Order records by published competition rank and stack each axis's weighted contribution for the first 20 markets.
#1 Northern Virginia (incl. ext.)4.63
#2 Tokyo4.57
#3 Singapore4.43
#4 London4.40
#5 Sydney4.26
#6 Dallas-Fort Worth4.18
#7 Sao Paulo4.18
#8 Jakarta4.15
#9 Phoenix4.10
#10 Mumbai4.09
#11 Atlanta4.06
#12 Chicago4.02
#12 Frankfurt4.02
#14 Paris3.89
#15 Melbourne3.83
#16 Osaka3.81
#17 Seoul3.75
#18 Hong Kong3.66
#19 Bangkok3.64
#20 Johor3.61
Axis A
Axis B
Axis C
Axis D
Published market dataset · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 08 · The framework
Axis weights in the composite score
weight bars
Render the four published axis weights, which sum to 100% of the composite.
Axis A · 40% of compositeCommercial colocation
Operational scale, committed growth, operator density and enterprise workload depth.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 09 · The framework
The 16 metrics and their intra-axis weights
metric grid
Group the 16 published metrics by axis and display each metric's weight within its axis.
A1Operational colocation capacity
30% of Axis A · 12.0% of composite
A2Committed colocation growth
20% of Axis A · 8.0% of composite
A3Competitive operator density
25% of Axis A · 10.0% of composite
A4Enterprise workload depth
25% of Axis A · 10.0% of composite
Operational scale, committed growth, operator density and enterprise workload depth.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 25 · Results
Tier counts by region
table
Count final-tier records within each published region and calculate Tier 1 share.
Tier 16Tier 210Tier 38Tier 41Total25
24% Tier 1 share
Region
T1
T2
T3
T4
Total
T1 share
North America
6
10
8
1
25
24%
EMEA
4
11
12
9
36
11%
Asia Pacific
10
6
9
10
35
29%
Latin America
1
2
1
1
5
20%
All markets
21
29
30
21
101
21%
Published market dataset · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 26 · Results
Tier composition by region
stacked bars
Normalize each region's four final-tier counts to 100%.
Tier 1
Tier 2
Tier 3
Tier 4
Published market dataset · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 27 · Results
Composite score range by region
range plot
Calculate the minimum, median and maximum unrounded composite score within each published region.
North AmericaMedian 2.84
Vancouver sets the low at 1.14; Northern Virginia (incl. ext.) sets the high at 4.63.
Published market dataset · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 01 · Executive summary
Tier distribution by region, all 101 markets
stacked bars
Count final-tier records within each published region; bar lengths use a common absolute scale.
North America25 markets
Tier 1: 6 · Tier 2: 10 · Tier 3: 8 · Tier 4: 1
Tier 1
Tier 2
Tier 3
Tier 4
Published market dataset · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 02 · Executive summary
Composite score distribution
histogram
Count unrounded composite scores in half-point bins from 0.0 through 5.0 and mark the fixed tier thresholds.
T3 · 1.50T2 · 2.50T1 · 3.50
Selected half-point bin
3.5 - 4.0
Lower bound included · upper bound excluded
9 markets
Paris3.89
Melbourne3.83
Osaka3.81
Seoul3.75
Hong Kong3.66
Bangkok3.64
Johor3.61
Silicon Valley3.61
Milan3.60
Scores are binned before display rounding. Before the gate the distribution is 22 / 28 / 30 / 21; after Bangkok's demotion it is 21 / 29 / 30 / 21.
Published market dataset · Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 03 · Why a four-axis classification
Commercial colocation against hyperscale commitment
scatter
Plot each market's Axis A and Axis B scores, calculate the least-squares fit and Pearson correlation, and preserve the report's labelled-market selection.
Pearson correlationr = 0.77
Axis B · Hyperscale cloud & self-build
Axis A · Commercial colocation
North America
EMEA
Asia Pacific
Latin America
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 04 · Why a four-axis classification
The AI-created markets, and where the framework puts them
table
Use published market scores and tiers for the six report-selected markets, with the report's profile notes.
Selected market
West Texas (incl. Abilene)
Largest operational AI campus in the US; no colocation market, interconnection ecosystem or cloud region.
Axis C
4.60
Axis A
2.50
Composite
2.46
Final tier
3
Six AI-created markets
Tier
4.60
2.50
2.46
Tier 3
3.60
2.00
1.57
Tier 3
3.20
1.20
1.40
Tier 4
2.90
0.80
1.48
Tier 4
2.20
0.70
0.84
Tier 4
1.60
0.20
0.56
Tier 4
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 05 · Why a four-axis classification
The constraint is inverted
stat comparison
Present the published aggregate counts for the weakest and strongest C4 power-runway groups; underlying C4 observations are not in the public market file.
16 markets · weakest power position
11 of 1669%
are Tier 1 or Tier 2
Moratoria, refusal regimes and multi-year grid queues concentrate in the largest established markets.
Blue cells are Tier 1 or Tier 2 markets.
Power runway is metric C4, scored on evidenced grid policy and physics rather than developer statements.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 06 · Why a four-axis classification
Cloud region depth across the classified set
category distribution
Present the five published B2 cloud-region-depth category counts; underlying platform classifications are not in the public market file.
55carry at least one full availability-zone region
28carry no cloud region at all
Selected category
22
3+ full-AZ platforms
Twenty-two markets carry full availability-zone regions from at least three distinct platforms.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 07 · Why a four-axis classification
Scale without regions, and regions without scale
comparison cards
Use the report-selected markets and their published axis scores, with report-curated installed-capacity and cloud-region annotations.
Scale without regions
Regions without scale
Scale without regions
Atlanta
891MW installed · no cloud region a customer can buy
Tier 14.06composite
Axis A · Commercial colocation4.75
Axis B · Hyperscale and self-build3.85
Axis C · AI market maturity4.30
Axis D · Connectivity and network gravity2.40
Axis scores run from 0–5: A commercial colocation, B hyperscale cloud and self-build, C AI maturity, and D connectivity and network gravity.
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 10 · The framework
Why the A1 bands are logarithmic
log band scale
Place the six A1 score bands on the report's logarithmic 1MW-to-2,000MW display domain.
110501003007002,000
A1 score 3100 to <300MW operational colocation
The scale is logarithmic from 1MW to 2,000MW; score 5 remains open-ended above 700MW.
Published methodology · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 11 · The framework
Scoring bands for Axis A
score band table
Render the six score definitions for metrics A1 through A4.
A1 · 30% of Axis AScore 3: 100 to <300
Operational commercial colocation capacity in 2025, measured in megawatts.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 12 · The framework
The evidence ladder
rule cards
Present the report's full-credit, half-credit and zero-credit evidence rules.
Selected rule: Full credit — Capacity operating in the market today. An operating presence, energised and serving workloads.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 13 · The framework
The B2 ladder
ladder
Render the six published B2 cloud-region-depth score definitions from zero through five.
The top two bands require full availability-zone regions.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 14 · The framework
B1 operator weighting
table
Present the published operator groups and B1 point values, including the contracted-position rule.
Anchor clouds4 points operating · 2 contracted
AWS, Microsoft
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 15 · The framework
How B2 classifies each platform’s regions
table
Present the report's provider-by-provider rules for establishing cloud-region depth.
Amazon Web Services
Every commercial region carries three or more availability zones and counts as a full region. Local Zones and Wavelength Zones are edge presence.
Microsoft Azure
Microsoft publishes an availability-zone support flag per region rather than a zone count. Flagged regions count as full; the rest are regions without zone redundancy.
Google Cloud
Every launched region carries three or more zones and counts as a full region. CDN, Media CDN and interconnect locations are edge presence.
Oracle OCI
Graded on availability domains. Only five OCI regions worldwide carry three; Oracle states that it launches new geographies with a single domain.
Alibaba Cloud, Tencent Cloud
Graded on the zone count each publishes per region. Three or more counts as a full region; one or two does not.
Huawei Cloud
Graded on per-region launch releases and Huawei’s published global-infrastructure dataset. Four regions whose count is unpublished or internally contradicted are treated as not zone-redundant.
IBM Cloud
Multizone regions count as full. Single-campus multizone regions and classic data centres do not.
CoreWeave
The only AI cloud that documents availability zones as independent failure domains, but no region carries more than two and most carry one, so every region counts as a region without zone redundancy. Regions reserved for selected customers score zero.
Nebius, Crusoe, Lambda
None publishes a zone construct — the region is the smallest unit each of them exposes — so every region counts as a region without zone redundancy. Nebius private regions score zero.
Not counted
AI clouds that publish no region identifiers: Nscale, Firmus, Yotta, Sharon AI, Together AI and Fluidstack. Where there is no published region list, there is nothing a reader can verify, so the metric scores zero rather than relying on an analyst estimate.
10 of 10 platform rules shown.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 16 · The framework
Scoring bands for Axis B
score band table
Render the six score definitions for metrics B1 through B4.
B1 · 45% of Axis BScore 3: 8-12 pts
Quality-weighted physical capacity from hyperscale, cloud, AI-lab and neocloud operators.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 17 · The framework
The five classes counted in C3
comparison cards
Present the five published AI-demand classes; each class counts once when physically present.
C3 AI demand-diversity score0 / 5
Each distinct class counts once, regardless of how many members are present.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 18 · The framework
Scoring bands for Axis C
score band table
Render the six score definitions for metrics C1 through C4.
C1 · 30% of Axis CScore 3: 20-60MW
Operational data-centre capacity serving AI workloads.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 19 · The framework
Strongest AI axis scores in the classification
ranked bars
Use published Axis C scores and tiers while preserving the report's selected eight-market display.
West Texas4.60Tier 3
Abu Dhabi4.40Tier 2
Johor4.40Tier 1
Atlanta4.30Tier 1
Bangkok4.20Tier 2
Riyadh4.10Tier 2
Paris4.00Tier 1
Phoenix3.90Tier 1
Published selection reproduced exactly: it includes Phoenix at 3.90 and omits Northern Virginia at 4.40, so it is not a recalculated top-eight ranking. Use the market explorer to sort all 101 records on Axis C.
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 20 · The framework
The D1 distance rule
rule cards
Present the published distance and construction-status credits for international subsea systems.
Landing within 100 km1.0
Financed and under construction: 0.5 points.
100 – 250 km0.5
Financed and under construction: 0.25 points — the discounts compound.
Beyond 250 km0
Announced systems score zero at any distance.
Credited D1 points per system1.00
Distance credit and construction status multiply; announced systems always score zero.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 21 · The framework
Scoring bands for Axis D
score band table
Render the six score definitions for metrics D1 through D4.
D1 · 25% of Axis DScore 3: 6-9 pts
International subsea systems credited by landing distance from the metropolitan market.
Published methodology · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 22 · The framework
The gate in practice — Bangkok
threshold bars
Compare Bangkok's unrounded composite with the Tier 1 threshold and the report's 86MW operational-colocation value with the 240MW Tier 1 gate.
Composite score3.64
3.50 Tier 1 threshold
Operational colocation86MW
240MW Tier 1 floor
Result under the 2026 gatesTier 2
The composite reaches Tier 1, but at least one Tier 1 gate condition fails.
Published market dataset · Published methodology · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 23 · The framework
How the rule applies, axis by axis
rule matrix
Present the report's full-credit, half-credit and zero-credit tests for B1, B3, B4, C1, C2 and D1.
B1 Hyperscalers · Full creditCapacity operating in the market, leased or self-build
Construction status and capital at risk govern the score; announcements alone do not.
Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 29 · Regional analysis
The Gulf momentum cluster
comparison cards
Use published Axis C and Axis A scores for the report-selected Gulf markets, with report-curated installed-capacity and profile notes.
Abu Dhabi · Tier 2Axis C − Axis A: +1.70
126MW installed in a single-operator-dominated market; three cloud regions, none zone-redundant.
Axis C4.40
Axis A2.70
Data note: Dubai's published Axis C score is 0.45 points below Axis A, unlike the other three markets in this cluster.
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
Exhibit 24 · The framework
The rule at the bottom of the table
comparison cards
Use published composite scores for the report-selected five Tier 4 markets, with the report's evidence notes.
Tasmania · Tier 40.84 composite
A hydro position and a consented site; short on tenancy and construction.
A0.70
B0.65
C2.20
D0.20
Published market dataset · Report-curated content · August 2026.Interactive controls change the view; every value and rule is drawn from the published exhibit data.
