How many 100 MW AI data centers will the world need by 2030?
A useful headline answer is about 1,560 units of 100 MW AI capacity.
But that does not mean 1,560 identical buildings. It is a capacity equivalent based on one global forecast.
The number to remember is 1,560. The phrase to remember is 100 MW capacity equivalents, not 1,560 literal data centers.
To understand the estimate, we need to separate two ideas: capacity and annual electricity use.
What Does One 100 MW Equivalent Mean?
In our earlier model, one data center has 100 MW of IT capacity. We assume an average load factor of 85% and a PUE of 1.30.
100 MW × 0.85 × 1.30 = 110.5 MW
IT capacity × load factor × PUE = average facility load
Over one year, that facility uses:
110.5 MW × 8,760 hours = 0.968 TWh per year
The 100 MW data center power model guide explains each step.
The Direct Answer: About 1,560 AI Capacity Equivalents
McKinsey’s continued-momentum scenario estimates that global AI-related data center capacity demand could reach 156 GW by 2030.
One 100 MW unit is 0.1 GW. So the conversion is simple:
156 GW ÷ 0.1 GW = 1,560 capacity equivalents
The same forecast puts AI-related capacity at 44 GW in 2025. That equals 440 units of 100 MW.
Under this scenario, AI capacity grows about 3.5 times from 2025 to 2030. AI would account for roughly 70% of total data center capacity demand.
Why 1,560 Does Not Mean 1,560 Buildings
Data centers do not come in one standard size.
A project may be a 50 MW facility. A large campus may reach several hundred megawatts. Some plans move toward 1 GW campuses built in phases.
The same 156 GW forecast produces very different site counts:
| Average AI capacity per campus | Equivalent campuses |
|---|---|
| 100 MW | 1,560 |
| 250 MW | 624 |
| 500 MW | 312 |
| 1 GW | 156 |
This table is not a construction schedule. It shows why the word equivalent matters.
The Energy View Gives a Different Number
Capacity is measured in GW. Annual electricity use is measured in TWh.
The International Energy Agency’s updated outlook expects total global data center electricity use to rise from 485 TWh in 2025 to about 950 TWh in 2030.
The IEA says electricity use from AI-focused data centers grows faster and roughly triples over the same period.
If we divide the total data center forecast by 0.968 TWh per 100 MW operating model, we get:
2025: 485 TWh ÷ 0.968 ≈ 501 operating equivalents
2030: 950 TWh ÷ 0.968 ≈ 981 operating equivalents
This implies about 480 additional operating equivalents between 2025 and 2030.
But this is a forecast for all data centers, not only AI facilities.
Why 981 and 1,560 Are Both Useful
The two answers measure different things.
| Forecast lens | Measures | 2030 equivalent |
|---|---|---|
| Energy lens | Total data center electricity used during a year | About 981 |
| Capacity lens | AI-related data center capacity demanded | About 1,560 |
They should not be averaged.
A capacity forecast depends on how capacity is defined. An energy forecast depends on utilization, PUE, operating hours, and the mix of facilities.
The safest summary is:
By 2030, one capacity forecast points to about 1,560 units of 100 MW AI capacity. A separate energy forecast points to about 981 operating equivalents for all data centers. These are scale estimates, not literal building counts.
Why the Forecast Could Move
AI infrastructure is growing quickly. It is also uncertain.
Demand could rise faster if AI agents, enterprise tools, and inference workloads spread widely. It could grow more slowly if projects face grid delays, chip shortages, financing limits, or weak business returns.
Better chips can lower energy per task. But cheaper computing can also encourage more AI use.
Forecasts are therefore scenarios, not promises.
What 1,560 Equivalents Would Require
The number is not only about servers.
It points to demand for land, buildings, generation, transmission lines, substations, transformers, backup systems, cooling, and skilled labor.
McKinsey estimates that meeting AI-related compute demand could require about $5.2 trillion in worldwide capital investment by 2030 under its central scenario.
That forecast includes chips and computing equipment, but also power, cooling, electrical systems, and site construction.
The next article follows that physical chain: What Must Be Built to Power the AI Data Center Boom?
Run the Conversion Yourself
The core calculation needs only a few lines of Python.
forecast_ai_capacity_gw = 156
site_capacity_gw = 0.1
capacity_equivalents = (
forecast_ai_capacity_gw
/ site_capacity_gw
)
print(capacity_equivalents)
# 1560This short code shows the core conversion. For the full power-cost model, including load factor, PUE, annual electricity use, and site comparison, use the AI Data Center Power Cost Calculator in Python .
The Number to Remember
The world may need about 1,560 units of 100 MW AI capacity by 2030 under one continued-growth forecast. The real number of campuses could be much smaller or larger because facilities will not all have the same size.
The forecast becomes easier to understand when we stop imagining 1,560 identical buildings.
Think instead of a global pool of AI capacity equal to 1,560 blocks of 100 MW.
Related Articles
- How to Understand a 100 MW Data Center Power Model
- How Much Power Is 100 MW? Compared with Entire Cities
- AI Data Center Power Cost Calculator in Python
Sources and Method
- International Energy Agency, Key Questions on Energy and AI , 2026.
- McKinsey & Company, Data Center Demands , 2025.
- The Contexta model: 100 MW IT capacity × 0.85 load factor × 1.30 PUE × 8,760 hours.
Published: July 2026 · Forecasts are scenarios. Equivalent counts are calculated estimates, not announced construction totals.