Australia reports about two years
AEMO says current large data center applications target roughly two years to energization. The estimate is still conditional on project readiness and local system conditions.
Grid access benchmark
How long does it take to connect AI infrastructure?
Connection speed is becoming a national AI competitiveness variable. This first comparable estimate covers the same 45 countries as the AI Cost Index and also models a 250 kW on-prem solution.
Median 100 MW central estimate 5 years
Central estimates span from about 2 years to 9 years. Individual projects can fall outside the published ranges.
Estimated time to connect 100 MW
A commercially plausible location within each country. Firm grid capacity, including dedicated high-voltage assets where required.
Plausible range Central estimate
Australia
2 yr
United Arab Emirates
2 yr
China
2.5 yr
Saudi Arabia
3 yr
Finland
3.5 yr
Iceland
3.5 yr
India
3.5 yr
Estonia
4 yr
Latvia
4 yr
Lithuania
4 yr
Brazil
4.5 yr
Bulgaria
4.5 yr
Croatia
4.5 yr
Greece
4.5 yr
Mexico
4.5 yr
Norway
4.5 yr
Portugal
4.5 yr
Romania
4.5 yr
Slovenia
4.5 yr
South Korea
4.5 yr
Austria
5 yr
Canada
5 yr
Chile
5 yr
Cyprus
5 yr
Czechia
5 yr
Hungary
5 yr
Japan
5 yr
Poland
5 yr
Singapore
5 yr
Slovakia
5 yr
Spain
5 yr
Sweden
5 yr
Taiwan
5 yr
United States
5 yr
Italy
5.5 yr
Switzerland
5.5 yr
Denmark
6 yr
France
6 yr
Malta
6 yr
Belgium
7 yr
Germany
7 yr
Luxembourg
7 yr
Ireland
9 yr
Netherlands
9 yr
United Kingdom
9 yr
Version 2026-07. National screening estimates, not connection offers.
What the comparison says
Speed to power is not the same as abundant electricity. It measures whether planning, grid capacity, equipment and institutions can turn demand into an energized connection.
China and the Gulf can move faster
State-led planning, dedicated substations and coordinated generation can compress delivery. The ranges remain wide because no national measured median is published.
European hubs are the warning case
The IEA reports seven to ten year averages in Frankfurt, London, Amsterdam, Paris and Dublin. National siting flexibility helps, but it does not remove grid construction time.
250 kW is a distribution problem
The on-prem case normally needs a transformer and local network work, not a transmission project. Administrative quality and DSO capacity therefore matter more than national generation volume.
How the estimates work
There is no global database of completed 100 MW demand connections. The benchmark is a reproducible screening model built from the best available evidence.
Complete application
Site control, a credible electrical design and all information needed for the utility or system operator to begin formal assessment.
Firm energization
Contracted capacity is physically connected and available for commissioning. A connection offer alone does not stop the clock.
Commercially plausible site
The estimate is not tied to a capital city or the worst queue. A developer may choose a credible location with power, land and fiber inside the country.
Firm grid service
Onsite generation may support the project, but the estimate does not assume that temporary generation bypasses the requested grid connection.
Estimation sequence
- Apply one project definition to all countries.
- Anchor the range in published operator timelines, regulatory deadlines and current queue or congestion evidence.
- Add the likely time for studies, dedicated connection assets, upstream reinforcement and constrained equipment supply.
- For 250 kW only, use the discontinued World Bank standardized warehouse series as a historical process proxy, never as a current measured answer.
- Review the central estimate within the evidence range using Utiligize grid-domain judgment.
Evidence grades
- A
- Current national or system-operator timeline directly matching the case
- B
- Current national evidence strongly constrains the estimate, but no complete national median exists
- C
- Regional or adjacent evidence plus Utiligize engineering judgment
- D
- Sparse evidence; wide model-based range
Why the range matters more than the ranking
A national estimate averages over thousands of utilities, network locations and project designs. A site beside spare high-voltage capacity may connect years before one that triggers a new substation or transmission line. The central estimate is useful for country screening. The low and high values are the honest decision range.
Country evidence register
Every central estimate is Utiligize's judgment. The grade indicates how directly public evidence constrains it.
Core sources
Official operators, regulators and international institutions are preferred. Commercial anecdotes are not used as numerical anchors.
IEA, Energy and AI
Transmission lines can take four to eight years in advanced economies; waits for transformers and cables have doubled.
IEA, Electricity 2026: Grids
Grid infrastructure takes 5–15 years, compared with 1–3 years for data centers; more than 2,500 GW is stalled in connection queues.
IEA, Overcoming energy constraints in European data centers
Connections in the Frankfurt, London, Amsterdam, Paris and Dublin hubs average seven to ten years.
European Commission, guidance on efficient and timely grid connections
Documents months- or years-long waits and the mismatch between 4–10 year network construction and 2–3 year demand projects.
World Bank, time required to get electricity
Historical standardized warehouse connection process. Used only as a small-connection process anchor; Doing Business was discontinued in 2021.
Danish Utility Regulator, review of Energinet large-connection prioritization
Records Energinet's three-month pause on new large connections in March 2026 and the subsequent prioritization model.
Danish Energy Agency, grid connection process for large consumption
A comparable transmission-connected consumption process is expected to take about four years from screening agreement to operation.
Fingrid, temporary connection constraints in southern Finland
Large consumption projects in parts of southern Finland face restrictions in 2025–27, with reinforcements underway.
Svenska kraftnät, connection queue statistics
Average time from accepted application to preliminary decision was 17 months for 2018–25 applications, before construction and energization.
Nordic TSOs, large connection requests exceed available capacity
Fingrid, Energinet, Statnett and Svenska kraftnät report unprecedented requests exceeding short- and medium-term capacity.
ACM, connection deadlines for large electricity connections
Sets 26- or 52-week work periods, potentially longer for complex work, plus waiting time and congestion delays.
CRU, large energy user connection policy
New data centers require matching generation or storage, renewable additionality and a location-specific constraint assessment.
UK government, demand connection queue reform
Demand applications grew 460% in six months and some projects face waits of up to 15 years.
FERC, action to speed large-load integration
Shows fragmented regional rules and active reform across six US regional grid operators.
AESO, interim approach to large-load connections
Alberta limited near-term large-load connections to 1.2 GW against more than 16 GW of requests.
AEMO, data center connections in Australia
Reports an indicative two-year application-to-energization target for large data centers, subject to project and system conditions.
SASAC, Ulaanqab green-powered computing base
Documents coordinated delivery of dedicated substations, lines, generation and storage for a national computing hub.
NDRC and NEA, guidance on high-quality grid development
National policy calls for anticipatory, coordinated grid development and better access for new electricity uses.
Taipower, areas with available grid capacity
Publishes capacity-available areas but warns that high-voltage works and rights-of-way remain project-specific and uncertain.
AEMO, approximate two-year large data center connection timeframe
Current experience indicates an approximate two-year application-to-energization timeframe, with project-specific variation.
Coordinador Eléctrico Nacional, data center connection planning
Models growing data center load and the resulting transformer and transmission constraints in metropolitan Santiago.
Open data