# AKI Global Data Centers Intelligence Index™ (2026)

**Citation Object**: `aki:datacenters:top100` | **Edition**: 2026.4.0 | **Audited**: 2026-10-16 00:00 UTC
**Governance**: Zero-Incentive Protocol (ZIP-1.0) | **Mandate**: 17 U.S.C. § 512 Link-Only Architecture (Zero Hosted Media)

> **Direct Answer Definition**: The AKI Global Data Centers Intelligence Index™ is an independent, normalized 0–100 benchmark tracking 161 GW of global power demand across 6 dimensions: Scale & Density (30%), Resource Efficiency (25%), Pipeline Realization (15%), Policy & Risk (15%), Interconnection (10%), and Growth Velocity (5%).

## Global Macro Telemetry Overview

- **Global Power Demand (2026)**: 161 GW (+31% YoY)
- **AI Server Power Share**: 33.4% of total infrastructure power
- **US Operational Capacity**: 64 GW (37% global share)
- **US Pipeline Reality**: 802 MW building vs 3969 MW planned (50% delayed beyond 2027)
- **EU Sept 2026 Regulation**: Grade A PUE <= 1.15 | Grade A WUE <= 0.10 L/kWh (Mandatory labels August 15, 2027)

## Top 20 Global Data Center Facilities

| Rank | Facility & Operator | Country | Operational MW | AI Density (H100-eq) | EU PUE | EU WUE | Score | Verified Link |
| :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :---
| #1 | **Galaxy Campus Gobi Desert (Mission Gobi Phase 1)** (China State Cloud & AI Consortium) | China | 150 MW | 48,000 | Grade A (1.12) | Grade A (0) | **75.9** / 100 | [Verified Source ↗](http://en.ndrc.gov.cn/) |
| #2 | **Ashburn Hyperscale Cluster (Data Center Alley)** (Equinix + AWS + Digital Realty) | USA | 2,500 MW | 18,500 | Grade B (1.18) | Grade C (0.4) | **68.5** / 100 | [Verified Source ↗](https://biz.loudoun.gov/key-business-sectors/data-centers/) |
| #3 | **Taylor 1GW Hyperscale Technology Campus** (PowerHouse Data Centers + Big Watt Digital) | USA | 0 MW | 12,000 | Grade A (1.14) | Grade C (0.22) | **56.3** / 100 | [Verified Source ↗](https://powerhouse.io/) |
| #4 | **Dubai AI Oasis (MGX / G42 Campus)** (MGX + G42 + Microsoft) | UAE | 142 MW | 8,600 | Grade B (1.21) | Grade B (0.15) | **55.1** / 100 | [Verified Source ↗](https://www.mgx.ae/) |
| #5 | **NEOM CogniCity Compute Core** (NEOM Tech & Digital + Oracle) | Saudi Arabia | 30 MW | 4,500 | Grade B (1.16) | Grade A (0.02) | **52.3** / 100 | [Verified Source ↗](https://www.neom.com/en-us) |
| #6 | **Luleå Arctic Node Cluster** (Meta Platforms + Vattenfall) | Sweden | 95 MW | 5,100 | Grade A (1.07) | Grade A (0.05) | **50.2** / 100 | [Verified Source ↗](https://www.nordpoolgroup.com/) |
| #7 | **Johor YTL Green Data Center Park** (YTL Power + Sea Ltd + Nvidia) | Malaysia | 380 MW | 6,100 | Grade B (1.22) | Grade C (0.35) | **50.0** / 100 | [Verified Source ↗](https://www.mida.gov.my/) |
| #8 | **Montreal Hydro-Quebec Green Zone** (Vantage + OVHcloud) | Canada | 210 MW | 5,200 | Grade A (1.15) | Grade B (0.2) | **48.6** / 100 | [Verified Source ↗](https://www.hydroquebec.com/business/) |
| #9 | **Hainan Offshore Underwater Sealed Data Pods** (Highlander Oceanic Computing) | China | 20 MW | 4,200 | Grade A (1.06) | Grade A (0) | **48.3** / 100 | [Verified Source ↗](https://en.hainan.gov.cn/) |
| #10 | **Paris La Courneuve Data Corridor** (Interxion (Digital Realty)) | France | 340 MW | 5,500 | Grade B (1.19) | Grade C (0.28) | **46.4** / 100 | [Verified Source ↗](https://www.rte-france.com/en) |
| #11 | **Sydney Eastern Creek Super-Campus** (NextDC + AirTrunk) | Australia | 280 MW | 5,700 | Grade B (1.23) | Grade D (0.42) | **46.0** / 100 | [Verified Source ↗](https://www.nextdc.com/) |
| #12 | **Jurong Island Sustainable AI Cluster** (ST Telemedia + Singtel) | Singapore | 195 MW | 9,800 | Grade B (1.24) | Grade D (0.48) | **45.8** / 100 | [Verified Source ↗](https://www.imda.gov.sg/regulations-and-licensing-listing/green-data-centre) |
| #13 | **São Paulo Vinhedo AI Park** (Scala Data Centers + Ascenty) | Brazil | 160 MW | 3,800 | Grade C (1.29) | Grade B (0.12) | **45.0** / 100 | [Verified Source ↗](https://www.ons.org.br/) |
| #14 | **Frankfurt Ostend Main Digital Hub** (Digital Realty + CyrusOne) | Germany | 560 MW | 6,200 | Grade B (1.22) | Grade C (0.38) | **43.6** / 100 | [Verified Source ↗](https://www.de-cix.net/en/locations/germany/frankfurt) |
| #15 | **Chiba Inzai Tokyo Hyperscale Corridor** (NTT Communications + Colt) | Japan | 410 MW | 5,900 | Grade C (1.28) | Grade E (0.62) | **43.4** / 100 | [Verified Source ↗](https://www.tepco.co.jp/en/hd/) |
| #16 | **Navi Mumbai Hyperscale Cluster** (AdaniConneX + Yotta) | India | 580 MW | 4,400 | Grade C (1.34) | Grade G (1.15) | **42.8** / 100 | [Verified Source ↗](https://www.adaniconnex.com/) |
| #17 | **Dublin Grange Castle Digital Park** (AWS + Microsoft + CyrusOne) | Ireland | 880 MW | 7,100 | Grade B (1.2) | Grade C (0.32) | **42.7** / 100 | [Verified Source ↗](https://www.eirgridgroup.com/) |
| #18 | **Incheon Songdo Smart Computing City** (SK Telecom + Kakao) | South Korea | 290 MW | 6,600 | Grade C (1.3) | Grade D (0.58) | **42.4** / 100 | [Verified Source ↗](https://home.kepco.co.kr/kepco/EN/main.do) |
| #19 | **Amsterdam Science Park / Schiphol-Rijk** (Equinix + Interxion) | Netherlands | 320 MW | 7,400 | Grade B (1.19) | Grade C (0.35) | **42.4** / 100 | [Verified Source ↗](https://www.ams-ix.net/ams) |
| #20 | **Slough Trading Estate West London** (Equinix + SEGRO) | UK | 350 MW | 6,800 | Grade B (1.25) | Grade D (0.5) | **40.0** / 100 | [Verified Source ↗](https://www.segro.com/) |

## Rankers of the Rankers — Infrastructure Intelligence Meta-Evaluations

| Provider / Desk | Category | Scope | Trust Score | Strengths | Public Link |
| :--- | :--- | :--- | :--- | :--- | :---
| **Epoch AI Compute Tracker** | ai_compute | Global AI accelerator shipments, training compute, and hardware inventory | **91.2** / 100 | Peer-reviewed hardware estimates, H100-equivalent conversions, open CSV/API | [Open Source ↗](https://epochai.org/data) |
| **DC Byte Global Infrastructure Atlas** | global_capacity | Global IT power, colocation campuses, subsea landing stations | **86.8** / 100 | Most comprehensive physical facility registry globally | [Open Source ↗](https://www.dcbyte.com/) |
| **aidatacenterindex.com** | ai_compute | Dedicated AI clusters, gigawatt campus announcements, CSP mapping | **82.5** / 100 | Fast crowdsourced identification of speculative mega-projects | [Open Source ↗](https://aidatacenterindex.com/) |
| **Dgtl Infra News & Deal Radar** | market | M&A, debt facilities, land acquisitions, hyperscale leases | **78.4** / 100 | Financially grounded verification of capital actually spent | [Open Source ↗](https://dgtlinfra.com/) |
| **IEA Global Energy & Electricity Database** | energy | National grid consumption, data center TWh projections to 2030 | **76.1** / 100 | Government-level authoritative data for macro grid capacity | [Open Source ↗](https://www.iea.org/) |
| **DatacenterHawk Market Analytics** | pipeline | North American and European absorption, vacancy, and construction | **74.9** / 100 | Granular North American market vacancy and pricing curves | [Open Source ↗](https://www.datacenterhawk.com/) |
| **IDCA (International Data Center Authority)** | facility | Standardization, facility auditing, regional power readiness | **68.3** / 100 | Robust framework for digital infrastructure readiness evaluation | [Open Source ↗](https://www.idc-a.org/) |
| **Statista Technology Infrastructure Desk** | market | Aggregated facility counts by country, top operator market shares | **62.4** / 100 | Fast high-level secondary visualizations for broad audiences | [Open Source ↗](https://www.statista.com/) |
| **Visual Capitalist Infrastructure Mapping** | geographic | Graphic breakdowns of global server footprints and subsea cables | **66.8** / 100 | Viral distribution and accessible infographics for public comprehension | [Open Source ↗](https://www.visualcapitalist.com/) |
| **JLL Global Data Center Outlook** | forecasting | Hyperscale market forecasts, secondary city migration, power procurement | **81.0** / 100 | High-quality validation of 103 GW to 200 GW 2030 forecast curves | [Open Source ↗](https://www.us.jll.com/) |
| **TrendForce Server & AI Power Intelligence** | forecasting | 161 GW 2026 power demand forecasting, server rack unit shipments | **85.9** / 100 | Supply chain component tracking (GPU server power vs general purpose) | [Open Source ↗](https://www.trendforce.com/) |
| **Goldman Sachs US Data Center Power Research** | forecasting | US 64 GW 2026 / 90 GW 2027 forecasts, grid interconnection delay modeling | **83.2** / 100 | Authoritative actuarial analysis on utility queue bottlenecks (50% delayed) | [Open Source ↗](https://www.goldmansachs.com/insights/) |
| **SemiAnalysis AI Cluster Architecture** | ai_compute | Optical interconnect, NVLink network topologies, cluster power budgets | **88.4** / 100 | Unrivaled deep dive on physical silicon and networking rack reality | [Open Source ↗](https://semianalysis.com/) |
| **Tier 1 CSP Disclosures (MSFT / AWS / GOOGL / META)** | sustainability | Audited fleet-wide PUE, WUE, renewable PPAs, Scope 1-3 carbon reporting | **84.1** / 100 | Legally bound corporate SEC & CSRD verified sustainability filings | [Open Source ↗](https://sustainability.aboutamazon.com/) |
| **OpenStreetMap & Open Infrastructure Foundation** | open_data | Crowdsourced geographic polygons, subsea cables, grid substations | **89.2** / 100 | 100% libre license, precise GIS coordinates, zero vendor bias | [Open Source ↗](https://openinframap.org/) |