The Race for More Electricity: Can Power Infrastructure Keep Up With AI, Data Centers, EVs and Electrification?
Engr. Muhammad Faisal Abbas
BSc Electrical Engineering
Table of Contents
- The New Electricity Race
- The Demand Shock Is Bigger Than AI
- AI and Data Centers: Small Globally, Huge Locally
- The Real Bottleneck: Time
- Generation Is Not the Same as Deliverable Power
- Transmission and Distribution: The Forgotten Infrastructure
- Transformers, Cables and the Physical Supply Chain
- EVs, Heat Pumps, Cooling and Industrial Electrification
- The Geography Problem: Power Where It Is Needed
- Can Renewables, Nuclear, Gas and Storage Keep Up?
- The Flexibility Challenge: When and How Much Power Is Needed
- Who Pays for the New Grid?
- The Risk of Building Too Much—and the Risk of Building Too Little
- What a Grid Designed for the Digital Economy Must Look Like
- The Critical Question: Can Physical Infrastructure Outrun Digital Growth?
- Conclusion: The AI Revolution May Ultimately Be a Grid Revolution
- References
The New Electricity Race
For much of the developed world, electricity demand was almost boring.
For years, efficiency improvements, industrial restructuring and the movement of economies away from heavy manufacturing kept electricity consumption relatively flat in many advanced economies. Utilities could plan around relatively predictable growth—or, in some cases, almost no growth at all.
That era is ending. The world is entering what the International Energy Agency (IEA) calls the “Age of Electricity.” Global electricity demand is forecast to grow by an average 3.6% per year from 2026 to 2030, around 50% faster than the average annual growth rate of the previous decade. The drivers are not just one technology: industry, electric vehicles, air conditioning, heat pumps, data centers and broader electrification are all pulling on the system simultaneously.
That distinction matters. The popular narrative is that artificial intelligence is suddenly consuming enormous quantities of electricity.
That is true—but incomplete. The deeper story is that several electricity-intensive transformations are arriving at the same time. Cars are becoming electric. Heating is becoming electric. Industrial processes are becoming electric. Cooling demand is increasing. Manufacturing is becoming more electricity-intensive. And computing is becoming dramatically more powerful. The result is a structural change in the relationship between economic growth and electricity.
The critical question is therefore not: How much electricity will AI consume?
It is: Can the physical electricity system expand quickly enough to support an economy that is becoming increasingly dependent on electricity?
The Demand Shock Is Bigger Than AI
AI receives most of the attention because its growth is spectacular. But electricity demand does not care which sector receives the headlines. A transformer sees load. A transmission line sees current. A substation sees power flow. A generating plant sees demand. And all of these have to work simultaneously.
The IEA’s latest electricity outlook shows how broad the demand increase is. Buildings, industry and transport are all contributing significantly, with cooling, heat pumps, EVs and data centres becoming increasingly important. Between 2025 and 2030, buildings alone are expected to contribute about 49% of additional global electricity demand. Transport’s contribution is expected to exceed 10%, more than double its share of growth during the previous five-year period.
The following simplified picture captures the fundamental problem:
| Electricity demand driver | What is changing? | Infrastructure consequence |
|---|---|---|
| AI/data centres | Rapid growth in computing | Large, concentrated loads |
| EVs | Millions of vehicles charging | Distribution and peak-load pressure |
| Heat pumps | Electrification of heating | Seasonal demand shifts |
| Air conditioning | Rising cooling demand | Extreme-weather peak loads |
| Industry | Electrification and reshoring | Large industrial connections |
| Manufacturing | Batteries, semiconductors, advanced factories | High-quality, reliable power |
| Digital services | More cloud and network activity | Continuous electricity demand |
The challenge is therefore cumulative. A grid that can comfortably handle one of these trends may struggle when all of them arrive together.
AI and Data Centers: Small Globally, Huge Locally
There is an important correction to the phrase “AI is consuming the world’s electricity.” It isn’t—at least not yet. The global electricity share of data centers remains relatively modest.
The IEA estimates that data centres consumed about 1.5% of worldwide electricity demand in 2025 and projects this to rise to approximately 3% by 2030, with global data-centre electricity consumption roughly doubling to around 950 TWh.
That is substantial, but it does not mean that data centres will overwhelm the global electricity system. The real problem is concentration. A data center is not like a million households each using a little more electricity. It can represent an enormous load concentrated at one location. And AI is pushing that concentration further.
The IEA’s analysis estimates that accelerated servers—primarily associated with AI—could grow electricity consumption at around 30% annually in its base case, substantially faster than conventional servers.
This produces an unusual infrastructure problem. A country can have sufficient generation capacity overall and still be unable to connect a particular data center. Why? Because national electricity capacity is not the same thing as local grid capacity.
The Real Bottleneck: Time
This is arguably the central issue in the entire debate. Digital infrastructure moves at digital speed. Electrical infrastructure does not. A technology company can design, finance and construct a data centre on a timescale measured in a few years. The grid operates according to a different clock.
The IEA estimates that planning, permitting and completing new grid infrastructure can take approximately 5–15 years, while data centres can be built in roughly 1–3 years. EV charging infrastructure may take only 1–2 years, and many renewable projects can be developed within 1–5 years. That creates a dangerous sequencing problem.
The infrastructure timing mismatch
DIGITAL / DEMAND SIDE
AI investment
│
▼
Data-centre planning
│
▼
Construction
│
▼
LOAD ARRIVES
│
│
│
▼
PHYSICAL GRID SIDE
Need identified
│
▼
Planning studies
│
▼
Permitting
│
▼
Land / rights-of-way
│
▼
Equipment procurement
│
▼
Construction
│
▼
Commissioning
│
▼
GRID CAPACITY ARRIVESThe two clocks are fundamentally misaligned. This is why the problem cannot be solved simply by saying: “Build more power plants.” The electricity has to arrive at the right location, at the right voltage, at the right time, with sufficient reliability.
Generation Is Not the Same as Deliverable Power
One of the most persistent misconceptions in electricity discussions is treating generation capacity as though it automatically translates into usable electricity everywhere. It doesn’t. Imagine a region with abundant solar generation. On paper, it has plenty of electricity. But suppose a new data centre requires 1 GW continuously, while the available transmission corridor is already operating close to its limits.
The problem isn’t necessarily generation. It is deliverability. This can be expressed conceptually as:
Usable power ≠ installed generation capacity
More realistically:
Deliverable power = generation × transmission availability × distribution capacity × system reliability
The actual engineering problem is even more complicated because voltage constraints, thermal limits, stability, contingencies, congestion, reserve requirements and timing all matter.
This is why the IEA reports that more than 2,500 GW of renewable, storage and large-load projects—including data-centre-type loads—are currently stalled in grid connection queues worldwide.
The queue itself is therefore becoming a form of infrastructure. Projects may exist financially and technologically, yet remain physically disconnected.
Transmission and Distribution: The Forgotten Infrastructure
For decades, electricity investment discussions often focused on generation. How many gigawatts of solar? How many gigawatts of wind? How much nuclear? How much gas? But the generation fleet is only one layer. The electricity system is a chain:
Generator → transmission → substation → distribution → customer
If one link is inadequate, the entire chain is constrained. The IEA estimates that around USD 400 billion per year is currently invested in electricity grids globally and says annual grid investment needs to rise by roughly 50% by 2030 to meet forecast demand.
The imbalance is striking. The IEA’s 2025 investment analysis puts annual global grid spending at roughly USD 400 billion compared with around USD 1 trillion for generation assets.
Simplified global investment imbalance
| Infrastructure | Approximate annual investment |
|---|---|
| Electricity grids | ~$400 billion |
| Generation assets | ~$1 trillion |
| Grid investment needed by 2030 | ~$600 billion/year or more |
The lesson is uncomfortable: Building generation faster while neglecting networks can actually increase the amount of generation that cannot be used. That is already happening in some markets through congestion and renewable curtailment.
Transformers, Cables and the Physical Supply Chain
Even if governments suddenly agreed to build more grids, another problem remains: The equipment has to exist.
Power systems depend on enormous quantities of highly specialized equipment:
- power transformers
- distribution transformers
- circuit breakers
- switchgear
- conductors
- underground cables
- overhead transmission lines
- substations
- power electronics
- protection systems
- control equipment
These are not software components that can be duplicated overnight.
The U.S. Department of Energy reports that critical grid equipment is experiencing supply-chain constraints, with some equipment facing lead times of two years or more. Distribution-transformer lead times increased from roughly 3–6 months in 2019 to 12–30 months in 2023.
This exposes a crucial vulnerability. The AI economy depends on semiconductor supply chains. But the AI economy also increasingly depends on transformer supply chains. A shortage of GPUs can delay a data centre. A shortage of transformers can delay the electrical infrastructure needed to power it. The bottleneck simply moves downstream.
EVs, Heat Pumps, Cooling and Industrial Electrification
It would be a mistake to treat data centres as the only new load. EVs alone are becoming a major electricity consumer. Global EV electricity consumption reached approximately 180 TWh in 2024, almost 60% higher than the previous year. Under the IEA’s stated-policies scenario, EV electricity demand could reach about 780 TWh by 2030. That is more than four times the 2024 level. But EVs create a fundamentally different grid challenge from data centres.
A data centre may operate continuously. EV charging can be shifted. That difference is extremely valuable. If millions of vehicles all charge immediately after people return home, the grid experiences a sharp evening peak. If charging is intelligently managed, much of that demand can be moved to periods when electricity is cheaper and network capacity is available. The same principle applies to heat pumps, industrial loads and some forms of refrigeration.
This leads to a critical concept:
The future grid does not merely need more electricity. It needs more controllable electricity demand.
The Geography Problem: Power Where It Is Needed
Electricity is not economically interchangeable simply because it exists somewhere on a national map. Location matters.
A wind farm may be located hundreds of kilometres from a major city. A solar plant may be built where land and sunlight are abundant. A hydroelectric plant may be located far from industrial centres.
A data centre, however, may want to locate near:
- fibre networks
- customers
- low-latency infrastructure
- land
- tax incentives
- reliable power
- cooling resources
- existing industrial infrastructure
These priorities do not necessarily overlap. This creates what might be called the geographical contradiction of electrification: The places with the cheapest or cleanest electricity are not always the places where the fastest-growing electricity loads want to locate.
China illustrates this issue particularly well. The IEA notes that data centres are heavily concentrated in eastern China, while policies are encouraging development in renewable-rich western regions. The same basic tension exists elsewhere.
Can Renewables, Nuclear, Gas and Storage Keep Up?
There is no single technology capable of solving the entire problem. And that is important. The electricity debate often degenerates into arguments over whether the answer is: solar, wind, nuclear, gas or batteries. The actual engineering answer is more complicated.
The system needs a portfolio. The IEA expects renewables to supply nearly half of the additional electricity demand from data centres through 2030, while natural gas and nuclear also play significant roles.
For data centres specifically, the IEA projects global electricity generation serving them to rise from around 460 TWh in 2024 to more than 1,000 TWh in 2030.
Different technologies solve different problems
| Technology | Major strength | Major limitation |
|---|---|---|
| Solar | Fast deployment, low-cost energy | Variable output |
| Wind | Large-scale generation | Variable output and transmission needs |
| Nuclear | Firm low-carbon power | Long development timelines |
| Natural gas | Dispatchable, flexible | Fuel cost and emissions |
| Batteries | Fast response, peak shifting | Duration and material constraints |
| Hydropower | Flexible generation | Geographic limitations |
| Demand response | Avoids or shifts load | Requires controllable demand |
| Transmission | Moves electricity geographically | Slow permitting/buildout |
The future system therefore cannot be designed around energy production alone. It must be designed around system flexibility.
The Flexibility Challenge: When and How Much Power Is Needed
A megawatt is not simply a megawatt. Timing matters. Location matters. Duration matters. Reliability matters.
Consider two customers:
Customer A:
1 GW for two hours during a predictable evening peak.
Customer B:
1 GW continuously, 24 hours a day, seven days a week.
They may have the same peak demand. But their effects on the grid are radically different. This is why future planning must move beyond annual energy consumption.
A useful conceptual framework is:
Grid requirement = magnitude × timing × duration × location × reliability
AI introduces another complication. AI workloads can involve highly concentrated computing loads and potentially significant changes in demand patterns as computational tasks are scheduled, scaled and migrated.
The IEA notes that newer electricity demands—including EVs, heat pumps and large loads such as data centres—are spatially and temporally concentrated, increasing the need for system flexibility. That makes advanced demand management increasingly important.
Who Pays for the New Grid?
This is where the technical discussion becomes political and economic. Suppose a hyperscale data centre requires billions of dollars of new transmission and substation infrastructure.
Who pays? The technology company? The utility? Existing electricity customers? Taxpayers? The government? Or some combination?
There is no universally correct answer. But there is a serious risk in socialising the cost while privatising the benefit. If one industrial customer triggers a major grid upgrade, regulators must determine whether the investment benefits the wider electricity system or primarily serves that customer.
The opposite problem is also possible. If utilities refuse to invest until every future load is contractually guaranteed, they may build too late. This creates a classic infrastructure dilemma:
Build ahead of demand → risk stranded investment.
Wait for demand certainty → risk infrastructure shortages.
The answer requires better planning rather than simply shifting the financial burden.
The Risk of Building Too Much—and the Risk of Building Too Little
The electricity race has two symmetrical dangers.
Risk 1: Underbuilding
If infrastructure cannot keep up:
- data-centre projects are delayed
- industrial investment moves elsewhere
- EV charging becomes constrained
- electricity prices can rise
- renewable generation is curtailed
- reliability margins shrink
- congestion increases
- fossil generation may be retained longer than planned
The economic consequence could be significant.
Electricity becomes a constraint on growth rather than an enabler of growth.
Risk 2: Overbuilding
But blindly building everything requested would also be irresponsible. AI projections contain considerable uncertainty. Computing efficiency can improve. Algorithms can become less energy-intensive. Workloads can migrate. Data-centre construction plans can be cancelled.
AI investment itself depends on whether the technology generates sufficient economic returns.The IEA’s updated analysis explicitly highlights uncertainty around the trajectory of AI adoption, efficiency improvements, investment and energy-sector bottlenecks.
Therefore, the answer is not simply: “Forecast huge demand and build huge infrastructure.”
It is: Build infrastructure that can adapt if the forecast changes.
That means modularity, flexibility, stronger interconnections, better forecasting and infrastructure that can serve multiple future uses.
What a Grid Designed for the Digital Economy Must Look Like
The traditional grid was largely designed around a relatively predictable relationship:
Generation → transmission → distribution → passive customer
That model is becoming obsolete.
The future grid will look more like a dynamic network containing:
- renewable generators
- nuclear plants
- gas plants
- batteries
- EVs
- smart chargers
- heat pumps
- distributed solar
- industrial loads
- data centres
- flexible loads
- microgrids
- storage
- advanced power electronics
The customer is no longer simply consuming electricity. The customer can increasingly respond to the grid. That changes the engineering philosophy.
From:
Build enough capacity for the maximum possible demand.
Toward:
Build sufficient physical capacity while actively managing when, where and how demand occurs.
This does not eliminate the need for infrastructure. It makes infrastructure smarter.
The Digital Economy May Need a Physical-Energy Operating System
There is a deeper implication here. The internet revolution made information abundant and cheap. AI is now making computation abundant and increasingly accessible. But computation is not weightless. Every AI model ultimately depends on:
chips → servers → cooling → buildings → electricity → generation → networks
The digital economy therefore rests on an enormous physical infrastructure stack. And the weakest layer can determine the speed of the whole system. This produces an unusual inversion.
For decades, electricity infrastructure was often treated as background infrastructure for the digital economy. Increasingly, electricity infrastructure may become one of the primary constraints on digital economic expansion.
The IEA’s analysis makes the contradiction particularly clear: a data centre can be operational in two to three years, while the broader energy system requires substantially longer planning and construction periods. The implication is profound.
The next competitive advantage may not simply be: Who has the best AI model?
It may increasingly be: Who can secure reliable, affordable and scalable electricity fastest?
The Infrastructure Race Is Already Visible
The United States provides an especially clear example.
After nearly two decades of relatively flat electricity demand, the country’s grid is experiencing rapid load growth. The U.S. Department of Energy says electricity demand has been increasing at close to 3% annually since 2023, driven partly by data centres, manufacturing and broader economic growth.
Meanwhile, DOE’s 2026 draft National Transmission Needs Study identifies pressing transmission needs associated with data centres, domestic manufacturing, large industrial loads and electrification.
The scale of the challenge can also be seen in U.S. generation interconnection queues.
Berkeley Lab’s 2026 edition reports roughly 8,200 projects seeking transmission interconnection at the end of 2025, representing approximately 1,312 GW of generation and 749 GW of storage. Importantly, these figures concern generation/storage requests and do not represent a direct measure of data-centre load queues. That distinction matters. A queue full of projects is not the same thing as installed infrastructure.
The Grid May Become the New Strategic Infrastructure
The consequences extend well beyond utilities.
Electricity infrastructure is increasingly becoming an issue of:
- industrial policy
- national security
- economic competitiveness
- climate policy
- technological leadership
- energy independence
The IEA estimates that global electricity-sector investment is set to reach about USD 1.5 trillion in 2025, approximately 50% more than investment in bringing oil, natural gas and coal to market. That is an extraordinary structural shift.
The energy economy is increasingly moving from a world dominated by fuel extraction toward one dominated by electricity production, networks, storage and electrified end uses. And grids are the connective tissue.
What Should Governments and Utilities Actually Do?
There is no single silver bullet. But several principles are increasingly difficult to ignore.
1. Plan grids before demand arrives
Waiting for confirmed demand can be fatal when grid infrastructure takes a decade to develop. Long-term planning must anticipate industrial clusters, data centres, EV adoption, electrified heating and new manufacturing.
2. Treat transmission as strategic infrastructure
Transmission is not simply an optional connection between generators. It is the mechanism that allows regions with different generation and demand profiles to support one another.
3. Modernise distribution networks
The distribution system is likely to experience enormous pressure from EVs, heat pumps, distributed generation and electrified buildings. The future grid problem is therefore not exclusively a high-voltage transmission problem.
4. Reform interconnection
Thousands of projects waiting in queues indicate that the connection process itself can become a bottleneck. Planning needs to move from project-by-project reaction toward anticipatory network development.
5. Expand grid supply chains
Transformers, cables, switchgear and other components need manufacturing capacity—not just theoretical procurement plans.
6. Reward flexibility
Customers who can shift electricity demand should have economic incentives to do so.
7. Make large loads responsible participants
Large data centres and industrial customers should not simply appear on the grid and expect unlimited capacity.
They can potentially contribute through:
- flexible demand
- onsite generation
- storage
- power-quality management
- long-term contracts
- transmission contributions
- demand-response arrangements
8. Diversify generation
A resilient electricity system should not depend on one technology. The objective is not ideological purity. It is reliability at acceptable cost and environmental impact.
The Most Important Metric May No Longer Be “How Many Gigawatts?”
For decades, energy planning often revolved around installed capacity. But the future requires more sophisticated metrics. We should increasingly ask:
How much firm capacity is available?
Where is it available?
When is it available?
For how long?
What transmission capacity connects it to demand?
How quickly can the system respond to unexpected changes?
How much demand can be shifted?
What happens under extreme weather or equipment failure?
This represents a transition from capacity planning to system capability planning. A grid can have enormous installed generation and still be fragile.
Conversely, a highly interconnected system with storage, flexible demand and diversified generation may provide more usable capacity from fewer nominal assets.
The Killer Question: What Happens When Digital Growth Outruns Physical Infrastructure?
This is the question that deserves much more attention. Suppose AI investment accelerates faster than expected. Data centres are built. EV adoption accelerates. Industrial electrification expands. Cooling demand rises.
But transmission projects remain trapped in permitting. Transformers remain constrained. Substations take years to build. Generation is constructed in regions where demand is not growing. What happens?
The digital economy encounters a physical ceiling. Developers begin competing for grid capacity. Electricity prices may become increasingly location-dependent. Data centres migrate toward power-rich regions. Utilities prioritise some loads over others. Governments intervene. Temporary gas generation may appear. Industrial projects may be delayed. Renewable projects may be curtailed. And electricity itself becomes a competitive economic resource.
This is not science fiction. The IEA is already identifying grid connection bottlenecks, long development timelines and supply-chain constraints as material obstacles to meeting rapidly growing electricity demand.
The Paradox of AI: It Could Help Solve the Problem It Helps Create
There is an interesting possibility hidden inside this crisis. AI is increasing electricity demand. But AI can also improve the operation of the electricity system.
It can potentially help with:
- demand forecasting
- renewable forecasting
- predictive maintenance
- outage detection
- transmission optimisation
- grid congestion management
- battery dispatch
- building energy management
- industrial load optimisation
- asset inspection
- power-quality monitoring
In other words: AI may become both a major new electricity consumer and an important tool for making electricity infrastructure more efficient.
The IEA identifies AI-enabled optimisation and innovation as an important potential benefit for the energy sector. But this should not become an excuse for magical thinking. AI cannot eliminate a missing transmission line. It cannot manufacture a transformer instantly. It cannot overcome a physical thermal limit indefinitely. Software can extract more capability from infrastructure. It cannot make physics disappear.
A More Realistic Future: The Grid as a Flexible Platform
The winning electricity systems of the next decade are unlikely to be those that simply build the most generation.
They will be the systems that combine generation + transmission + distribution + storage + flexibility + digital control + demand management into one coordinated architecture.
The IEA’s 2026 electricity analysis explicitly places grid expansion and system flexibility at the centre of the new electricity era. This changes how infrastructure should be evaluated. A transmission project should not be viewed only as a cost. A battery should not be viewed only as an energy-storage device. An EV charger should not be viewed only as transport infrastructure. A data centre should not be viewed only as a technology investment. They are increasingly components of a single electricity ecosystem.
The Race Is Not Really for More Electricity
This is the central conclusion. The world does need more electricity. But more electricity alone is not enough.
What is needed is:
more generation, in the right locations;
more transmission, before congestion becomes critical;
more distribution capacity, before electrification overwhelms local networks;
more transformers and equipment;
more storage;
more flexible demand;
better forecasting;
faster permitting;
better interconnection planning;
and investment mechanisms capable of financing infrastructure years before demand becomes obvious.
The IEA’s estimate that grid investment needs to rise substantially by 2030 is therefore not merely an infrastructure statistic. It is a warning about the physical foundation of the emerging economy.
Conclusion: The AI Revolution May Ultimately Be a Grid Revolution
The most important electricity story of the coming decade may not be the number of AI models being trained. It may be the number of substations being built. It may not be the next generation of GPUs. It may be the availability of transformers. It may not be how quickly a company can build a data centre. It may be how quickly the surrounding electrical infrastructure can connect it. That is the uncomfortable reality behind the AI electricity boom.
Digital infrastructure can scale extraordinarily quickly. Physical infrastructure cannot. A server can be manufactured in a factory. A software model can be deployed globally in hours. A data centre can be constructed in a few years. But a major transmission corridor can require a decade or more of planning, permitting, procurement and construction.
That mismatch creates the central infrastructure challenge of the AI era: What happens when the digital economy grows faster than the electrical system underneath it?
The answer will determine much more than whether data centres receive enough electricity.
It will influence where industries locate, how quickly EVs are adopted, whether electrification delivers its promised benefits, how much renewable energy can actually be used, how electricity prices evolve, and which countries are able to capture the economic value of AI.
The race for electricity is therefore not really a race for electrons. It is a race to build the physical system capable of delivering those electrons reliably, affordably and at the exact moment and location where the new economy needs them. And that race has already begun.
Bottom line: The electricity challenge of AI is not primarily a question of whether humanity can generate enough electrons. It is whether generation, transmission, distribution, equipment supply chains, financing, regulation and demand flexibility can scale on a timetable compatible with the speed of digital and physical electrification. That is where the real race is being won—or lost.
Key Takeaways
| Indicator | Current / projected figure | What it tells us |
|---|---|---|
| Global electricity-demand growth, 2026–2030 | 3.6%/yr | Demand growth is accelerating |
| Data-centre electricity demand, 2025 | ~485 TWh | Already a major concentrated load |
| Data-centre electricity demand, 2030 | ~950 TWh | Roughly doubles in five years |
| Data-centre share of global electricity, 2030 | ~3% | Globally manageable, locally disruptive |
| Global EV electricity use, 2024 | ~180 TWh | Rapidly growing new load |
| EV electricity demand, 2030 STEPS | ~780 TWh | Transport electrification becomes significant |
| Global annual grid investment today | ~$400 billion | Current infrastructure spending |
| Additional grid investment needed by 2030 | ~50% higher | Networks need to catch up |
| Projects stalled in global grid queues | >2,500 GW | Connection capacity is a major bottleneck |
| U.S. generation/storage projects in queues, end-2025 | ~1,312 GW generation + ~749 GW storage | Large development pipeline does not equal connected capacity |
A Simple Systems View
THE ELECTRICITY RACE
AI / DATA CENTRES ───────┐
│
ELECTRIC VEHICLES ───────┤
│
HEAT PUMPS ──────────────┤
│
COOLING ─────────────────┤
▼
┌──────────────┐
│ ELECTRICITY │
│ DEMAND │
└──────┬───────┘
│
┌───────────┴───────────┐
▼ ▼
GENERATION FLEXIBILITY
Solar / Wind Batteries
Nuclear Demand response
Gas Smart charging
Hydro Thermal storage
│ │
└───────────┬───────────┘
▼
┌──────────────────┐
│ TRANSMISSION │
└────────┬─────────┘
▼
┌──────────────────┐
│ SUBSTATIONS │
│ & TRANSFORMERS │
└────────┬─────────┘
▼
┌──────────────────┐
│ DISTRIBUTION │
└────────┬─────────┘
▼
┌──────────────────────────┐
│ DIGITAL & ELECTRIFIED │
│ ECONOMY │
└──────────────────────────┘
THE BOTTLENECK CAN OCCUR ANYWHERE.The Core Numbers at a Glance
| Indicator | Current / projected figure | What it tells us |
|---|---|---|
| Global electricity-demand growth, 2026–2030 | 3.6%/yr | Demand growth is accelerating |
| Data-centre electricity demand, 2025 | ~485 TWh | Already a major concentrated load |
| Data-centre electricity demand, 2030 | ~950 TWh | Roughly doubles in five years |
| Data-centre share of global electricity, 2030 | ~3% | Globally manageable, locally disruptive |
| Global EV electricity use, 2024 | ~180 TWh | Rapidly growing new load |
| EV electricity demand, 2030 STEPS | ~780 TWh | Transport electrification becomes significant |
| Global annual grid investment today | ~$400 billion | Current infrastructure spending |
| Additional grid investment needed by 2030 | ~50% higher | Networks need to catch up |
| Projects stalled in global grid queues | >2,500 GW | Connection capacity is a major bottleneck |
| U.S. generation/storage projects in queues, end-2025 | ~1,312 GW generation + ~749 GW storage | Large development pipeline does not equal connected capacity |
The figures above come primarily from the IEA’s Electricity 2026, Energy and AI, Global EV Outlook 2025, and World Energy Investment 2025 analyses, together with U.S. Department of Energy and Berkeley Lab grid studies.
References & Further Reading
- International Energy Agency — Electricity 2026: the most directly relevant current global assessment of electricity demand, generation, grids and flexibility through 2030. IEA — Electricity 2026
- IEA — Electricity 2026: Grids: particularly useful for the infrastructure lead-time mismatch, grid queues, investment requirements and supply-chain constraints. IEA — Electricity 2026: Grids
- IEA — Energy and AI: detailed analysis of data-centre electricity consumption, AI workloads, generation requirements and energy-system implications. IEA — Energy and AI
- IEA — Key Questions on Energy and AI: updated 2026 assessment of data-centre demand, bottlenecks, onsite generation and AI’s longer-term electricity trajectory. IEA — Key Questions on Energy and AI
- IEA — Global EV Outlook 2025: electricity-demand implications of EV adoption through 2030. IEA — Global EV Outlook 2025
- IEA — World Energy Investment 2025: global investment trends across generation, grids, storage and electrification. IEA — World Energy Investment 2025
- U.S. Department of Energy — National Transmission Needs Study: current U.S. assessment of transmission requirements associated with load growth, data centres, manufacturing and electrification. DOE — National Transmission Needs Study
- U.S. Department of Energy — Grid Supply Chain: evidence on transformer and other grid-equipment supply constraints and lead times. DOE — Supply Chain and Market Analysis
- Lawrence Berkeley National Laboratory — Queued Up 2026: U.S. generation and storage interconnection-queue data through the end of 2025. Berkeley Lab — Queued Up 2026
- IEA — Electricity Mid-Year Update 2026: latest mid-year assessment of electricity-demand growth in 2026–2027. IEA — Electricity Mid-Year Update 2026