million of 2025 spend to USD 587.02 million in 2035, 32.12% a year.

Honam Region (Jeollanam-do and Gwangju)

SK Telecom names the southwest as its third AI data center hub after Gasan and Ulsan (Source: SK Telecom, November 2025). Jeollanam-do is the fastest-growing province at 37.42% a year, from USD 40.73 million of 2025 spend to USD 978.36 million in 2035, as renewable power and land draw GW-scale campuses.

Gangwon and Jeju

Gangwon-do’s cooler climate and hydro power support smaller AI halls; the region grows 21.39% a year to USD 349.42 million by 2035.

Which Korean Provinces and Cities Host the Most AI Data Center Spend?

Gyeonggi Province hosts the most South Korea AI data center spend with 30% of 2025 spend (USD 301.74 million), and Jeollanam-do grows fastest among large provinces at 37.42% a year.

Rank Province or city Region 2025 (USD million) Share 2025 2035 (USD million) CAGR 2025-2035
1 Gyeonggi Province Capital Region 301.74 30.00% 1,090.18 13.71%
2 Seoul Capital Region 229.32 22.80% 712.81 12.01%
3 Incheon Capital Region 72.42 7.20% 293.51 15.02%
4 Sejong City Chungcheong Region 63.37 6.30% 317.97 17.50%
5 Gangwon-do Gangwon and Jeju 50.29 5.00% 349.42 21.39%
6 Gwangju Honam Region 49.79 4.95% 419.30 23.75%
7 Busan Yeongnam Region 48.28 4.80% 366.89 22.48%
8 Daejeon Chungcheong Region 42.24 4.20% 211.98 17.50%
9 Jeollanam-do Honam Region 40.73 4.05% 978.36 37.42%
10 Ulsan Yeongnam Region 36.21 3.60% 587.02 32.12%
11 Gyeongsangbuk-do Yeongnam Region 36.21 3.60% 269.05 22.21%
12 Chungcheongbuk-do Chungcheong Region 35.20 3.50% 227.12 20.49%

By 2035 Jeollanam-do moves from ninth to second place behind Gyeonggi, the largest shift in the table, because the southwest hub and the national center both need land and power that the capital area cannot offer.

What If Grid Approvals Speed Up or Stall? South Korea AI Data Center Scenarios to 2035

DC Market Insights puts the South Korea AI Data Center Market between USD 4.54 billion and USD 6.93 billion in 2035, around a base case of USD 5.82 billion.

Scenario 2035 (USD million) CAGR 2025-2035 Assumption
Low 4,542.41 16.27% Grid assessments delay a fifth of AI completions after 2028; power prices stay above 168 KRW/kWh
Base 5,823.60 19.20% 260,000 GPUs housed by 2028; about half of SK Telecom’s 5 GW domestic plan built by 2035
High 6,930.08 21.29% Faster KEPCO connections in Honam and Yeongnam; AI share of new capacity reaches 75%

The base case keeps the AI share of new capacity at 65% in 2035; each 5-point change in that share moves 2035 spend by about USD 0.35 billion.

Who Builds and Equips South Korea AI Data Centers, and How Concentrated Is the Spend?

SK Group (SK Telecom and SK Broadband, now SK Horizon) accounts for about 24% of 2025 AI facility spend, and the top five builders hold about 62% (DC Market Insights estimate, based on GPU commitments and named capacity).

  • SK Telecom and SK Broadband (SK Horizon)
  • KT Corporation and KT Cloud
  • NAVER Cloud
  • Samsung SDS and Samsung C&T
  • LG Uplus and LG CNS
  • Kakao
  • Hyundai Motor Group
  • Amazon Web Services (AWS)
  • Microsoft
  • Equinix
  • Digital Realty
  • LS Electric
  • Hyosung Heavy Industries
  • LG Electronics
  • Schneider Electric
  • Vertiv

SK Telecom reported AI data center revenue of KRW 136.2 billion in the second quarter of 2026, up 92.5% from a year earlier (Source: SK Telecom). KT signed a five-year partnership with Microsoft in September 2024 that includes a sovereign cloud for Korea (Source: Microsoft). On the supply side, LS Electric and Hyosung Heavy Industries supply transformers and switchgear, LG Electronics supplies chillers, and Schneider Electric and Vertiv supply UPS systems and CDUs; DC Market Insights estimates that the five largest equipment suppliers hold about 45% of 2025 equipment spend.

Which Dated Announcements Moved the South Korea AI Data Center Market in 2025 and 2026?

Six dated events moved the South Korea AI Data Center Market toward USD 1.74 billion in 2026.

  • In September 2025, MSIT set out plans for more than 50,000 GPUs and over KRW 2 trillion of investment in the National AI Computing Center (Source: MSIT).
  • In October 2025, NVIDIA announced more than 260,000 GPUs for Korean government and industry partners (Source: NVIDIA).
  • In January 2026, the AI Basic Act came into force with support measures for AI data centers (Source: MSIT).
  • In June 2026, NAVER and NVIDIA set out the GAK Sejong AI factory: 55 MW in the first half of 2027 and 200 MW by 2028 (Source: NAVER).
  • In July 2026, SK Telecom set out up to 15 GW of AI data center capacity, 5 GW of it in Korea from 2029 (Source: SK Telecom).
  • In August 2026, KKR and IMM agreed a KRW 3.08 trillion investment in SK Horizon, which will grow to 318 MW (Source: SK Telecom).

Methodology: How Is the South Korea AI Data Center Model Calculated?

The South Korea AI Data Center Market model multiplies megawatts by money: DC Market Insights runs 16 yearly completion cohorts through 14 inputs. The inputs are new IT MW completed in Korea (120 MW in 2020, 230 MW in 2025, 470 MW in 2030, 510 MW in 2035), the same capacity base DC Market Insights uses for Korean cooling demand; the AI share of new capacity (2%, 25%, 60%, 65%); AI facility content of USD 11.00 million per MW in 2025 rising 2% a year, booked 50% in the completion year and 50% in the year before; services of USD 0.35-0.50 million per installed AI MW a year (1% of the base retired a year); five GPU commitments; and KEPCO tariffs at KRW 1,400 per USD.

Two ceilings and one floor test the total. First, the 2026 build capital of USD 1.64 billion equals 29% of the full cost of the 114.0 MW of AI capacity due in 2026 at SK Telecom’s KRW 70 trillion per GW (about USD 5.70 billion), so facility spend stays a minority of the full AI data center bill (ceiling). Second, the 2,624 MW installed AI base in 2035 is about half of SK Telecom’s 5 GW domestic plan, so the model stays below one operator’s stated ambition (ceiling). Third, the 260,000 announced GPUs need about 468 MW of halls, and the model installs 624.5 MW of AI capacity by 2028 (floor). Component, workload, type, region and province tables all add up to the yearly total, and each CAGR comes from the 2025 and 2035 endpoints; method notes sit on the DC Market Insights research methodology page.

Table of contents

  • 13 chapters
  • 1 figures
  • 11 data tables
  1. 01Introduction
    • 1.1Market Definition and Scope
    • 1.2Included and Excluded Elements
    • 1.3How the Segments Fit Together
    • 1.4Why South Korea AI Data Center Market Estimates Differ
  2. 02Research Methodology
    • 2.1Bottom-Up Model: AI MW × Content per MW
    • 2.2Bound Checks
    • 2.3Source Ledger
  3. 03Executive Summary
    • 3.1Market Snapshot, 2025, 2026 and 2035
    • 3.2Key Market Statistics
    • 3.3Key Takeaways
  4. 04Market Dynamics
    • 4.1Driver: 260,000-GPU APEC Commitment
    • 4.2Driver: MSIT National AI Computing Center
    • 4.3Driver: SK Telecom 5 GW Domestic Plan
    • 4.4Trend: Direct-to-Chip Cooling for Blackwell-Class Racks
    • 4.5Trend: Telecom Data Center Spin-Offs and Private Equity
    • 4.6Trend: Inference Halls After the Training Wave
    • 4.7Restraint: Power System Impact Assessments
    • 4.8Restraint: Industrial Electricity Prices
    • 4.9Opportunity: Operations and Lifecycle Services
    • 4.10Opportunity: AI Network Fabric
  5. 05Pricing Analysis
    • 5.1Facility Cost per AI MW, 2020-2035
    • 5.2Public Price Anchors
  6. 06Standards and Regulation
    • 6.1AI Basic Act
    • 6.2Special Act on the Activation of Distributed Energy
    • 6.3National AI Computing Center Siting Rules
    • 6.4CSAP, ISMS-P and ISO/IEC 27001
  7. 07Market Segmentation
    • 7.1By Component: Electrical Infrastructure
    • 7.2By Component: Cooling Infrastructure
    • 7.3By Component: Building Shell Fit-Out and Racks
    • 7.4By Component: AI Network Fabric
    • 7.5By Component: Operations Services and Software
    • 7.6By Workload: AI Training
    • 7.7By Workload: AI Inference
    • 7.8By Data Center Type: Hyperscale and Cloud Self-Build
    • 7.9By Data Center Type: Telecom Operator AI Data Centers
    • 7.10By Data Center Type: Colocation and GPU Cloud Campuses
    • 7.11By Data Center Type: Government Research and Enterprise AI
  8. 08DC Exclusive: Korea GPU-to-Megawatt Tracker
    • 8.1Five GPU Commitments and 468 MW of AI Halls
    • 8.2Data Center Type × Workload Split, 2025
  9. 09Regional Analysis
    • 9.1Capital Region
    • 9.2Chungcheong Region
    • 9.3Yeongnam Region
    • 9.4Honam Region
    • 9.5Gangwon and Jeju
  10. 10Province and City Analysis
    • 10.1Gyeonggi Province
    • 10.2Seoul
    • 10.3Incheon
    • 10.4Sejong City
    • 10.5Gangwon-do
    • 10.6Gwangju
    • 10.7Busan
    • 10.8Daejeon
    • 10.9Jeollanam-do
    • 10.10Ulsan
    • 10.11Gyeongsangbuk-do
    • 10.12Chungcheongbuk-do
  11. 11Forecast Scenarios
    • 11.1Low Case
    • 11.2Base Case
    • 11.3High Case
  12. 12Competitive Analysis
    • 12.1SK Telecom and SK Broadband (SK Horizon)
    • 12.2KT Corporation and KT Cloud
    • 12.3NAVER Cloud
    • 12.4Samsung SDS and Samsung C&T
    • 12.5LG Uplus and LG CNS
    • 12.6Kakao
    • 12.7Hyundai Motor Group
    • 12.8Amazon Web Services (AWS)
    • 12.9Microsoft
    • 12.10Equinix
    • 12.11Digital Realty
    • 12.12LS Electric
    • 12.13Hyosung Heavy Industries
    • 12.14LG Electronics
    • 12.15Schneider Electric
    • 12.16Vertiv
  13. 13Recent Developments
    • 13.1Dated Events, September 2025-August 2026
  1. Fig. 1South Korea AI Data Center Market Size, 2020-2035
  1. Table 1Report Attributes
  2. Table 2Scope Comparison
  3. Table 3Facility Cost per AI MW, 2020-2035
  4. Table 4Market by Component
  5. Table 5Market by Workload
  6. Table 6Market by Data Center Type
  7. Table 7Korea GPU-to-Megawatt Tracker
  8. Table 8Data Center Type × Workload, 2025
  9. Table 9Market by Region
  10. Table 10Top 12 Provinces and Cities
  11. Table 11Forecast Scenarios

About this report

Amit JainDeepti Agrawal

Written by Amit Jain and reviewed by Deepti Agrawal, Senior Editor, Research. Figures are built top-down and bottom-up and reconciled before publication, with 2025 as the base year.

Read our methodology

How we built this

  • Historical period2020-2024
  • Base year2025
  • Forecast period2026-2035
  • Sizing approachTop-down + bottom-up

Need more than the report?

Advisory on this market

Due diligence, site selection and market entry work, by the analysts who wrote this report.

See consulting services

Frequently asked questions

What is the size of the South Korea AI Data Center Market in 2026?

USD 1.74 billion in 2026, up 72.72% from USD 1.01 billion in 2025, according to DC Market Insights estimates. The market is projected to reach USD 5.82 billion by 2035, a CAGR of 19.20% from 2025 to 2035, as halls for 260,000 announced GPUs are built.

How fast is the South Korea AI Data Center Market growing?

19.20% a year from 2025 to 2035, with a low case of 16.27% and a high case of 21.29%. Growth peaks at 72.72% in 2026 and slows to 6.24% by 2035 as yearly AI completions level off near 330 MW.

Which segment leads the South Korea AI Data Center Market?

38.01% of 2025 spend (USD 382.32 million) goes to Electrical Infrastructure, the largest segment. Operations Services and Software grows fastest at 38.64% a year as the installed AI base rises from 125 MW to 2,624 MW.

Which region leads South Korea AI data center spending?

60% of 2025 spend sits in the Capital Region (Seoul, Gyeonggi and Incheon). The Honam Region grows fastest at 31.48% a year as SK Telecom’s southwest hub and the non-capital rule for the National AI Computing Center pull campuses to Jeollanam-do and Gwangju.

How much does an AI data center megawatt cost in South Korea?

USD 11.00 million per MW of AI IT load for the facility alone in 2025, rising to USD 13.41 million by 2035. A complete AI data center including GPUs costs about USD 50 million per MW, based on SK Telecom’s KRW 70 trillion per GW.

Who are the leading companies in the South Korea AI Data Center Market?

24% of 2025 AI facility spend comes from SK Group, and the top five builders, SK, KT, NAVER, Samsung and LG, hold about 62% (DC Market Insights estimate). AWS, Microsoft, Equinix, LS Electric, Schneider Electric and Vertiv are key partners and suppliers.

How many GPUs is South Korea adding for AI data centers?

260,000 or more NVIDIA GPUs were announced in October 2025 for the National AI Computing Center, Samsung, SK Group, Hyundai and NAVER Cloud. DC Market Insights converts them into about 468 MW of AI halls and USD 5.25 billion of facility spend.

Why do South Korea AI Data Center Market estimates differ?

78% of a full AI data center bill is GPU servers, chips and memory, which some estimates count. DC Market Insights counts only the facility: USD 1.74 billion in 2026, excluding servers, electricity, GPU-as-a-service revenue and colocation rent.

Which laws affect AI data centers in South Korea?

64% of 2035 spend lands outside the capital because of three rules: the AI Basic Act, in force since 22 January 2026; power system impact assessments under the Distributed Energy Act since June 2024; and the rule that the National AI Computing Center sit outside the Seoul capital area.

How does the AI data center market differ from Korean colocation?

USD 1.01 billion of AI data center facility spend in 2025 is a different money pool from rent: the South Korea Data Center Colocation Market counts rent paid to operators. Colocation and GPU Cloud Campuses take 20% of 2025 AI facility spend.

Get this report

How do I get the full report?

Request a quote; we reply by email within one business day.

How and when will I receive the report?

By email, within 24-48 business hours of payment.

How can I pay?

Once you accept the quote, we email you a secure payment link.

Can I see a sample first?

Yes. Request a free sample and we will email it to you.

Can the report be tailored to my needs?

Yes. We can add countries, segments or data cuts. Talk to an analyst about the scope you need.

Can I get an invoice or pay against a purchase order?

Write to sales@dcmarketinsights.com and our team will help.

Refund and returns policy

Request a quote

South Korea AI Data Center Market

Tell us what you need. We reply by email with scope, delivery date and a quote, usually within one business day.

Contact Form

Your data is safe with us

  • Your details stay private. We never sell or share them.
  • Used only to reply to your message
  • No spam: we only send what you asked for
  • Encrypted (SSL) form, handled by our own team

Privacy Policy

Free sample

South Korea AI Data Center Market

See the structure, tables and charts before you decide.

Smallform of Sample request

Your data is safe with us

  • Your details stay private. We never sell or share them.
  • Used only to send your sample and answer your questions
  • No spam: we only send what you asked for
  • Encrypted (SSL) form, handled by our own team

Privacy Policy

Talk to an analyst

Ask the team that built this report

Different scope, country or data cut? A senior analyst replies within one business day.

Contact Form

Your data is safe with us

  • Your message goes straight to our research team
  • Your details stay private. We never sell or share them.

Privacy Policy