An AI data center can be built surprisingly fast.
Its power connection may take much longer.
That sounds strange. If a region has power plants, why can’t a new data center simply plug in?
Because having electricity somewhere is not the same as having enough electricity at this site, at this time, through equipment that can safely carry it.
AI can move at software speed. The grid still moves at infrastructure speed.
That mismatch is becoming one of the most important physical constraints behind AI, electrification, new factories, batteries, and renewable power.
Is This Really a Power Shortage?
Sometimes. But often the immediate problem is grid access.
A region may produce enough electricity in total and still be unable to connect a new 100 MW, 500 MW, or larger load quickly.
The project can face three different limits:
- Generation: Is enough electricity available?
- Grid capacity: Can the network carry it to the site?
- Connection equipment: Are the substations, transformers, and switchgear ready?
These problems are related. They are not the same.
Why Can the Building Be Ready Before the Grid?
The main difference is project speed. The IEA compares typical lead times for several technologies and infrastructure projects.[1]
Figure 1. The Electric Age Is Moving Faster Than the Grid Typical lead-time ranges published by the IEA. Chart recreated by The Contexta from the source data.
A data center may be ready in a few years. A major new high-voltage line can take far longer.
The building can be finished before the electricity path is finished.
What Does an Interconnection Queue Actually Do?
It is not just a waiting line.
It is a process for answering engineering and financial questions before a large project connects.
Grid operators may need to ask:
- How will power flows change?
- Will existing lines overload?
- Is a new substation needed?
- Does a transformer need to be replaced?
- Will reliability get worse if another facility fails?
- Which upgrades are required?
- Who should pay for those upgrades?
Large projects can also affect one another. When one project leaves the queue, earlier studies may need to be updated.
The IEA reports that more than 2,500 GW of renewables, storage, and large-load projects are stalled in grid-connection queues worldwide.[1]
That number is not a forecast. Many projects will never be built.
It is a sign of how much pressure is building around the connection process.
Is the U.S. Queue Getting Better?
There are some signs of improvement, but the backlog is still enormous.
Berkeley Lab’s 2026 update, using data through the end of 2025, found about 8,200 projects actively seeking U.S. grid connection.[2]
They represented roughly:
- 1,312 GW of generation
- 749 GW of storage
That is about 2,061 GW in total. Active queue volume fell from the year before, partly because many projects withdrew.
But projects that do get built are still taking longer on average to move through studies and reach operation.
Are Transformers Really a Bottleneck?
Yes. But the shortage is broader than AI alone.
Data centers are adding a new wave of demand for transformers, circuit breakers, and switchgear.
Utilities also need the same equipment for:
- renewable generation
- industrial electrification
- electric vehicles
- grid expansion
- replacement of aging equipment
- storm and disaster recovery
The IEA says it can take two to three years to procure some transmission cables and up to four years for large power transformers. Some specialised direct-current cables can take more than five years.[3]
Reuters reported in July 2026 that lead times for some high-voltage transformers in the United States had reached about 160 weeks.[4]
IEA industry data also show cable prices nearly doubled from 2019 to 2024, while power-transformer prices rose by about 75%.[3]
A grid is not built with money alone. It also needs factories, materials, skilled workers, and time.
Is Permitting the Main Problem?
It is important, but it is not the only problem.
New transmission can cross farms, towns, forests, rivers, and several legal jurisdictions.
Projects may need:
- route studies
- land rights
- environmental review
- local and national permits
- engineering studies
- equipment procurement
- construction crews
- agreement on cost allocation
The IEA identifies permitting as a major cause of transmission delay in many advanced economies.
But faster permits do not instantly create transformers, cables, rights-of-way, engineers, or financing.
Do We Really Have to Wait for New Power Lines?
Not always.
Some capacity can be unlocked from infrastructure that already exists.
| Grid option | Typical lead time | What it does |
|---|---|---|
| Dynamic line rating | 1–2 years | Uses weather and line conditions to estimate safe real-time capacity |
| Topology optimisation | 1–2 years | Changes network configuration to reduce congestion |
| Advanced power-flow control | 2–3 years | Directs electricity toward less-congested paths |
| Reconductoring | 3–4 years | Replaces wires with higher-capacity conductors on existing routes |
| Voltage uprating | 4–7 years | Raises the voltage of parts of the existing network |
| New high-voltage line | 7+ years | Adds a completely new transmission path |
These ranges come from the IEA’s 2026 grid analysis.[1]
The point is not that one technology solves everything.
Different grid problems run on different clocks.
Can Batteries Replace Transmission Lines?
No. They solve a different problem.
Batteries move electricity through time.
Transmission lines move electricity through space.
Batteries can save power for later. They cannot move a distant power plant closer to a city.
Storage can reduce congestion, smooth peaks, and sometimes delay a network upgrade.
But if a region needs a much larger path to move electricity over distance, transmission still matters.
Who Pays for the Grid Upgrade?
This is becoming one of the hardest questions.
If a new data center requires a substation, transformer, or major transmission upgrade, someone must pay.
Possible answers include:
- the data-center developer
- the utility
- other electricity customers
- taxpayers
- a negotiated combination
In June 2026, the U.S. Federal Energy Regulatory Commission ordered all six regional grid operators under its jurisdiction to justify or reform the rules for connecting data centers and other large loads.[5]
FERC emphasized both speed-to-power and protection for existing ratepayers.
The grid bottleneck is an engineering problem and a money problem at the same time.
What Does This Mean for AI and Industry?
An AI project needs more than chips and capital.
It also needs:
- a site
- power generation or contracts
- a grid connection
- transformers and switchgear
- cooling
- construction labor
- network capacity
A missing part can delay the whole system.
Reuters reported in July 2026 that U.S. utilities and developers were ordering critical grid equipment years in advance as data-center demand tightened supply.[4]
Speed-to-compute now depends partly on speed-to-power.
What Should You Watch Next?
- Connection time: How long does a large load wait between application and commercial power?
- Transformer lead time: Are delivery times falling as factories add capacity?
- Queue reform: Are serious projects moving faster and speculative projects leaving sooner?
- Grid-enhancing technology: Are utilities using dynamic ratings, power-flow controls, and reconductoring?
- Cost allocation: Who pays for upgrades required by new data centers and other large users?
How Should an Ordinary Reader Prepare?
When you see a large data-center announcement, do not stop at the investment number.
Look for five details:
- requested MW or GW
- the utility or grid operator
- expected connection date
- required substation, transformer, or transmission upgrades
- who pays for those upgrades
A project with land and chips is not the same as a project with secured power.
The Main Idea
The grid is not one bottleneck.
It is a chain of bottlenecks running on different clocks.
Some can be improved in one or two years. Others need factories, permits, land, equipment, and construction that can take most of a decade.
The next AI race is not only about who can buy the most chips. It is also about who can connect power fast enough.
Continue the Electric Age Series
- Solar and Batteries Got Much Cheaper. Why Isn’t Electricity Always Cheap? — See why cheap components do not automatically create a cheap electricity bill.
- From Electricity to Intelligence: How the Grid Becomes a Data System — Continue to the digital layer of the electric system.
- Why Power, Not Chips, May Limit the AI Data Center Boom — Apply the grid bottleneck directly to AI infrastructure.
- The Electric Age Explained: From Falling Energy Costs to AI and Global Power — Read the complete Electric Age guide.
Key English Words
- grid access: the ability to connect a project to the electricity network
- interconnection queue: the process and waiting list for projects seeking grid connection
- transformer: equipment that changes voltage so electricity can move and be used safely
- lead time: the time between ordering or starting something and receiving or completing it
- reconductoring: replacing existing wires with higher-capacity conductors, often on existing towers
- cost allocation: the rules that decide who pays for required grid upgrades
Sources
- IEA — Electricity 2026: Grids — global queues, lead times, and grid-enhancing technologies.
- Lawrence Berkeley National Laboratory — Queued Up 2026 Edition — U.S. generation and storage interconnection data through the end of 2025.
- IEA — Building the Future Transmission Grid — cable and transformer prices, procurement times, and permitting constraints.
- Reuters — U.S. power companies scramble to secure equipment — 2026 evidence on transformer, switchgear, and circuit-breaker lead times.
- FERC — Targeted action to speed large-load integration — June 2026 action on data-center and other large-load connection rules.
- U.S. Department of Energy — Distribution Transformers — supply-chain constraints and industry efforts to reduce production delays.
Figure 1 was retained from the original article and recreated by The Contexta from IEA lead-time ranges. Queue totals are indicators of connection pressure, not forecasts that every project will be built. Sources checked September 2026.