COUNTRIES / APAC + MIDDLE EAST · 20 MARKETS IN SCOPE
Hover or focus a map marker for a snapshot. Select a country to open its findings, charts and source records.
Explore documented data-centre activity across 20 markets. Available facility details include dated sources; an operating label retains the date of its source.
474campus / site groups in research
568research records, including supporting rows
474 research campus / site groups · Inventory snapshot . Source dates vary; coverage is incomplete.
Country aggregates · no facility pins
On smaller screens, scroll the map sideways to view every country.
474 research campus / site groups.
568 research records across all lifecycle states in the inventory, including building / phase observations and unresolved identities. These records are not additional campuses.
Research totals describe NTRY’s documented inventory. They are not a complete market census, fresh operating verification or accepted evidence. Source and identity checks determine which named details can be released.
SSD-73761707 · Status evidenced
Planning approval only. Construction, commissioning, operation and available power are not established. Department registry metadata; third-party plans and images excluded.
View primary source ↗NSW Crown Copyright – Department of Planning and Environment · CC BY 3.0 Australia · Source reuse policy
Campus / site groups are counted once; supporting building and phase rows do not add campus counts. Named facility details are a subset of this inventory and must not be added to its counts. Planning approval is not construction or operation. Accepted measurements remain separate from facility counts.
Source-linked measurements and news provide context for the facility inventory. Evidence records do not add sites to its counts.
Select a country on the map to inspect its dated measurements and research context.
These accepted measurements can describe a campus, phase or wider programme. They are not additional facilities and are not added to the inventory. A site relationship is shown only after its identity is matched.
Planned project power capacity ceiling · Evidence 01 Oct 2026
JERA announced the Chiba project on 1 October 2026 with up to 400 MW of planned power capacity and operations targeted around 2028. JERA provides generation, Dell the compute platform and RHAELM project delivery. This is a site-level power ceiling, not IT load, operating capacity, binding secured-power evidence or a Japan market total. The framework is an MoU; financing closure, construction start and commissioning remain unverified. Behind-the-meter delivery may shorten grid dependence, but the claimed acceleration is not an observed outcome.JERA Chiba AI infrastructure MoU release 2026 source ↗Development pipeline capacity · Evidence 31 Aug 2026
NEOM reports construction underway for the first 100 MW of a 360 MW Oxagon phase, with availability anticipated in 2028. This is under-construction pipeline, not operating capacity. The official release does not establish an IT-load basis, so the basis remains publisher-reported and unspecified.NEOM HUMAIN DataVolt Oxagon release source ↗Development pipeline capacity · Evidence 21 Aug 2026
STTGDC was selected under DC-CFA2 to develop 50 MW of new Singapore data-centre capacity. The award remains subject to approval and implementation processes and is not operating capacity; the publisher does not label the figure as operating IT load.STTGDC Singapore DC-CFA2 release source ↗Development pipeline capacity ceiling · Evidence 13 Aug 2026
STT Johor is planned for up to 166 MW of IT load. The normalized 166 MW value is an explicit publisher-reported upper bound for one named campus, not exact pipeline capacity, operating capacity or a Malaysia market total.STTGDC Johor financing release source ↗Estimated development pipeline capacity · Evidence 28 Apr 2026
The official groundbreaking release describes the Visakhapatnam AI hub as nearly 1 GW in one location. The normalized 1,000 MW value is a publisher approximation for one groundbroken hub, not exact pipeline capacity, operating capacity or an India market total.Adani Google Visakhapatnam AI Hub groundbreaking source ↗Development pipeline capacity ceiling · Evidence 18 Dec 2025
Saudi Press Agency reported that the HUMAIN-center3 partnership would develop and operate infrastructure capable of supporting up to 1 GW of AI workloads. The normalized 1,000 MW value is an explicit partnership-programme ceiling, not exact site capacity, operating capacity or a Saudi market total.Saudi Press Agency HUMAIN center3 JV release source ↗Development pipeline capacity ceiling · Evidence 17 Dec 2025
AirTrunk announced that OSK2 in East Osaka would deliver up to 100 MW of total campus IT capacity. The value is an upper bound for one named announced campus, not operating capacity or a Japan market total.AirTrunk OSK2 announcement 2025 source ↗Development pipeline capacity · Evidence 05 Nov 2025
Microsoft and G42 announced 200 MW of additional UAE data-centre capacity delivered through Khazna, with initial capacity expected online before the end of 2026. It is announced pipeline, not operating capacity or a UAE market total.Microsoft G42 UAE capacity expansion release source ↗Development pipeline capacity · Evidence 16 Feb 2025
Saudi National Infrastructure Fund reported financing for a first phase of four Alfanar Projects data centres in Riyadh and Dammam totaling 19.2 MW, or 4.8 MW per centre. This is a financed planned phase, not operating capacity, delivered power, a commissioning milestone or a Saudi market total.Saudi National Infrastructure Fund Alfanar financing release source ↗One physical facility map; lifecycle and evidence review remain separate. Power allocations, MW, capital plans and programme measurements never become extra facility counts. The measurement feed is partial; some evidence records may be absent.
80 cited research updates · 54 contextual charts · latest evidence date 2026-11-10
Contextual research informs diligence; it does not change accepted observations, governed coverage or rankings. Governed views use the returned evidence feed; missing or bounded records are not proof of no market activity.
Google says it plans to add more than 1 GW of new clean-energy capacity to Malaysia’s National Grid through generation development under CRESS, working with TNB. The plan accompanies a US$15 billion investment over 2027–2028, with the operational Elmina campus and new campuses planned in Selangor and Negeri Sembilan. The Star reports that two Kedah solar projects totalling about 50 MW are part of the existing clean-power context.
CAVEAT Announced future grid capacity, not verified operating generation, firm campus supply or a time-matched clean-power share. Project-level capacity, commissioning dates, Google campus load and hourly matching were not disclosed. The roughly 50 MW Kedah figure is separately reported by The Star and is not the total 1 GW plan. No campus-level offset or grid headroom is inferred.
The Star / Reuters · 10 Nov 2026 · Context only · excluded from capacity totals CLOUD FOOTPRINT / REGIONS NOT FACILITIESSoutheast AsiaCONTEXT ONLYAlibaba Cloud lists six Southeast Asian cloud regions in five countries; Johor AZ count conflictsThe official locations pages checked on 10 November 2026 UTC list Singapore, Kuala Lumpur, Johor, Jakarta, Bangkok and Manila: six cloud regions across five countries. Both pages list Singapore 4, Kuala Lumpur 3, Jakarta 3, Bangkok 2 and Manila 2 availability zones. One page lists Johor 3 while the other lists 2; that conflict is unresolved.
CAVEAT This is a dated website snapshot, not a publisher release date or a physical-facility inventory. A cloud region is a service geography; an availability zone may contain one or multiple data centres. The chart counts listed cloud regions only. Johor is excluded from any asserted exact AZ total because official pages disagree between 2 and 3. Do not turn regions or AZs into physical sites, ownership claims, MW, utilisation, operating-capacity totals or facility pins.
Alibaba Cloud · Website checked 10 Nov 2026 · Context only · excluded from capacity totals SITE STATUS / PERMITTINGSouth KoreaCONTEXT ONLYSaemangeum data-centre preparation advances; construction remains targetedSouth Korea’s Saemangeum agency said on 30 October that Hyundai opened a local project office. Architectural review for the proposed AI data centre was completed on 16 October; building-permit and industrial-estate procedures were still being progressed toward a targeted March 2027 construction start.
CAVEAT Architectural review is not a building permit or observed construction. The office is a project coordination base, not an operating data centre. No commissioning, delivered capacity or facility-count increase is established.
Saemangeum Development and Investment Agency / Korea Policy Briefing · 30 Oct 2026 · Context only · excluded from capacity totals0 recent additions since opening. Showing the latest three items. Source facts remain separate from NTRY analysis; every item includes a caveat. Hover or focus to pause. Checks every two hours while this page is open; new items appear only after review. Announcement dates are not refresh times.
Each line shows the path from cited context to a genuinely comparable governed series. It is a coverage graphic, not an attractiveness score or ranking.
Mint segments are visible in the current dated response; outlined segments are not visible; hatched segments cannot be verified because a required governed feed is unavailable. Missing coverage is never plotted as zero and does not mean no market activity. Malaysia’s evidence count includes Peninsular Malaysia; each measurement retains its original scope.
Choose a core market to open its supply–demand charts, accepted evidence and missing proof. Operating supply, contracted power, observed demand and forecasts retain their separate meanings.
Other markets remain available through the country selector; their governed supply–demand interpretation is not yet mapped.
Compare population scale and internet participation, then inspect business adoption within each market. These are demand proxies; they do not measure local compute workloads or available data-centre capacity.
Horizontal axis: population, logarithmic scale. Vertical axis: individuals using the internet, % of population. Select a point or choose a market below.
Panel medians are 31.2m people and 95.5% internet participation. They are descriptive reference lines—not targets, scores, demand forecasts or investment thresholds.
World Bank WDI / ITU · fixed 2024 comparison · 18 matched economies. Dashed guides are panel medians. Unverified or conflicting cells are omitted, never plotted as zero.
Market size and population growth answer different questions. Switch between percentage change and people added; select a bar to inspect its source. Market order is alphabetical.
Shared zero line · left: decrease · right: increase · percent change from the 2023 base. Bar lengths use one common scale. Mint: increase · slate: decrease · amber: selected market.
Source: World Bank WDI population estimates, CC BY 4.0; exact source links appear in each profile. NTRY calculation: (2024 − 2023) / 2023 × 100; absolute change is 2024 − 2023. Taiwan’s year-end registered-population series remains separate in its profile. Population change does not establish customer spend, local workloads or causation.
Compare reported digital-participation change on its own, or place it beside population movement. Select any point or bar to inspect the exact dated history and source.
Horizontal: 2023→2024 population change, %. Vertical: 2019→2024 internet-participation change, percentage points. Same source series within each axis; periods intentionally remain distinct. Amber: selected market.
Population uses 2023→2024 midyear estimates; digital participation uses the 2019→2024 internet-use series. Different periods and units remain separate.
Sources: World Bank WDI population estimates and WDI / ITU indicator “Individuals using the Internet (% of population)”, CC BY 4.0; exact source links appear in each profile. NTRY calculations: 2024 population − 2023 population as a percent of the 2023 base; 2024 internet-use value − 2019 value. The relationship view has 18 matched markets. China and Taiwan remain outside it because their retained internet observations use unresolved or different denominators. Survey methods and historical revisions can create apparent movement; neither view measures enterprise adoption, compute demand or causation.
35,126,298 → 35,557,673 persons. World Bank country/economy midyear estimates on the displayed source basis.
Population estimates may be modelled and revised; not a census count or measure of local compute demand. Internet-use survey quality, frames and reference periods can differ among reporting economies. Annual population and internet percentages must not be interpreted as enterprise AI adoption, spending, or campus-ready power.
84.2% → 98% of population on the displayed World Bank / ITU indicator basis. Historical revisions and survey-method changes remain possible.
| Metric | Reference period | Percent |
|---|---|---|
| Individuals using the internet | 2019 | 84.2% |
| Individuals using the internet | 2020 | 89.6% |
| Individuals using the internet | 2021 | 96.8% |
| Individuals using the internet | 2022 | 97.4% |
| Individuals using the internet | 2023 | 97.7% |
| Individuals using the internet | 2024 | 98% |
Source-reported contextual history. Historical revisions and national survey differences remain possible.
June 2026 · 18/20 markets. Compare the work-related share of consumer ChatGPT messages, then select a market to inspect its monthly history. The percentage describes messages within each market; it is not a business-adoption rate or provider market share.
Common 0–100% scale · alphabetical order · mint: published share · amber: selected market · unknown: no released observation. Select a row to update all demand evidence.
24/24 monthly observations in this release. Automated publisher classification; messages, not users or businesses.
| Month (UTC) | Work-related share | Published fraction |
|---|---|---|
| Jul 2024 | 56.9% | 0.569 |
| Aug 2024 | 49.5% | 0.495 |
| Sep 2024 | 45.1% | 0.451 |
| Oct 2024 | 45.7% | 0.457 |
| Nov 2024 | 47.2% | 0.472 |
| Dec 2024 | 45.9% | 0.459 |
| Jan 2025 | 45.3% | 0.453 |
| Feb 2025 | 39.5% | 0.395 |
| Mar 2025 | 35.3% | 0.353 |
| Apr 2025 | 33.8% | 0.338 |
| May 2025 | 37.5% | 0.375 |
| Jun 2025 | 36.7% | 0.367 |
| Jul 2025 | 33.0% | 0.330 |
| Aug 2025 | 30.9% | 0.309 |
| Sep 2025 | 31.7% | 0.317 |
| Oct 2025 | 32.4% | 0.324 |
| Nov 2025 | 35.6% | 0.356 |
| Dec 2025 | 35.0% | 0.350 |
| Jan 2026 | 34.5% | 0.345 |
| Feb 2026 | 30.1% | 0.301 |
| Mar 2026 | 30.5% | 0.305 |
| Apr 2026 | 37.3% | 0.373 |
| May 2026 | 33.2% | 0.332 |
| Jun 2026 | 35.0% | 0.350 |
Lower rank number means higher consumer ChatGPT message use relative to 2023 population within the publisher's country panel.
| Quarter | Published rank | Population basis |
|---|---|---|
| Jan 2025 | #51 | 2023 World Bank population |
| Apr 2025 | #46 | 2023 World Bank population |
| Jul 2025 | #39 | 2023 World Bank population |
| Oct 2025 | #54 | 2023 World Bank population |
| Jan 2026 | #60 | 2023 World Bank population |
| Apr 2026 | #59 | 2023 World Bank population |
Work-related share shows the composition of messages within a market; the ordinal rank shows population-adjusted consumer ChatGPT message use across countries. Together they can prioritise deeper demand diligence, but neither measures adoption, users, absolute message volume, paid enterprise demand, provider market share, local workload hosting or compute demand.
OpenAI population-adjusted consumer message-use rank paired with Microsoft's estimated share of people ages 15–64 using any generative-AI product.
Dashed lines are the descriptive medians of the 17 matched markets: rank #59 and 27.9% estimated AI use. They are not targets or thresholds.
The relationship can flag markets for deeper diligence when population-adjusted ChatGPT message use and the broader Microsoft generative-AI-use estimate diverge or align. Different products, populations, methods and denominators make this contextual triangulation—not provider market share, adoption proof, a composite score, causal evidence, compute demand or an allocation rule.
Claude and Gemini: comparable country-by-month measures are not included in this view. No cross-provider share or composite adoption score is calculated.
Consumer ChatGPT messages from Free, Go, Plus and Pro accounts with declared age at least 18; excludes Enterprise and Codex, deleted accounts and users who opt out of training.
300,000 messages sampled globally per month; day-weighted, automatically classified work-relatedness; differential privacy and rounding to 3 decimals. Country is associated with request. Rank uses 2023 World Bank population. Values are publisher estimates; granular country slices have greater sampling variance. Changes may reflect user composition, classification and restatements. This is external contextual history, with no semantic acceptance or NTRY inference.
Chatterji, Cunningham, Deming, Hitzig, Johnston, Richmond, Ong, Shan and Wadman, OpenAI Signals v2.0 · CC BY 4.0. NTRY modifications: selected country rows, conversion of fractions to percentages, monthly chart, quarterly ordinal-rank chart, cross-publisher descriptive scatter and missing-value display. Snapshot checked 12 October 2026; data ends June 2026. The public downloads are verified source material; this contextual panel does not create accepted observations or NTRY inference.
Q2 2026 · 18/20 NTRY markets · 72 dated estimates. Microsoft estimates the share of people ages 15–64 who used a generative AI product. This is a broad usage measure—not ChatGPT, Claude or Gemini market share.
Ranked Q2 2026 estimate · common 0–100% scale · mint: published estimate · amber: selected market · unknown: no economy estimate in this release.
Horizontal: 2024 population, logarithmic scale. Vertical: Microsoft Q2 2026 AI Diffusion estimate. Dashed guides are the 18-market panel medians—not targets, scores or allocation thresholds. Select a point to inspect its evidence.
1.0× the panel median population · -5.1 percentage points versus the panel median AI-use estimate.
The axes intentionally retain different reference periods and definitions. Their position is descriptive context only: it does not estimate addressable users, enterprise demand, compute load, revenue, hosting location, power need or causation.
Horizontal: Microsoft Q2 2026 AI Diffusion estimate. Vertical: NTRY research-inventory count of campus groups tracked before operation across announced, proposed, approved or permitted, financed, groundbroken and under-construction stages. Amber: selected market. Dashed guides: panel medians.
18 additional campus groups are classified operating in the research inventory. Counts describe the bounded research footprint, not market capacity, MW, utilisation, completeness or causation.
Markets above the horizontal median have a larger tracked development footprint; markets right of the vertical median have higher estimated consumer generative-AI use. Investigate apparent mismatches with enterprise demand, workload location, occupied IT load, power access and delivery evidence before drawing an allocation conclusion.
NTRY calculation: sum of six explicitly pre-operational lifecycle groups in the bounded physical-facility research inventory dated 2026-11-06. Campus counts are not capacity and research coverage is not a census. Microsoft usage and NTRY facility research remain separate evidence families.
4 published estimates · H1 2025 18.3% → Q2 2026 22.6% · +4.3 percentage points.
| Period | Estimated adoption |
|---|---|
| H1 2025 | 18.3% |
| H2 2025 | 19.7% |
| Q1 2026 | 21.8% |
| Q2 2026 | 22.6% |
Use this as a comparable adoption proxy beside population, digital reach and official enterprise surveys. It can prioritize questions about paid business demand, local workload placement and connectivity. It cannot size compute, power, revenue, capacity headroom or hyperscaler allocation.
The population×AI scatter and ranked bars are used instead of a radar chart. They preserve actual magnitudes, ordering and unknown values without normalizing unrelated evidence into a composite score.
Estimated share of people ages 15–64 who used a generative AI product during the period. Microsoft aggregates and anonymizes consumer AI telemetry, then adjusts estimates for operating-system and device-market share, internet penetration and population.
Microsoft describes this as an estimate from aggregated, anonymized consumer telemetry. Results can move with product coverage, calibration, internet access and historical revisions. The measure is usage—not skill, productivity, business adoption, spending, provider share, compute demand, hosting location or infrastructure capacity.
NTRY modifications: exact selection of the 20-market universe, labels aligned to NTRY market names, percentage parsing, ranking and selected-market trend. Snapshot checked 2026-11-05. The public dataset is verified source material; this contextual panel does not create accepted observations or NTRY inference.
Population is country/economy context, not a utility service territory. World Bank WDI population values are midyear estimates; Taiwan uses separately sourced year-end registered population and is excluded from the shared scatter. Internet use is the share of individuals who used the internet. World Bank / ITU attribution, CC BY 4.0. NTRY modifications: 2024 selection, rounding, chart rendering and within-series percentage-point changes and 2023–2024 population differences. These sources provide context only; this panel does not create accepted observations or an investment conclusion.
Study sample · historical climate baseline · country-level context
Historical 2001–2010 water conditions applied to the study’s operational facility inventory; not current water availability.
Share of each market’s study sample exposed for at least one month annually. Separate samples and facility counts; this is not an investment ranking.
75 / 229 study facilities
0 / 5 study facilities
146 / 344 study facilities
0 / 92 study facilities
173 / 238 study facilities
35 / 137 study facilities
95 / 149 study facilities
4 / 4 study facilities
0 / 84 study facilities
1 / 54 study facilities
12 / 15 study facilities
3 / 23 study facilities
10 / 10 study facilities
10 / 34 study facilities
0 / 74 study facilities
6 / 77 study facilities
0 / 21 study facilities
31 / 38 study facilities
50 / 54 study facilities
0 / 25 study facilities
Rosa, Sangiorgio, Scanlon, Mazzotti, Berry and Michalak (2026). Published dataset: 23 September 2026; accessed 1 November 2026. NTRY selected twenty markets and calculated exposed counts and shares. No endorsement implied.
Exposure combines the 1–4, 5–8 and 9–12 month bins. +3°C results average five climate models using H08, retaining fractional counts. Historical climate conditions and future scenarios are separate from the facility inventory date, which remains unverified.
China, Hong Kong and Taiwan retain separate source rows; detailed geographic boundary methods remain unverified. Aggregate exposure does not identify individual sites, permits, consumption, capacity or water entitlement. These figures are separate from NTRY’s lifecycle inventory and governed evidence totals. No proprietary coordinates are provided or inferred.
Modelled zero exposure does not establish water security, including local supply, desalination, cross-border supply or operational drought risk. Water-use sector overlap describes spatial context, not a data centre’s cooling system. Aqueduct scores have different definitions and must be interpreted separately.
Source-reported developments, with project stage and caveats. These updates do not change accepted measurements, capacity totals or rankings.
Visible contextual charts only—not governed market×dimension coverage, a completeness score or a ranking.
OFFICIAL WEBSITE SNAPSHOT · SIX CLOUD REGIONS · FIVE COUNTRIES · JOHOR AZ 2 / 3 UNRESOLVED · NOT FACILITY COUNTS
The official locations pages checked on 10 November 2026 UTC list Singapore, Kuala Lumpur, Johor, Jakarta, Bangkok and Manila: six cloud regions across five countries. Both pages list Singapore 4, Kuala Lumpur 3, Jakarta 3, Bangkok 2 and Manila 2 availability zones. One page lists Johor 3 while the other lists 2; that conflict is unresolved.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is a dated website snapshot, not a publisher release date or a physical-facility inventory. A cloud region is a service geography; an availability zone may contain one or multiple data centres. The chart counts listed cloud regions only. Johor is excluded from any asserted exact AZ total because official pages disagree between 2 and 3. Do not turn regions or AZs into physical sites, ownership claims, MW, utilisation, operating-capacity totals or facility pins.
REPORTED BUILT-CAPACITY HISTORY · ANNOUNCED 11 GW SEPARATE · CONTEXT ONLY · NOT GOVERNED CAPACITY
The report's page 60 series shows Southeast Asia built data-centre capacity at 0.8 GW in 2020, 0.9 GW in 2021 and 2022, 1.6 GW in 2023, 2.4 GW in 2024, 4.0 GW in 2025 and 5.8 GW YTD 2026. It separately labels total announced capacity at 11 GW; NTRY does not combine the two statuses.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Built and announced are separate publisher statuses. Built is the report label and is not independently verified by NTRY as operating or commissioned capacity; announced capacity is not delivered capacity. The values are regional, not country totals, a site inventory, contracted power or available headroom. YTD 2026 is the report's latest stock observation, not a full-year flow. Source note: Bain analysis; datacenterHawk. Required citation: Google, Temasek and Bain, e-Conomy SEA 2026. NTRY selected and redrew the values without reproducing the source chart.
JULY 2026 SURVEY SNAPSHOT · SEA-6 WORKING POPULATION · LEADING INDICATOR · NOT INFRASTRUCTURE DEMAND
The July 2026 matrix reports penetration / weekly / daily use at 70% / 62% / 21% in Singapore, 68% / 57% / 20% in Malaysia, 65% / 60% / 26% in Vietnam, 63% / 55% / 22% in Thailand, 54% / 45% / 16% in Indonesia and 53% / 44% / 14% in the Philippines.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Penetration is the share of the working population using a standalone GenAI app or website at least once in the previous 30 days; embedded AI features are excluded. Weekly and daily are frequency shares of the working population in the same matrix. This is a July 2026 survey snapshot, not a repeated series, enterprise token workload, data-centre demand or capacity. It covers SEA-6 only and must not be extrapolated to other NTRY markets. Source note: Google/YouGov, Uncovering AI Adoption, Trust, and Autonomy in SEA, SEA-6, July 2026, n=9,006 AI users; Bain analysis. Required citation: Google, Temasek and Bain, e-Conomy SEA 2026. NTRY redrew the comparison without reproducing the source chart.
GLENDENNING APPROVED 14 SEP 2026 · PROJECT MARS WITHDRAWN · NEITHER VERIFIED OPERATING
The NSW Planning Portal records the Glendenning Road Data Centre as approved on 14 September 2026 with 202.4 MW of proposed power consumption. A separate Greater Sydney proposal, Project Mars, is currently marked withdrawn and described as having approximately 90 MW of overall power consumption.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: These are source-reported planning power scopes for two different named projects, not IT load, firm grid allocation, delivered power or operating capacity. The values must not be added or treated as a repeated series. Approval does not prove construction, energisation, commissioning, occupancy or service commencement. The Project Mars landing page does not state the withdrawal date or reason, while the Glendenning page currently lists no post-approval documents.
1,789 MW OPERATING · 509 MW UNDER CONSTRUCTION · 3,351 MW PLANNED · H1 2026 SNAPSHOT
An Invest India-hosted paper produced with Cushman & Wakefield reports 1,789 MW of operating IT capacity, 509 MW under construction and 3,351 MW planned at H1 2026. The 3,860 MW combined pipeline was about 2.2 times the operating base; planned projects represented 86.8% of that pipeline.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: The capacity table is attributed to Cushman & Wakefield, not an operator-by-operator government register, and the PDF does not reconcile named facilities. Operating, under-construction and planned capacity are separate lifecycle buckets and must not be added to other national estimates without matching methodology and dates. The paper's 5,649 MW end-2030 baseline assumes every identified under-construction and planned project becomes ready for service; that is a scenario, not delivered capacity. Its H1 2026 1,789 MW operating estimate is methodologically distinct from the Government of India's 1.57 GW installed-power figure for August 2026.
8-YEAR FACILITY AGREEMENT · Q1 2028 ENERGISATION TARGET · NO OFFTAKE · NOT OPERATING
Bitdeer AI says it entered an eight-year data-center services agreement for the 60 MW A103 facility at its Cyberjaya campus, with energization expected in Q1 2028. The company says A103 lifts the future Cyberjaya campus scope to about 71.5 MW and its global secured AI Cloud facility portfolio to about 333.5 MW, or approximately 95% of an up-to-350 MW Q1 2028 target.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: All figures are company-reported critical IT load and describe nested or overlapping scopes that must not be added. 'Secured' means owned capacity or capacity under an executed data-center services agreement; it does not mean constructed, energized, commissioned, operating, available or customer-contracted capacity. Bitdeer states that no A103 offtake commitments existed as of 12 October 2026. The 71.5 MW campus scope includes A103 alongside A101 and A102; the 333.5 MW portfolio spans Malaysia, Norway and the United States; and 350 MW is a global ceiling target. The more-than-US$10 billion active pipeline is a management estimate, not backlog, revenue, an executed contract or a binding commitment.
OFFICIAL ANNUAL CONSUMPTION SERIES · NATIONAL ALL-SECTOR DEMAND · COMPLETE 2024 YEAR · NOT DATA-CENTRE-SPECIFIC
Singapore's Energy Market Authority reports total electricity consumption rising from 52 TWh in 2019 to 58 TWh in 2024, an 11.5% increase on the published annual totals. EMA reports 4.0% growth in 2024 and identifies information and communications as one of the faster-growing sub-sectors at 12.3%, without isolating data-centre load.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: These are rounded annual national consumption totals across all sectors. The information-and-communications growth rate includes activities beyond data centres and is not a data-centre electricity-use figure. Consumption is not peak demand, generation capacity, reserve margin, grid headroom, contracted power or campus-available supply, so it cannot size a data-centre supply gap or enter operating-capacity totals.
OFFICIAL ANNUAL RENEWABLE OUTPUT · NATIONAL TOTAL · 2021–2022 · NOT CAMPUS-AVAILABLE POWER
The Ministry reports renewable electricity generation rising from 11,537 GWh in 2021 to 12,892 GWh in 2022, an 11.7% increase. Over the same period, installed renewable capacity rose from 2,997.7 MW to 3,599.7 MW, or 20.1%—so nameplate capacity expanded faster than annual output.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: These are annual national totals, not a capacity factor, hourly generation profile, curtailment measure, data-centre allocation, contracted clean energy or campus-available power. Technology mix, commissioning dates, resource conditions, outages, curtailment and reporting scope can explain the different growth rates. The GWh and MW figures must not be divided into a site-availability claim or added to operating data-centre capacity.
OFFICIAL ANNUAL INSTALLED-CAPACITY SERIES · NATIONAL RENEWABLE FLEET · ENDS 2022 · NOT DATA-CENTRE-SPECIFIC
The official UAE series reports total renewable-energy installed capacity rising from 133.1 MW in 2016 to 3,599.673 MW in 2022—about 27.0 times the 2016 base, or a 2,604.5% increase. Selected intermediate points were 597.4 MW in 2018 and 1,931.5 MW in 2019.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is a historical national utility-scale installed-capacity series ending in 2022, not generation delivered, reserve margin, available grid capacity, data-centre allocation, contracted clean energy, hourly matching, transmission headroom, connection capacity, additionality or operating data-centre capacity. It cannot size a data-centre supply-demand gap or enter operating-capacity totals.
OFFICIAL REGULATOR SERIES · NATIONAL PEAK AND AVAILABLE CAPACITY · 2020–2024 · NOT SITE HEADROOM
Saudi Arabia's electricity regulator reports national peak load rising from 62.3 GW in 2020 to 74.8 GW in 2024, a 20.1% increase. Available capacity reached 80.4 GW in 2024, leaving a 5.6 GW national margin; the same official sequence shows that margin falling as low as 2.4 GW in 2022.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Peak load, available capacity and their national annual margin are system-level observations, not regional or substation headroom, spare data-centre capacity, firm connection capacity, contracted power or campus-available supply. The annual margin is not a guaranteed reserve at a named site and cannot be allocated to Riyadh, Jeddah, Eastern Province, NEOM or Oxagon without local network and connection evidence.
OFFICIAL ANNUAL GRID SERIES · NATIONAL PEAK · 2024–25 HISTORICAL · NOT DATA-CENTRE-SPECIFIC
India's Central Electricity Authority reports all-India peak electricity demand rising from 183,804 MW in 2019–20 to 249,856 MW in 2024–25, a 35.9% increase. Peak met reached 249,854 MW in 2024–25, leaving a reported national shortfall of 2 MW (0.0%)—strong system-level context that does not establish grid headroom at a data-centre site.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is an annual all-India peak-demand series, not data-centre load, available generation, regional reserve margin, connection capacity or deliverable site power. National peak met can coexist with local transmission constraints, curtailment, long connection queues and non-coincident regional peaks. The 2024–25 observation is historical, not a current 2026 adequacy reading, and cannot establish data-centre causation or enter operating-capacity totals.
OFFICIAL ANNUAL GRID SERIES · NATIONAL AVERAGE · NOT DATA-CENTRE-SPECIFIC
Official Singapore data show the national grid emission factor at 0.409 kg CO₂/kWh in 2019, peaking at 0.417 in 2022 before falling to 0.412 in 2023 and 0.402 in 2024. EMA attributes the latest decline to a higher solar share and lower generator emissions intensity, while natural gas still supplied 94.0% of the 2024 electricity mix.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is a generation-weighted annual national-grid average, not a data-centre-specific footprint, contracted clean-energy share, hourly marginal-emissions profile or proof of renewable additionality. It does not identify which generators serve a facility, distinguish imported electricity, or show DC-CFA2 compliance, so it cannot be converted into workload MW, operating capacity or an allocation score.
OBSERVED MONTHLY INDUSTRY INDEX · THROUGH AUG 2026 · 2015 = 100 · NON-SEASONALLY ADJUSTED · NOT CAPACITY
The official monthly production index for MSIC 26101 was 215.793 in August 2026, with 2015 = 100. The chart shows the latest twelve months of the retained January 2015–August 2026 series. This measures real industrial production relative to the base year; it is not an absolute output or capacity figure.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National industry context, not AI-chip output, wafer starts, fab or packaging capacity, exports, revenue, site load or data-centre demand. The non-seasonally adjusted item indices must not be added together or used to reconstruct overall manufacturing. The five-digit dataset covers selected industries; recent observations can be revised. No cloud or data-centre causation is inferred. Contains DOSM data accessed 1 November 2026, under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/); the displayed twelve-month excerpt is an NTRY selection. No endorsement implied.
OBSERVED MONTHLY INDUSTRY INDEX · THROUGH AUG 2026 · 2015 = 100 · NON-SEASONALLY ADJUSTED · NOT CAPACITY
The official monthly production index for MSIC 26102 was 188.889 in August 2026, with 2015 = 100. The chart shows the latest twelve months of the retained January 2015–August 2026 series. This measures real industrial production relative to the base year; it is not an absolute output or capacity figure.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National industry context, not AI-chip output, wafer starts, fab or packaging capacity, exports, revenue, site load or data-centre demand. The non-seasonally adjusted item indices must not be added together or used to reconstruct overall manufacturing. The five-digit dataset covers selected industries; recent observations can be revised. No cloud or data-centre causation is inferred. Contains DOSM data accessed 1 November 2026, under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/); the displayed twelve-month excerpt is an NTRY selection. No endorsement implied.
OBSERVED OFFICIAL MONTHLY SERIES · JAN 2018–JUN 2026 · H1 2026 PROVISIONAL · NATIONAL SCOPE · NOT FIRM POWER
Official monthly records show total national electricity consumption increasing from 94.587 TWh in H1 2025 to 100.381 TWh in H1 2026, a source-derived 6.1% increase using the same January–June window. The dataset’s broad commercial category increased from 65.841 TWh to 70.390 TWh, or 6.9%. The companion supply ledger reports the same total monthly values as the consumption ledger; this is an accounting relationship, not evidence of spare power. Both public datasets run monthly from January 2018 through June 2026.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National, multi-sector context—not data-centre consumption, Johor or Peninsular load, firm MW, reserve margin, connection availability, curtailment, tariff economics or site headroom. The commercial category includes industrial and mining demand and cannot be attributed to data centres. H1 totals are calculated by summing the six published monthly MKWh observations and dividing by 1,000; percentages use unrounded source values. The latest six months are provisional and may be revised. Identical supply and consumption totals must not be subtracted to manufacture a headroom figure.
OFFICIAL MONTHLY GRID SERIES · JUNE COMPARISON · 2026 PROVISIONAL · NOT DATA-CENTRE-SPECIFIC
Malaysia's official monthly dataset reports total electricity consumption of 13,290.32 million kWh in June 2020 and 17,075.87 million kWh in June 2026, a 28.5% increase across the same-month comparison. June 2026 was 6.6% above June 2025, showing a rising system-demand context without assigning that growth to data centres.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is an all-sector national consumption series, not data-centre electricity use, peak demand, available generation, grid headroom or deliverable power. The dataset does not isolate data centres from commercial, industrial, mining or domestic load, and its latest six months are provisional and may be revised. The sequence therefore cannot establish causation, size the data-centre supply-demand gap or enter operating-capacity totals.
REPORTED PASSED 7 OCT 2026 · COMMENCEMENT UNVERIFIED · MINISTRY EXPECTED 2027 · 2024 NATIONAL ELECTRICITY USE · RATED-LOAD LICENSING GATES
CNA reported Parliament passed the Bill on 7 October 2026. The 6 October opening speech sets sustainability licensing for new and existing data centres at ≥3 MW rated critical IT load, including own-use enterprise facilities. The ≥10 MW security/resilience regime covers cloud and co-location data-centre facility services. Critical IT load is maximum design IT/network power, not instantaneous consumption. The 7 October closing speech reports about 6.5 TWh used in 2024, slightly more than 10% of national electricity consumption, and PUB's WUE target of ≤2 m³/MWh for new and existing data centres. MDDI expected staged commencement in 2027; existing operators would have six months from commencement to apply, with continued operation while IMDA considers their application.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Passage is not commencement: current Gazette commencement dates, final Codes of Practice and substantive compliance deadlines were not verified. The announced 2027 timing is a ministry expectation, not an appointed date; the six-month window is for licence applications, not a deadline to meet every requirement. The 2024 demand statements are rounded national aggregates, not facility load, grid headroom, occupancy or utilisation. The WUE figure is a policy target, not measured performance or a verified mandatory threshold. Neither a repeatable annual series nor a published calculation method is supplied; this remains contextual and excluded from governed longitudinal totals. The licensing regimes have different service scope and exemptions; a ≥10 MW own-use enterprise facility is not automatically a major third-party service.
OFFICIAL 2024 STATE COMPARISON · TOP SIX PUBLISHED ROWS · PROVISIONAL · NOT DATA-CENTRE WATER USE
The official dataset reports 5,077 MLD of national non-domestic water consumption in 2024. Selangor, which includes Kuala Lumpur and Putrajaya in this dataset, represented 1,591 MLD or 31.3%; Johor was second at 610 MLD or 12.0%, just ahead of Sarawak at 586 MLD or 11.5%. The three largest published rows represented 54.9% of the national total.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Point-in-time 2024 comparison of broad non-domestic customer accounts, not a trend and not data-centre water use. The chart shows the six largest published state/territory rows rather than the complete national distribution. Selangor includes W.P. Kuala Lumpur and W.P. Putrajaya, so it is not a standalone state-only value. Consumption is not water-system capacity, allocation, entitlement, drought resilience, potable or non-potable supply, cooling-water availability or campus deliverability. The source marks 2024 as provisional and subject to revision. CC BY 4.0.
OFFICIAL 2019–2024 JOHOR NON-DOMESTIC SERIES · 2024 PROVISIONAL · NOT DATA-CENTRE WATER USE
Malaysia's official open dataset reports Johor non-domestic water consumption at 610 million litres per day in 2024, up 6.6% from 572 MLD in 2023 and 20.8% from the 2021 trough. The 2024 level was only 1.8% above 2019, preserving the pandemic-era dip and subsequent recovery in one scope-consistent state series.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Johor annual non-domestic customer-account context only. The category includes all non-domestic accounts and does not isolate data centres, hyperscalers, industrial parks or cooling demand. It is consumption, not water-system capacity, allocation, entitlement, drought resilience, potable or non-potable supply, cooling-water availability or campus deliverability. The source marks the most recent year, 2024, as provisional and subject to revision. CC BY 4.0.
OFF-TAKE COMMITTED · Q1 2027 ENERGISATION TARGET · NOT OPERATING
Bitdeer reports off-take commitments covering more than 70% of A201's approximately 21.7 MW capacity in Johor Bahru, with more than US$1.7 billion of expected five-year site revenue and energisation targeted for Q1 2027.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: A201 is not yet energised. The approximately 21.7 MW figure is planned site capacity, while more than 70% is a company-reported commitment floor—not an exact contracted-load value. Applying that floor to the rounded site capacity implies more than approximately 15.2 MW; this is transparent NTRY arithmetic, not a separately source-reported measurement. The uncommitted remainder is not evidence of available or deliverable capacity. Expected revenue, delivery timing, customer performance, equipment deployment and power availability remain forward-looking.
TERM SHEET ENTERED 28 SEP 2026 · DEVELOPMENT STAGE · 2029 CONDITIONAL COD · NOT OPERATING
SD Guthrie’s Bursa filing says G3nerasi Kinta entered a BESC term sheet on 28 September 2026 for a proposed hybrid utility-scale solar PV plant in Perak with total minimum net capacity of 680 MWac, coupled with a four-hour battery energy storage system. An unnamed US-based multinational technology company is the intended data-centre offtaker under CRESS; definitive BESC and CRESS agreements are targeted for Q1 2027, and COD is scheduled for 2029 subject to TNB access and required approvals.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is a term sheet and the project remains at development stage. The 680 MWac is minimum net renewable-generation capacity—not data-centre IT load, contracted grid headroom, average delivered power, firm capacity or operating capacity. The filing discloses four-hour BESS duration but not its MW or MWh size. The offtaker is unnamed; definitive BESC and CRESS agreements remain unexecuted; and the 2029 COD depends on a Renewable Energy Supply Access Agreement with TNB and required CRESS approvals. SD Guthrie separately describes an up-to-1.2 GW solar-assets partnership target, but the sources do not establish a common capacity basis or confirm that the 680 MWac counts toward it, so no ratio, residual or additive total is shown.
OFFICIAL NATIONAL SECTOR SNAPSHOT · 2024 · POINT-IN-TIME COMPARISON · CONTEXT ONLY
Six official national sector records sum to 155,279.18 GWh in 2024. Commercial use is the largest reported category at 40.6%, followed by residential at 31.1% and industry at 17.4%. The chart compares sectors within the same national year; it is a demand-composition snapshot, not a trend.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Annual electricity consumption is energy, not peak MW, installed capacity, reserve margin or campus-ready power. The source does not isolate data centres or establish their sector classification. National records are for 2024; the workbook's emirate breakdown is for 2023 and is excluded from this comparison. A missing emirate value is unknown, not zero. Shares use only the six reported UAE sector rows; no EtihadWE service-area series is added. Source: Ministry of Energy and Infrastructure via Bayanat, CC BY-SA 4.0; label trimming, sorting, share calculations and chart rendering are NTRY modifications.
GOVERNMENT SECTOR SNAPSHOT · INSTALLED / FORECAST / ANNOUNCED SEPARATE
The Government of India reports 1.57 GW of installed data-centre power capacity as of August 2026, projects nearly 8 GW by 2030, and distinguishes about US$70 billion of investment underway from a further US$90 billion of announced projects.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: The installed-capacity figure is a national aggregate with no operator or site reconciliation. The 8 GW figure is a projection; investment underway is not verified spend or delivered capacity, and announced projects are not committed or operating assets.
CAPACITY SUPPORT CONDITIONAL · PRIMARY GRID SUPPLY · LOW-CARBON REPORTING REQUIRED · NOT OPERATING
Singapore's energy ministry says the four supported DC-CFA2 data centres, equal to about 12% of national data-centre capacity, will draw electricity primarily from the national grid with on-site solar supplementation. More than half of the supported capacity is committed to Singapore-based low-carbon generation, and operators must report total electricity use and substantiate the low-carbon share to IMDA and EDB.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: The 200 MW is supported data-centre capacity, not measured electricity demand or operating IT load. The reply gives no grid-versus-solar MW split, commissioning dates, delivered power, fuel-specific allocation or utilisation. The more-than-half low-carbon share is a future condition of capacity support, not verified generation or consumption today.
OFFICIAL VERSION 4 FORECAST · SYSTEM-WIDE RESERVE MARGIN · 2030 UPPER-CASE SHORTFALL · NOT SITE HEADROOM
EMA's 8 September 2026 Version 4 RFP forecasts upper-bound system peak demand rising from 9.2 GW in 2026 to 14.4 GW in 2034. Against projected supply, the corresponding reserve margin is 27% in 2029, 18% in 2030, 9% in 2031 and -1% in 2034. EMA is seeking one new CCGT unit of at least 600 MW for 2031 and up to five units of at least 500 MW each for 2032.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: These are national planning forecasts and RFP requirements, not observed demand, awarded or constructed generation, operating capacity, firm supply, connection headroom or a facility SLA. The upper-bound demand scenario includes prospective projects across energy-intensive sectors such as semiconductors and data centres, but it does not quantify data-centre load or establish causation. Reserve margin is system-wide, not campus-ready power. The supply projection excludes imports until they are firm, and the requested 2031–2032 units receive zero operating-capacity credit. Values use Version 4's current figures; redlined superseded values are excluded.
GOVERNMENT EVENT SUMMARY · MULTIPLE FUTURE COMMITMENTS · NOT OPERATING
Saudi MCIT reports a November 2026 availability target for Microsoft Azure, a December 2026 launch target for AWS, up to 50 MW for the AWS–HUMAIN AI zone by 2028, and separate expansion ambitions of 192 MW for Al Moammar Information Systems and 65 MW for Khuzam Digital Valley by 2033.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: These are government-reported announcements and delivery commitments, not one comparable capacity series. The 50 MW is planned for 2028; 192 MW is an intended post-expansion capacity; 65 MW is an expandability ceiling; and cloud-region availability dates do not disclose delivered power, utilisation or customer absorption. The event's nearly US$15 billion total spans several technology sectors and is not data-centre spend.
NON-BINDING MOU · DEFINITIVE AGREEMENTS TARGET Q4 2026 · SITE UNDISCLOSED · NOT OPERATING
The parties target approximately 300 MW of total IT load within five years from an unspecified commercial launch, 100% renewable-energy penetration, at least 50% self-generated on-site or associated green power, and PUE of no more than 1.20. Feasibility, site planning, engineering and commercial structuring remain next steps; definitive agreements are targeted by the end of Q4 2026.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is a preliminary, non-binding MoU-stage target with no named site or commercial-launch date. The approximately 300 MW figure is planned IT load—not delivered or contracted power, grid draw, renewable generation, firm supply, constructed, commissioned or operating capacity—and receives zero operating-capacity credit. The renewable-energy and PUE figures are targets, not measured performance. Definitive agreements have not been executed; feasibility, site, grid connection, regulatory approvals, financing, construction, energisation, commissioning, customers and operating evidence remain unverified.
GOVERNMENT-REPORTED H1 2026 ABSORPTION SHARE · MW BASIS UNDISCLOSED
India's Ministry of Commerce and Industry reports that hyperscalers represented more than 80% of data-centre absorption in the first half of 2026. The same roundtable identified reliable power, Day-1 sanctioned load, dual feeders and power-ready land as constraints on faster deployment.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: Absorption is a market-share indicator, not disclosed occupied MW, utilisation, contracted workload or future demand. The government release does not publish the underlying denominator, methodology, geography split or operator reconciliation, so this signal cannot close the numerical supply-demand gap.
CONSTRUCTION ANNOUNCED · FIRST 100 MW TARGET 2028 · 360 MW PHASE · NOT OPERATING
HUMAIN and DataVolt say construction is underway on 100 MW of AI-ready capacity at Oxagon, inside a stated 360 MW first phase targeted for service in 2028. DataVolt places that phase within a wider planned 1.5 GW campus, separating the portion now described as under development from the longer-term masterplan.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is a company construction announcement, not independent evidence of completed works, energisation, customer occupancy or utilisation. The 100 MW is under development and not operating; 360 MW is the stated first-phase plan and 1.5 GW is the wider planned campus ceiling. The 2028 service date, renewable-energy integration and cooling design remain forward-looking.
100% OFF-TAKE COMMITTED · Q1 2027 SERVICE TARGET · NOT ENERGISED OR OPERATING
Bitdeer's official releases moved A102 from approximately 50% contracted on 19 August 2026 to sold out on 31 August. Five-year off-take commitments cover the planned 9.5 MW site capacity and represent more than US$800 million of expected contract revenue; service is targeted for Q1 2027.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is a two-disclosure company-reported commitment-coverage sequence, not observed customer load, utilisation, energised capacity, operating capacity, delivered power or a facility SLA. The 19 August value was stated as approximately 50%, while the 31 August release said A102 was sold out; the chart must not be used to infer a precise daily contracting rate. The 9.5 MW is planned site capacity, and revenue, service timing, customer performance, customer prepayments and delivery remain forward-looking.
PROVISIONAL 50 MW ALLOCATION · DELIVERY DATE UNDISCLOSED · NOT OPERATING
Digital Realty says Singapore's DC-CFA2 process provisionally allocated 50 MW for a planned fourth facility on Jurong Island, designed to power more than half of its capacity through green-energy pathways. The company separately reports three existing Singapore data centres with about 84 MW of combined capacity.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: The 50 MW allocation is provisional, is nested within the 200 MW DC-CFA2 tranche and must not be added to that tranche. The new facility is not operating; no final approval, commissioning date, delivered power, customer occupancy or utilisation is disclosed. The green-energy share is a design commitment, while the existing 84 MW is a company-reported combined platform figure—not available capacity.
OFFICIAL 2015–2025 NATIONAL SERIES · CONTEXT ONLY · NOT DATA-CENTRE-SPECIFIC
The official HEESI 2025 table reports national electricity utilization increasing from 289,311 GWh in 2019 to 455,001 GWh in 2025. Industry accounted for 232,277 GWh, or 51.0% of the 2025 total, across PLN sales and non-PLN consumption.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National annual electricity-utilization context only. It combines PLN electricity sales with non-PLN consumption and does not isolate data centres, regional grid constraints, connection queues, available capacity, peak demand or site-level power headroom. The publication page does not state a reuse licence.
GOVERNMENT-REPORTED INSTALLED BASE · SOURCE-DATED SNAPSHOT
A Lok Sabha response published by PIB reports national installed capacity at about 1,575 MW, up from 375 MW in 2020, and identifies AI and high-performance computing as demand drivers.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is a government-reported national aggregate, not a reconciled operator or site inventory, available tenant capacity or an AI-ready subset. The release does not specify the capacity basis.
DC BYTE Q1 2026 SNAPSHOT · LIVE + UNDER CONSTRUCTION + COMMITTED · MIXED STATUS · CONTEXT ONLY
DC Byte's Q1 2026 Melbourne snapshot reports 1,858.6 MW of wholesale colocation capacity (83%), 321.6 MW of self-build public-cloud capacity (14%) and 45.49 MW of retail colocation capacity (2%), with three smaller self-build categories. The report ranks Melbourne sixth in APAC at 4.44 as a Balanced Growth Metro and eighth in APAC's separate Future Growth Metro list. DC Byte describes recent demand as hyperscale and AI-led while identifying grid delivery and site connectivity as the main supply constraints.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is a proprietary, point-in-time market-composition snapshot covering Live, Under Construction and Committed projects together. It is not an operating-capacity total, available supply, vacancy, utilisation, contracted customer load, firm grid headroom, market share of demand or a repeated series. The category totals must not be added to other Melbourne or Australian figures without matching facility, status and effective-date scope. Percentages are source-rounded, and operator-type composition does not identify individual counterparties or prove operator concentration, energisation or delivery timing.
DC BYTE 2026 RIYADH SNAPSHOT · REPORTED LIVE PROJECT STATUS · CONTEXT ONLY · NOT CAPACITY HEADROOM
DC Byte's 2026 report ranks Riyadh tenth within EMEA at 3.70 and classifies it as an Emerging Growth Metro. Its Riyadh market-share table, filtered to colocation and self-build facilities with project status Live, reports Center3 at 41.4 MW (46%), Others at 22.39 MW (25%), Saudi Central Bank at 8.1 MW (9%), Mobily at 6.93 MW (8%), Ezditek at 6 MW (7%) and DAMAC Digital at 5 MW (6%). The listed capacity values total 89.82 MW.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is a proprietary, point-in-time metro snapshot—not an NTRY score, a national capacity total, spare capacity, utilization or a repeated series. DC Byte labels the table's project status Live, but the report does not disclose the capacity basis, facility-level reconciliation or observation timestamp beyond publication; the values must not be compared or added to government-reported national operating capacity or DC Byte total supply across mixed lifecycle stages. Percentages are source-rounded and sum to 101%. The table does not prove customer concentration, available headroom, contractual access or the delivery of any planned cloud region.
DC BYTE 2026 APAC INDEX · TOKYO CONSTRAINED CORE · CONTEXT ONLY · NOT OPERATING CAPACITY
DC Byte's APAC-only 2026 metro index ranks Johor first at 4.78, Sydney second at 4.76 and Tokyo fourth at 4.61, ahead of Mumbai at 4.40 and Kuala Lumpur at 4.18. Tokyo is classified as a Constrained Core Metro: demand and/or existing depth are stronger than delivery. The report says power connections in Tokyo can take 8–10 years and points to Osaka, Kyushu and other regional markets as possible outlets for the next growth wave.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is a proprietary, point-in-time regional composite—not an NTRY score, market share, capacity total, capacity headroom or repeated trend. Scores are benchmarked only against APAC peers and must not be compared across regions. The weighted index uses demand 40%, delivery 35% and depth 25%; delivery combines under-construction and committed MW, neither of which is guaranteed operating capacity. The selected chart is not the complete top-ten ranking, and DC Byte's category does not prove that any named project will be delayed or displaced.
ANNOUNCED INVESTMENT · 2026–2030 · US$13BN AI/CLOUD INCREMENT · NOT OPERATING CAPACITY
Amazon says it will invest an additional US$13 billion to expand AI and cloud infrastructure in India, including AWS data-centre capacity in Mumbai and Hyderabad. The additional commitment takes Amazon's stated 2026–2030 India investment across all businesses to US$48 billion. The scopes are retained separately: US$13 billion is the incremental AI/cloud commitment; US$48 billion is the broader all-business total.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is a company-announced spending plan, not money already deployed, operating data-centre capacity, IT load, contracted power, available grid supply, connection headroom or a delivery schedule. The US$13 billion AI/cloud commitment is incremental within the broader US$48 billion 2026–2030 India plan; the two values are overlapping scopes and must not be added. Amazon does not disclose MW, annual spend, named-facility commissioning dates or the portion attributable to compute hardware, buildings, land, energy or networking.
OFFICIAL 2021–2024 SSIC 2020 SERIES · BROAD INFORMATION & COMMUNICATIONS CATEGORY · NOT DATA-CENTRE LOAD
The official SINGSTAT dataset reports electricity consumption by the broad Information and Communications category increasing from 4,029.0 GWh in 2021 to 6,072.0 GWh in 2024. Its share of national electricity consumption rose from 7.53% to 10.53% over the same SSIC 2020 classification window.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National annual industry-classification context only. The Information and Communications category is broader than data centres and includes other information, communications and digital-service activities; the source does not isolate data-centre electricity use. The series is demand, not available capacity, reserve margin, grid-connection headroom, contracted power or deliverable site power. Values before 2021 use SSIC 2015, so this chart begins in 2021 and does not bridge the classification break. No hyperscaler causation is inferred.
OFFICIAL 1975–2025 NATIONAL GENERATION SERIES · 2025 CONSUMPTION UNAVAILABLE · CONTEXT ONLY
The official SINGSTAT dataset reports Singapore electricity generation increasing from 54,142.3 GWh in 2019 to 60,410.4 GWh in 2025, an 11.6% rise. Generation increased 1.3% from 2024 to 2025; the same dataset marks 2025 electricity consumption as unavailable.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National annual generation context only. Generation is not available capacity, reserve margin, grid-connection headroom, contracted or delivered site power, or data-centre load. The source reports 2025 electricity consumption as unavailable, so no 2025 generation-consumption balance or surplus is inferred. Dataset classification footnotes are retained; sector rows spanning 2020–2021 require care because the source changes from SSIC 2015 to SSIC 2020.
COMPANY-REPORTED OPENINGS · DATED 2025 / 2026 MILESTONES · FACILITY IDENTITIES UNRESOLVED · CONTEXT ONLY
Alibaba Cloud said on 9 June 2026 that it had opened two new data centres in Johor, bringing its Malaysia total to five. Its 2 July 2025 release had reported a third Malaysian data centre opening on 1 July 2025. The chart compares these two dated company-reported milestones, not a continuously observed facility census.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Company-reported data-centre counts are not an independently audited physical-facility inventory. The two Johor openings are included in the Malaysia total of five and must not be added again. Individual facility names, addresses, ownership versus leased premises, MW, utilisation and commissioning evidence are unresolved. Cloud regions and availability zones are different units; one availability zone may contain one or multiple data centres. No facility pins, capacity totals, accepted observations or Signals are created.
GOVERNMENT-REPORTED OPERATING BASE · NATIONAL AGGREGATE · BASIS UNDISCLOSED
Saudi Arabia's Ministry of Communications and Information Technology reports national operational data-centre capacity of 68 MW in 2021, more than 440 MW in 2025 and 467 MW in Q1 2026. The ministry also reports more than 60 data centres operated by more than 20 companies.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: These are government-reported national aggregates. The capacity basis and facility-level reconciliation are not disclosed, and the 2025 value is a lower bound. The sequence is useful context but is not yet a governed comparable series or evidence of available IT load.
JV ANNOUNCED · TWO PHASES · OPERATING STATUS UNVERIFIED
The official release describes a project of up to 60,000 square metres, with 29 MW of stated first-phase power followed by 100 MW of additional committed power.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: The joint-venture announcement does not prove that 29 MW is energised or occupied. The additional 100 MW is committed power, not operating capacity or verified IT load, and delivery dates are undisclosed.
OFFICIAL ANNUAL ENTERPRISE ADOPTION SERIES · DEMAND PROXY · NOT WORKLOAD MW
Official annual data show business usage of cloud-computing services rising from 19.6% in 2020 to 41% in 2025, while reported business use of artificial intelligence rose from 3.9% to 23.5%. The sequence strengthens Singapore's demand-side case as an adoption signal, but does not size data-centre demand.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is a national enterprise technology-adoption series covering all enterprises, not contracted workload, cloud revenue, data-centre occupancy, utilisation, electricity demand or MW. Usage may be served from infrastructure outside Singapore, so the series cannot close the numerical supply-demand gap or enter capacity totals.
OFFICIAL ANNUAL OWN-GENERATION HISTORY · 2020–2025 · SOURCE BREAK UNEXPLAINED · CONTEXT ONLY
EtihadWE's official annual workbooks report own electricity generation of 439.0515 GWh in 2020 and 505.2375 GWh in 2021, followed by 3.333 GWh in 2022, 0.2174 GWh in 2023 and zero in 2024 and 2025. Separately, EtihadWE's service-area consumption workbook reports consumption rising to 20,475 GWh in 2025. The divergence makes the source of served electricity a material diligence gap; the files do not reconcile purchases, transfers or network supply.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is EtihadWE-reported generation from its own power plants, not total electricity supplied to its service area, installed or available generation capacity, contracted power, reserve margin, network imports, regional headroom or campus-ready supply. The generation and consumption workbooks cannot be subtracted to infer a deficit because purchases, transfers, losses and boundary continuity are not reconciled. The sharp 2021–2022 break may reflect an operational or reporting change; the source does not explain it. The 2024 workbook says “Jan. 2024” in metadata despite containing a 2024 row, a disclosed timing inconsistency. Source zeros are preserved without imputation. Source: EtihadWE via Bayanat, CC BY-SA 4.0; chart selection and percentage calculation are NTRY modifications.
OFFICIAL ANNUAL SERVICE-AREA ENERGY · 2020–2025 · SOURCE-REPORTED HISTORY · CONTEXT ONLY
The official 2025 workbook reports EtihadWE service-area electricity consumption of 20,475 GWh, up from 18,820 GWh in 2024 and 14,409.24 GWh in 2020. The file contains annual observations from 2010 to 2025; the chart shows its latest six years, with all sectors kept within the publisher's service-area scope.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: This is annual electrical energy used across all sectors within EtihadWE's service area, not a UAE national total, data-centre load, peak MW, installed or operating capacity, reserve margin or available power at a campus. Historical service-boundary and methodological continuity are not independently established. Do not subtract this consumption from EtihadWE own generation to infer a shortage: purchased electricity and network transfers are not reconciled. Values in the chart are rounded to the nearest GWh; source decimals are retained in the evidence record. Source: EtihadWE via Bayanat, CC BY-SA 4.0; chart selection, rounding and growth calculations are NTRY modifications.
OFFICIAL 2020–2024 NATIONAL RELIABILITY SERIES · SUSTAINED INTERRUPTIONS · NOT A FACILITY SLA
EMA reports a national System Average Interruption Duration Index (SAIDI) of 0.26 minutes per customer in 2024 and a System Average Interruption Frequency Index (SAIFI) of 0.006 interruptions per customer. The five-year SAIDI sequence was 0.17, 0.13, 0.16, 0.14 and 0.26 minutes from 2020 to 2024.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: National annual customer-average reliability context only. SAIDI and SAIFI cover sustained interruptions; they do not measure voltage dips, momentary events, power quality, event concentration, substation redundancy, data-centre availability, contractual SLA performance or a named campus's resilience. The 2024 increase from 0.14 to 0.26 minutes is one annual movement and is not treated as a structural trend.
OFFICIAL ANNUAL STATISTICS · NATIONAL NETWORK ENERGY · 2023–2024 · CONTEXT ONLY
Saudi Arabia's official electricity workbook reports annual electrical energy sent to the national network rising from 380,878 GWh in 2023 to 402,628 GWh in 2024, a 5.7% increase. This is system throughput across all uses, not a measure of data-centre demand or power available to a new campus.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Annual energy sent to the national network is not installed capacity, available capacity, reserve margin, regional or substation headroom, contracted power, data-centre electricity demand or campus-ready supply. The annual totals do not reveal coincident peak conditions, local constraints, losses, curtailment or connection timing and cannot size a data-centre supply-demand gap or enter operating-capacity totals.
OFFICIAL 2020–2024 REGIONAL PEAK SERIES · NETWORK AREAS · NON-COINCIDENT PEAKS · NOT SITE HEADROOM
GASTAT's official workbook reports regional peak load increasing between 2020 and 2024 from 19.8 to 24.7 GW in the Central region, 19.0 to 22.1 GW in the Eastern region, 15.8 to 19.3 GW in the Western region and 6.4 to 8.2 GW in the Southern region. Southern growth was fastest at 28.1%, while Central and Eastern represented 62.9% of the four published 2024 regional peaks.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Regional annual peaks may occur at different times, so their sum is not a coincident national peak and is not compared arithmetically with national available capacity. The regions are electricity-network reporting areas, not named substations or campus service territories. Peak load is demand, not reserve margin, spare capacity, connection headroom, contracted power or deliverable site power; no data-centre causation is inferred.
COMPANY PROJECTION · >US$7.3BN EXPECTED BY END-2025 · 2026–2029 PLAN · NOT CONFIRMED SPEND OR OPERATING CAPACITY
Microsoft's 3 November 2025 disclosure projected that, by the end of 2025, it would have invested and spent just over US$7.3 billion in the UAE. That year-end projection included more than US$4.6 billion of capital expenditure on advanced AI and cloud data centres; neither amount was a confirmation of actual spend completed on the publication date. Microsoft separately planned to spend more than US$7.9 billion during 2026–2029, including more than US$5.5 billion of AI/cloud infrastructure capex, within a US$15.2 billion 2023–2029 programme.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is a 3 November 2025 company projection, not audited deployment evidence. The >US$7.3 billion total and >US$4.6 billion AI/cloud data-centre capex component were amounts Microsoft expected to have invested and spent by the end of 2025—neither was confirmed actual spend as of publication. The US$15.2 billion programme total comprises that projected end-2025 amount and the planned 2026–2029 phase; the phase totals and their AI/cloud infrastructure capex components are nested scopes and must not be added as separate commitments. The end-2025 amount also included a US$1.5 billion G42 equity investment and more than US$1.2 billion of local operating expenses and cost of goods sold; the later phase included almost US$2.4 billion of planned local operating expenses and cost of goods sold. Microsoft did not disclose facility-level allocations, MW, power connections, annual deployment cadence, commissioning dates, operating data-centre capacity, IT load, delivered compute, contracted power, available grid supply or site readiness.
OFFICIAL 2023–2024 SEC DISTRIBUTION RELIABILITY COMPARISON · NATIONAL SYSTEM AVERAGE · NOT A FACILITY SLA
SERA's official 2024 statistical booklet reports the SEC distribution system's general-average SAIDI improving from 82.16 minutes per consumer in 2023 to 72.98 minutes in 2024. SAIFI also improved from 1.81 to 1.58 interruptions per consumer, a 12.7% reduction.
Historical series · repeated dated points · context only · excluded from governed totals
Caveat: Two-point SEC distribution-system general averages, not a long-run trend and not a measure of a named data-centre campus. SAIDI and SAIFI cover sustained customer interruptions; they do not establish voltage quality, momentary-event exposure, substation or feeder redundancy, event concentration, connection headroom, backup performance, contractual availability or facility SLA compliance. The values cannot be used to infer spare capacity or causal effects from data-centre load.
ANNOUNCED 1 GW CLUSTER · FIRST 200 MW EXPECTED 2026 · DELIVERY NOT VERIFIED
OpenAI announced a planned 1 GW Stargate UAE compute cluster in Abu Dhabi, with an initial 200 MW expected to go live in 2026. The announcement positions the cluster as the first international Stargate deployment and says its potential service radius could reach up to half the world's population.
Stage comparison · statuses are not additive · context only · excluded from governed totals
Caveat: This is OpenAI's May 2025 partnership announcement, not evidence that construction, energisation, commissioning, customer occupancy or utilisation has occurred. The 1 GW is the planned cluster scale and the 200 MW is an expected initial tranche, not verified operating capacity. The 2,000-mile reach is described as potential coverage, not measured demand or service availability.
OFFICIAL REGIONAL ALL-SECTOR HISTORY · 2012–2021 · 2021 COMPARISON · CONTEXT ONLY
The official regional dataset reports 126,488.92 GWh of UAE electricity consumption in 2021. Abu Dhabi contributed 56,134.80 GWh and Dubai 44,993 GWh, together representing 79.9% of the four published regional totals. The national total rose 44.1% from 87,789.82 GWh in 2012, but the history ends in 2021 and does not isolate data-centre load.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: This is historical annual all-sector electricity consumption by published service region, not current demand, data-centre load, peak MW, available generation, reserve margin, substation headroom, contracted power or campus-ready supply. The history ends in 2021. Abu Dhabi sector detail changes sharply between 2019 and 2020 while the publisher's all-sector total remains coherent, so NTRY uses the supplied all-sector rows rather than reconstructing totals. Northern Emirates 2021 values exactly repeat 2020 and may be carried forward; no freshness inference is made. Regional service boundaries and methodology continuity are not independently established. Source: Ministry of Energy & Infrastructure via Bayanat, CC BY 4.0; label normalization, chart selection, share and growth calculations are NTRY modifications.
OFFICIAL DUBAI SECTOR HISTORY · 2019–2021 · TWO SEPARATE SECTORS · CONTEXT ONLY
The official workbook reports Dubai/DEWA commercial consumption of 22,223 GWh in 2019, 21,160 GWh in 2020 and 24,205 GWh in 2021. Residential consumption rose from 13,144 to 13,992 to 14,853 GWh over the same years. Commercial use rebounded 14.4% in 2021. The chart shows the latest three years in two separate sector histories; all 20 source observations for 2012–2021 are retained in the evidence record.
Point-in-time comparison · not a trend · context only · excluded from governed totals
Caveat: Historical annual energy only; this workbook ends in 2021 and does not establish current demand. The commercial category does not isolate data centres or contracted cloud workload. Opposite sector changes do not identify a cause, substitution, peak-load shape or spare grid capacity. Sector definitions, service boundaries and methodological continuity are not independently established. Do not join the two sectors into one time series, add national totals or connect these figures to the separate UAE 2024 sector snapshot without reconciliation. Agriculture N/A records, other entities and all-sector/total rows are excluded. Source: Ministry of Energy and Infrastructure via Bayanat, CC BY 4.0; sector-label trimming, year selection, chart rendering and percentage calculations are NTRY modifications.
Source-linked context and NTRY interpretation. Open each finding to inspect its limits.
Track the CRESS project awards, financing, connection and commissioning milestones alongside each Google campus’s construction, energisation and measured load. Underwrite firm site power only against utility-dated capacity and delivery evidence.
Named CRESS projects and awarded MW; signed offtake and grid-connection terms; generation and battery commissioning dates; TNB network-upgrade milestones; and campus-level contracted and measured demand with time-matched clean-energy supply.
Project stage and scopeANNOUNCED · >1 GW PLANNED CLEAN-ENERGY CAPACITY · CRESS / TNB · NOT OPERATING SUPPLY
Announced future grid capacity, not verified operating generation, firm campus supply or a time-matched clean-power share. Project-level capacity, commissioning dates, Google campus load and hourly matching were not disclosed. The roughly 50 MW Kedah figure is separately reported by The Star and is not the total 1 GW plan. No campus-level offset or grid headroom is inferred.
Evidence period / effective date: 2026-11-04
Use the country and region list to organize facility verification and cloud-market diligence. Keep service topology separate from physical-site inventory, and withhold any exact SEA AZ total until Alibaba reconciles Johor.
A consistent dated official AZ table; named buildings and operators; tenancy or ownership evidence; facility-level permits, commissioning and MW disclosures; and the effective dates of each region's service availability.
Project stage and scopeOFFICIAL WEBSITE SNAPSHOT · SIX CLOUD REGIONS · FIVE COUNTRIES · JOHOR AZ 2 / 3 UNRESOLVED · NOT FACILITY COUNTS
This is a dated website snapshot, not a publisher release date or a physical-facility inventory. A cloud region is a service geography; an availability zone may contain one or multiple data centres. The chart counts listed cloud regions only. Johor is excluded from any asserted exact AZ total because official pages disagree between 2 and 3. Do not turn regions or AZs into physical sites, ownership claims, MW, utilisation, operating-capacity totals or facility pins.
Evidence period / effective date: 2026-11-10
Use the built series as regional context for the 2027–2031 build-out question. Allocate by country or site only after lifecycle-consistent facility, firm-power, connection, construction and commissioning evidence is available.
A later edition using the same built-capacity definition; country-level status-separated capacity; named-project construction and commissioning; firm grid connections; and delivered-power evidence.
Project stage and scopeREPORTED BUILT-CAPACITY HISTORY · ANNOUNCED 11 GW SEPARATE · CONTEXT ONLY · NOT GOVERNED CAPACITY
Built and announced are separate publisher statuses. Built is the report label and is not independently verified by NTRY as operating or commissioned capacity; announced capacity is not delivered capacity. The values are regional, not country totals, a site inventory, contracted power or available headroom. YTD 2026 is the report's latest stock observation, not a full-year flow. Source note: Bain analysis; datacenterHawk. Required citation: Google, Temasek and Bain, e-Conomy SEA 2026. NTRY selected and redrew the values without reproducing the source chart.
Evidence period / effective date: 2026-10-21
Choose a market on the map, preparedness graphic or demand charts to open its evidence and decision analysis. Cross-market values retain their own dates, units and definitions; they are not an allocation ranking.
THE UAE EVALUATION SYNOPSIS