Within Local Power Strain
Why AI Data Centres Cannot Plug In Quickly
Large AI facilities can be delayed for years because grid operators must study, approve and build extensive local upgrades before connection.
On this page
- What happens inside a large load connection study
- Why transformers, substations and permits take years
- How queues reshape where AI facilities are built
Page outline Jump by section
Introduction
AI data centres are often built far more quickly than they can be connected to the electricity network. A modern AI facility may be physically complete within one to three years, yet still wait several more years before it can receive its full electricity supply. The reason is not usually a lack of generating capacity alone. Instead, grid operators must determine whether the local network can safely absorb a new electricity demand that may rival a medium-sized city, identify every piece of equipment that would become overloaded, assign responsibility for upgrades, and then build those upgrades before the connection is energised.[IEA]iea.orgGrids – Electricity 2026 – Analysis - IEA…
Within the wider discussion of AI electricity demand and local grid bottlenecks, this waiting period has become one of the main constraints on AI infrastructure. If advanced AI is expected to accelerate science, medicine and productivity over coming decades, then the speed at which transmission networks, substations and transformers can be expanded becomes part of the practical limit on how quickly those benefits can spread. The bottleneck is therefore not simply “more electricity”, but “getting large amounts of electricity to the right place safely and on time”.[IEA]iea.orgExecutive summary – Energy and AI – Analysis - IEA…
What happens inside a large-load connection study?
Connecting a very large AI data centre is unlike connecting an office building or factory. A hyperscale AI facility can require hundreds of megawatts of continuous power, with future campuses increasingly measured in gigawatts. Such demand changes how electricity flows across an entire regional network rather than just within a neighbourhood.[IEA]iea.orgAI and energy security – Energy and AI – AnalysisAI and energy security – Energy and AI – Analysis - IEA…
Before approving a connection, the transmission operator typically carries out a sequence of engineering studies designed to answer increasingly detailed questions.
Feasibility studies examine whether the proposed location appears technically viable and identify obvious constraints.
System impact studies simulate how the new demand would affect voltages, transmission lines, substations and system stability under both normal conditions and equipment failures. Engineers model scenarios such as the loss of a major transmission line or transformer to ensure the wider grid remains reliable.
Facilities studies determine the precise upgrades required, estimate costs and establish which party is responsible for funding new infrastructure.
Each stage depends on the previous one. If an earlier applicant changes its plans or withdraws from the queue, later studies may have to be repeated because the network assumptions have changed. This is one reason why connection queues can move slowly even when no physical construction has begun.[IEA]iea.orgAI and energy security – Energy and AI – AnalysisAI and energy security – Energy and AI – Analysis - IEA…
Why transformers, substations and permits take years
Once studies identify the required upgrades, the real waiting often begins.
A large AI connection may require:
- a new high-voltage substation;[techcrunch.com]techcrunch.comAI data centers just got a government-mandated fast lane to the grid | TechCrunchJune 18, 2026 — Image: High Voltage Power Transmission L…
- larger power transformers;
- additional transmission circuits;
- upgraded protection and control equipment;
- new switching stations and circuit breakers.
Many of these components cannot simply be ordered from stock. Large transformers are custom-built pieces of equipment weighing hundreds of tonnes. Manufacturing capacity is limited, supply chains are stretched and delivery times have lengthened substantially. At the same time, transmission projects must pass planning approvals, environmental assessments, land acquisition and construction before they can carry electricity.[IEA]iea.orgExecutive summary – Energy and AI – Analysis - IEA…
The timing mismatch is striking. The International Energy Agency notes that data centres are commonly built within one to three years, whereas planning, permitting and constructing major grid infrastructure often takes between five and fifteen years. Even where funding exists, the physical network simply cannot expand at the same pace as digital infrastructure.[IEA]iea.orgGrids – Electricity 2026 – Analysis - IEA…
Why connection queues keep growing
AI developers are joining an already crowded queue.
Grid operators are simultaneously processing requests from:
- renewable energy projects;[theguardian.com]theguardian.comWith 315 datacentres in the queue demanding 73GW—nearly 30GW above the country's peak demand—Ofgem aims to discourage speculative project…
- battery storage facilities;
- new manufacturing plants;
- transport electrification;
- hydrogen and industrial decarbonisation projects;
- large AI and cloud data centres.
Every major new connection changes assumptions about future electricity flows. As a result, applications often interact with one another rather than being assessed independently. If several large customers all choose the same region, the combined impact may require entirely different upgrades than any single project would have triggered alone.[IEA]iea.orgGrids – Electricity 2026 – Analysis - IEA…
Some jurisdictions have also experienced speculative applications, where developers reserve grid capacity before projects are fully financed. Regulators have increasingly introduced stricter milestones, deposits or queue reforms to ensure that inactive projects do not occupy scarce connection capacity indefinitely.[IEA]iea.orgAI and energy security – Energy and AI – AnalysisAI and energy security – Energy and AI – Analysis - IEA…
How long waits change where AI facilities are built
Electricity availability is becoming a site-selection criterion almost as important as fibre connectivity or land prices.
Historically, hyperscale data centres clustered around established digital hubs with excellent internet connectivity and existing cloud infrastructure. Those same locations are now frequently among the most constrained parts of the electricity network. The International Energy Agency estimates that around half of data centres under development in the United States are located within existing large clusters, increasing the risk of local bottlenecks.[IEA]iea.orgExecutive summary – Energy and AI – Analysis - IEA…
Developers increasingly respond in several ways:
- choosing regions where spare transmission capacity already exists;
- building campuses in phases as additional power becomes available;
- investing in on-site generation or battery systems to reduce dependence on immediate grid reinforcement;
- negotiating more flexible connection arrangements that allow limited operation before permanent upgrades are complete.
These decisions reshape the geography of AI infrastructure. Instead of concentrating solely around traditional technology hubs, new facilities increasingly seek locations where “time to power” is shorter, even if other advantages are slightly weaker.[IEA]iea.orgAI and energy security – Energy and AI – AnalysisAI and energy security – Energy and AI – Analysis - IEA…
Can the process become faster?
Governments and grid operators are experimenting with reforms that speed connections without compromising reliability.
One approach is to replace purely first-come, first-served queues with systems that prioritise projects that are genuinely ready to build. Another is to offer non-firm or conditional connections, allowing customers to connect earlier while accepting occasional restrictions during periods of network stress. The IEA estimates that regulatory reforms, grid-enhancing technologies and more flexible connection arrangements could unlock enough hosting capacity to connect hundreds of gigawatts of projects currently waiting in queues.[IEA]iea.orgGrids – Electricity 2026 – Analysis - IEA…
Researchers are also exploring whether some AI computing workloads can become more flexible. Training large models is often less time-critical than real-time AI services, creating opportunities to reduce consumption temporarily during periods of grid stress or shift certain workloads between geographically separated data centres. Although such flexibility cannot eliminate the need for new transmission infrastructure, it could reduce the scale of some upgrades and shorten connection delays in suitable cases.[arXiv]arxiv.orgarXiv Power-Flexible AI Data Centers: A New Paradigm for Grid-Responsive ComputePower-Flexible AI Data Centers: A New Paradigm for Grid-Responsive ComputeJune 23, 2026…
Why these delays matter for AI-enabled abundance
Connection queues may appear to be a narrow engineering issue, but they influence how quickly AI infrastructure can expand in practice.
The optimistic case for AI-enabled abundance assumes that increasingly capable systems will accelerate scientific research, industrial productivity and technological progress. Those capabilities depend on computing infrastructure, which in turn depends on reliable electricity. If transmission networks, substations and specialist equipment cannot be expanded at a comparable pace, computing capacity grows more slowly than demand.
That does not mean AI-driven progress is impossible. It means the limiting factor shifts from software to physical infrastructure. Building more generation helps, but it is only part of the solution. Equally important are faster permitting, expanded grid investment, improved connection procedures and better choices about where large AI facilities are located. The pace at which these practical bottlenecks are resolved will influence how rapidly the broader benefits of advanced AI can become widely available.[IEA]iea.orgAI and energy security – Energy and AI – AnalysisAI and energy security – Energy and AI – Analysis - IEA…
Amazon book picks
Further Reading
Books and field guides related to Why AI Data Centres Cannot Plug In Quickly. Use these as the next step if you want deeper reading beyond the article.
The Grid
One of Bill Gates's Favorite Books of 2016 A revelatory look at our national power grid--how it developed, its current flaws, and how it...
How the World Really Works
INSTANT NEW YORK TIMES BESTSELLER “A new masterpiece from one of my favorite authors… [How The World Really Works] is a compelling and hi...
The Big Switch
A hundred years ago, companies stopped producing their own power with steam engines and plugged into the newly built electric grid. The c...
Power Hungry
The promise of "green jobs" and a "clean energy future" has roused the masses. But as Robert Bryce makes clear in this provocative book,...
eBay marketplace picks
Marketplace Samples
Live-tested eBay searches with available results related to this page.
Selected fromserver rack model oneBay.co.uk.
Endnotes
1.
Source: iea.org
Link:https://www.iea.org/reports/electricity-2026/grids
Source snippet
Grids – Electricity 2026 – Analysis - IEA...
2.
Source: iea.org
Title: AI and energy security – Energy and AI – Analysis
Link:https://www.iea.org/reports/energy-and-ai/ai-and-energy-security
Source snippet
AI and energy security – Energy and AI – Analysis - IEA...
3.
Source: iea.org
Link:https://www.iea.org/reports/energy-and-ai/executive-summary%C2%A0
Source snippet
Executive summary – Energy and AI – Analysis - IEA...
4.
Source: iea.org
Link:https://www.iea.org/reports/electricity-2026/executive-summary
Source snippet
Executive summary – Electricity 2026 – Analysis - IEA...
5.
Source: arxiv.org
Title: arXiv Power-Flexible AI Data Centers: A New Paradigm for Grid-Responsive Compute
Link:https://arxiv.org/abs/2606.25098
Source snippet
Power-Flexible AI Data Centers: A New Paradigm for Grid-Responsive ComputeJune 23, 2026...
Published: June 23, 2026
6.
Source: arxiv.org
Link:https://arxiv.org/abs/2604.05376
7.
Source: iea.org
Title: Key Questions on Energy and AI – Analysis
Link:https://www.iea.org/reports/key-questions-on-energy-and-ai
8.
Source: iea.org
Link:https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions?__cf_chl_f_tk=mPlSI9KgDo0inioCRhuh6kU7U80qA.4CGq1WWVhvZcU-1783084545-1.0.1.1-072zjVfFsXXPnHR.edts3ccVzFC5OHDol3NSdHt3s8Y
9.
Source: iea.org
Link:https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary?_bhlid=10646f272364cf3af59c0fa8f3886b1cfe01e627
10.
Source: theguardian.com
Link:https://www.theguardian.com/uk-news/2026/jul/29/datacentre-projects-fees-grid-access-connection-queue-ofgem
Source snippet
With 315 datacentres in the queue demanding 73GW—nearly 30GW above the country's peak demand—Ofgem aims to discourage speculative project...
Additional References
11.
Source: ferc.gov
Link:https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration
Source snippet
FERC Launches Aggressive Targeted Action to Speed Large Load Integration | Federal Energy Regulatory CommissionJune 18, 2026 — FERC LAUNC...
Published: June 18, 2026
12.
Source: techcrunch.com
Link:https://techcrunch.com/2026/06/18/ai-data-centers-just-got-a-government-mandated-fast-lane-to-the-grid/
Source snippet
AI data centers just got a government-mandated fast lane to the grid | TechCrunchJune 18, 2026 — Image: High Voltage Power Transmission L...
Published: June 18, 2026
13.
Source: youtube.com
Link:https://www.youtube.com/watch?v=1o7WuKEbznw
Source snippet
140 UK Data Centres Are Queuing for Power That Doesn't Exist. Here's Where the Money Goes...
14.
Source: youtube.com
Link:https://www.youtube.com/watch?v=g27MDRzWnfw
Source snippet
The $12 Trillion Interconnect: Solving the AI Load Crisis...
15.
Source: youtube.com
Title: The Economics of Owning a Data Center
Link:https://www.youtube.com/watch?v=zwb1rXftbYs
Source snippet
The Data Center Power Bottleneck, Explained (And What Developers are Doing About it)...
16.
Source: carbon-direct.com
Link:https://www.carbon-direct.com/press/carbon-direct-releases-new-analysis-of-power-grid-interconnection-queues-pjm-ercot
17.
Source: sciencedirect.com
Title: Scalable data centers – Power generation and delivery challenges and solutions
Link:https://www.sciencedirect.com/science/article/pii/S2352484726001472
18.
Source: youtube.com
Link:https://www.youtube.com/watch?v=oued4ayuYPU
19.
Source: youtube.com
Title: The $12 Trillion Interconnect: Solving the AI Load Crisis
Link:https://www.youtube.com/watch?v=4JfZ8LlNIl8
Source snippet
The Economics of Owning a Data Center...
20.
Source: slideshare.net
Title: iea publication electricity year 2026 full report
Link:https://www.slideshare.net/slideshow/iea-publication-electricity-year-2026-full-report/285971204