# Data centers — overview of infrastructure, expansion, and limits

Public Dossier DOS-1 · Overview

Schema: 1.0.0
Dossier ID: DOS-1
Slug: data-centers-infrastructure-expansion-limits
Public version: 1.0.0
Locale: en
Intent: explain
Audience: Readers, researchers, investors, operators, and policymakers.
Last verified: 2026-07-29
Canonical: https://blog.nilu.pro/en/blog/data-centers-infrastructure-expansion-limits/dossier/
Article summary: https://blog.nilu.pro/en/blog/data-centers-infrastructure-expansion-limits/
Markdown: https://blog.nilu.pro/en/blog/data-centers-infrastructure-expansion-limits/dossier.md
JSON: https://blog.nilu.pro/en/blog/data-centers-infrastructure-expansion-limits/dossier.json

## Executive summary

- Data centers are physical systems that combine servers, accelerators, memory, storage, and networks with buildings, power, cooling, connectivity, capital, and authorization. The sector's expansion cannot be explained by a single component. AI accelerates part of the demand; power has become a significant constraint; chips, networks, electrical equipment, territory, and regulation remain interdependent.
- Announced, contracted, installed, energized, usable, and actually consumed capacity are distinct states.
- Measurement gaps prevent rankings and false precision, but they do not prevent explanations of known mechanisms.

## Scope

How digital demand becomes physical infrastructure; which resources and bottlenecks constrain expansion; and how costs, impacts, and power are distributed.

**Thesis:** Data centers are not abstract digital infrastructure. They are cyber-physical systems whose expansion depends on the convergence of compute, power, cooling, connectivity, capital, equipment, territory, and authorization.

**Out of scope**

- Technical, environmental, fiscal, or legal details remain in the specialized Dossiers.
- The Overview does not reproduce the complete evidence bases of the four specialized Dossiers.

## Specialized Dossiers

- DOS-6 — [Data centers — how they work and why they are expanding](https://blog.nilu.pro/pt-br/blog/datacenters-sistema-fisico-demanda/)
- DOS-2 — [Data centers — energy and environmental impacts](https://blog.nilu.pro/pt-br/blog/datacenters-energia-impactos-ambientais/)
- DOS-3 — [Data centers — political economy, investment, and local impacts](https://blog.nilu.pro/pt-br/blog/datacenters-economia-politica-impactos-locais/)
- DOS-4 — [Data centers — geography, geopolitics, and sovereignty](https://blog.nilu.pro/pt-br/blog/datacenters-geografia-geopolitica-soberania/)

## System map

```text
demand for digital services
        ↓
workloads + delivery model
        ↓
servers, accelerators, memory, storage, and networking
        ↓
racks, critical power, cooling, and building
        ↓
grid connection + fiber + cables + territory
        ↓
contracts, financing, incentives, and regulation
        ↓
installed → energized → usable → actually consumed capacity
        ↓
economic, environmental, territorial, and strategic effects
```

## DATA — evidence and measurements

### DC-OV-D01 — A data center combines IT and supporting infrastructure

A data center combines servers, storage, and networking with power supply, conversion, redundancy, cooling, security, and a building.

- **Types:** fact
- **Status:** verified
- **Confidence:** high
- **Last reviewed:** 2026-07-25
- **References:** [src-epa-energy-star-datacenter-faq-pue](#src-epa-energy-star-datacenter-faq-pue), [src-lbnl-us-data-center-energy-2024](#src-lbnl-us-data-center-energy-2024)

**Limitations**

- Implementations vary by facility type and availability requirements.

### DC-OV-D02 — Cloud is not a workload

Cloud is a delivery and operating model; AI, video, databases, web services, analytics, HPC, and storage are applications or workload classes that can share the same infrastructure.

- **Types:** fact, taxonomy
- **Status:** verified
- **Confidence:** high
- **References:** [src-nist-sp800-145-cloud-definition](#src-nist-sp800-145-cloud-definition)

**Limitations**

- Application categories can overlap in their use of resources, making mutually exclusive shares difficult to determine.

### DC-OV-D03 — Global projections are specific to an edition and scenario

The IEA estimated about 415 TWh for data centers in 2024 in the 2025 Energy and AI edition; a later update used approximately 485 TWh in 2025 and a central scenario close to 950 TWh in 2030.

- **Types:** estimate, prediction
- **Status:** contextual
- **Period:** Base years 2024 and 2025; central scenario for 2030.
- **Confidence:** high for attribution; medium for projection
- **Risk of misinterpretation:** Do not present 2030 as a future fact.
- **References:** [src-iea-energy-and-ai-2025](#src-iea-energy-and-ai-2025), [src-iea-key-questions-energy-ai-2026](#src-iea-key-questions-energy-ai-2026)

**Limitations**

- The values belong to different editions, base years, and assumptions; they do not form a single range.

### DC-OV-D04 — Concentration makes local impact disproportionate

Data centers accounted for 23% of Ireland's measured electricity consumption in 2025, compared with 5% in 2015.

- **Types:** fact
- **Status:** verified
- **Measurement basis:** Metered electricity consumption.
- **Period:** 2015–2025
- **Confidence:** high
- **References:** [src-cso-ireland-datacenter-electricity-2025](#src-cso-ireland-datacenter-electricity-2025)

**Limitations**

- This is not a share of all final energy and is not a global measure.

### DC-OV-D05 — An AI-dense configuration does not represent a typical rack

NVIDIA documentation reports approximately 120 kW for an NVL72 configuration; in the 2025 Uptime survey, 82% of 709 responding operators did not have a rack above 30 kW as the highest-density rack in the facility.

- **Types:** comparison
- **Status:** contextual
- **Measurement basis:** Specification for one NVL72 configuration compared with a survey of 709 operators.
- **Period:** Uptime 2025 survey.
- **Confidence:** high for the sources; low for global generalization
- **References:** [src-nvidia-dgx-gb-rack-hardware-guide](#src-nvidia-dgx-gb-rack-hardware-guide), [src-uptime-global-datacenter-survey-2025](#src-uptime-global-datacenter-survey-2025)

**Limitations**

- A product specification and an operator survey cover different populations; the survey is not a distribution of all racks.

### DC-OV-D06 — TDP and benchmarks do not measure annual site consumption

TDP is a component property; MLPerf Power measures a system's AC power during a defined benchmark; neither measure, in isolation, provides a data center's annual consumption.

- **Types:** comparison
- **Status:** verified
- **Measurement basis:** Component TDP and system AC power during a defined benchmark.
- **Confidence:** high
- **References:** [src-nvidia-h100-specifications](#src-nvidia-h100-specifications), [src-mlcommons-mlperf-inference](#src-mlcommons-mlperf-inference), [src-mlcommons-mlperf-power](#src-mlcommons-mlperf-power)

### DC-OV-D07 — Water use depends on boundaries and conditions

An LBNL study found a variation of more than four orders of magnitude in water consumption per workload among the evaluated combinations of equipment, utilization, cooling, climate, power grid, and hardware refresh.

- **Types:** estimate, comparison
- **Status:** contextual
- **Measurement basis:** Water consumption per workload among the evaluated combinations.
- **Confidence:** medium-high
- **References:** [src-lbnl-water-use-workloads-2025](#src-lbnl-water-use-workloads-2025)

**Limitations**

- This does not represent the observed distribution of global sites and does not justify a universal number 'per query.'

### DC-OV-D08 — A renewable contract does not prove hourly physical supply

Market-based Scope 2 records contractual instruments; on its own, it does not demonstrate that consumption was supplied by carbon-free generation in every hour and location.

- **Types:** comparison
- **Status:** verified
- **Confidence:** high
- **References:** [src-ghg-protocol-scope2-guidance-2015](#src-ghg-protocol-scope2-guidance-2015)

**Limitations**

- Distinguish the current rule from future proposals for hourly matching and deliverability.

### DC-OV-D09 — In the Virginia case, modeled economic impact is predominantly construction

The Virginia model estimated 74,000 jobs associated with the sector: 59,000 tied to construction and 15,000 to operations, including direct, indirect, and induced effects.

- **Types:** estimate
- **Status:** contextual
- **Measurement basis:** Modeled jobs, including direct, indirect, and induced effects.
- **Confidence:** high for attribution
- **References:** [src-jlarc-data-centers-virginia-2024](#src-jlarc-data-centers-virginia-2024)

**Limitations**

- This is not equivalent to permanent employees of operators and does not represent other jurisdictions.

### DC-OV-D10 — Use of an incentive does not prove additionality

The Washington audit found fewer than ten beneficiaries, only refurbishments, no new construction under the preference, and an estimated US$42.4 million in savings between 2023–2026.

- **Types:** fact, estimate
- **Status:** verified Case Study
- **Period:** 2023–2026
- **Confidence:** high
- **References:** [src-wa-jlarc-urban-data-centers-tax-preference-2026](#src-wa-jlarc-urban-data-centers-tax-preference-2026)

**Limitations**

- This does not measure every incremental investment and does not support national generalization.

### DC-OV-D11 — Regional electricity consumption is not a share of compute

For 2024, the IEA attributed 45%, 25%, and 15% of estimated data-center electricity consumption to the United States, China, and Europe.

- **Types:** estimate
- **Status:** contextual
- **Measurement basis:** Share of estimated data-center electricity consumption.
- **Period:** 2024
- **Confidence:** high for the published estimate
- **References:** [src-iea-energy-and-ai-2025](#src-iea-energy-and-ai-2025)

**Limitations**

- This does not measure GPUs, FLOPs, servers, or usable capacity.

### DC-OV-D12 — Cables are critical, but cable count does not measure resilience

The ITU estimates that submarine cables carry more than 99% of international data traffic; faults, repairs, capacity, route diversity, and unavailability are distinct metrics.

- **Types:** estimate, comparison
- **Status:** contextual
- **Measurement basis:** Estimated share of international data traffic.
- **Confidence:** medium-high
- **References:** [src-itu-submarine-cable-resilience-2026](#src-itu-submarine-cable-resilience-2026)

### DC-OV-D13 — Location is not comprehensive sovereignty

Keeping data and data centers within a territory does not, on its own, determine effective jurisdiction, operational control, keys, portability, technological substitutability, or continuity.

- **Types:** comparison
- **Status:** verified as framework
- **Confidence:** high
- **References:** [src-oecd-data-localisation-2020](#src-oecd-data-localisation-2020)

## INFO — derived syntheses

### DC-OV-I01 — Expansion is the coordination of dependencies

Chips without power create unproductive inventory; power without hardware, connectivity, or demand creates idle capacity; neither becomes operational without permits and contracts.

- **Derived from:** DC-OV-D01, DC-OV-D05, DC-OV-D06

### DC-OV-I02 — The bottleneck moves through the system

The constraint can move among chips, HBM, packaging, networking, electrical equipment, grid connection, cooling, construction, and commissioning.

- **Derived from:** DC-OV-D01, DC-OV-D05
- **Provenance note:** The canonical derivation also points to the specialized Dossiers on Operations and Energy, without identifying additional claims in the Overview.

### DC-OV-I03 — Unit efficiency and aggregate consumption can diverge

Lower energy use per task does not determine total consumption; volume, price, utilization, and new uses also change.

- **Derived from:** DC-OV-D03, DC-OV-D06

### DC-OV-I04 — Small global impacts can be large locally

A global share of electricity or water does not describe pressure on a grid node, watershed, or municipality.

- **Derived from:** DC-OV-D04, DC-OV-D07, DC-OV-D11

### DC-OV-I05 — Benefits and costs appear in different institutions

A state tax expenditure, municipal revenue, utility investment, electricity rates, and temporary employment can coexist; they should not be added together as a net benefit without a counterfactual.

- **Derived from:** DC-OV-D09, DC-OV-D10

### DC-OV-I06 — Location is simultaneously an advantage and a dependency

Power, fiber, and regulation attract capacity, but chips, cloud, software, cables, and jurisdiction preserve external dependencies.

- **Derived from:** DC-OV-D11, DC-OV-D12, DC-OV-D13

### DC-OV-I07 — A projection or incentive does not demonstrate delivered capacity

Demand projections and incentive programs do not, on their own, demonstrate operational capacity; deployment still depends on financing, equipment, grid connection, permits, construction, and operations.

- **Derived from:** DC-OV-D03, DC-OV-D10

## Uncertainties

### DC-OV-UNC-01 — Global share by workload

- **Status:** not publicly measured
- **What we know:** Adoption and presence are not equivalent to energy, servers, or compute share.
- **Editorial treatment:** Do not publish an artificial global breakdown.
- **Confidence:** high confidence in the gap
- **Last reviewed:** 2026-07-25

### DC-OV-UNC-02 — Installed versus usable capacity

- **Status:** not publicly measured
- **What we know:** Power, cooling, space, networks, and reliability can produce stranded capacity.
- **Editorial treatment:** Explain the mechanisms without a global percentage.
- **Last reviewed:** 2026-07-25

### DC-OV-UNC-03 — Comparable environmental impact by site

- **Status:** incompatible methods
- **What we know:** PUE, WUE, water, emissions, and embodied carbon depend on boundary, climate, and operations.
- **Editorial treatment:** Preserve the site, period, denominator, and method.
- **Last reviewed:** 2026-07-25

### DC-OV-UNC-04 — Net fiscal return

- **Status:** not publicly measured
- **What we know:** Gross revenue and tax expenditures exist at different levels of government.
- **Editorial treatment:** Do not produce an incentive ranking.
- **Last reviewed:** 2026-07-25

### DC-OV-UNC-05 — Global compute and sovereignty

- **Status:** imperfect proxy
- **What we know:** Electricity, racks, facilities, and location measure partial dimensions.
- **Editorial treatment:** Use a multidimensional framework, not a single index.
- **Last reviewed:** 2026-07-25

## Methodology

- **DATA:** DATA records verifiable claims with explicit sources, measurement context, limitations, and confidence.
- **INFO:** INFO records synthesis or interpretation and always declares the IDs of the claims from which it is derived.
- **Uncertainty:** An Uncertainty records a gap as an editorial result; it remains explicit and is not filled with a fragile estimate.
- **Update rule:** A review may correct or update the projection only from the canonical Dossier with editorial authorization; material changes increment the public version and are recorded in the corrections history.

## Version and corrections

- **Public version:** 1.0.0
- **Last verified:** 2026-07-29
- **Review after:** 2026-08-24
- No public corrections recorded.

## References

1. <a id="src-epa-energy-star-datacenter-faq-pue"></a>**src-epa-energy-star-datacenter-faq-pue.** [U.S. EPA ENERGY STAR — Data center rating FAQ and PUE definition](https://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=P100TSZE.TXT). U.S. EPA. Accessed 2026-07-24.
1. <a id="src-lbnl-us-data-center-energy-2024"></a>**src-lbnl-us-data-center-energy-2024.** [LBNL — 2024 United States Data Center Energy Usage Report](https://eta-publications.lbl.gov/publications/2024-lbnl-data-center-energy-usage-report). Lawrence Berkeley National Laboratory. Accessed 2026-07-24.
1. <a id="src-nist-sp800-145-cloud-definition"></a>**src-nist-sp800-145-cloud-definition.** [NIST — The Definition of Cloud Computing (SP 800-145)](https://csrc.nist.gov/pubs/sp/800/145/final). NIST. Accessed 2026-07-24.
1. <a id="src-iea-energy-and-ai-2025"></a>**src-iea-energy-and-ai-2025.** [IEA — Energy and AI (2025)](https://www.iea.org/reports/energy-and-ai). IEA. Accessed 2026-07-22.
1. <a id="src-iea-key-questions-energy-ai-2026"></a>**src-iea-key-questions-energy-ai-2026.** [IEA — Key Questions on Energy and AI (2026)](https://www.iea.org/reports/key-questions-on-energy-and-ai). IEA. Accessed 2026-07-24.
1. <a id="src-cso-ireland-datacenter-electricity-2025"></a>**src-cso-ireland-datacenter-electricity-2025.** [CSO Ireland — Data Centres Metered Electricity Consumption 2025](https://www.cso.ie/en/releasesandpublications/ep/p-dcmec/datacentresmeteredelectricityconsumption2025/). Central Statistics Office Ireland. Accessed 2026-07-24.
1. <a id="src-nvidia-dgx-gb-rack-hardware-guide"></a>**src-nvidia-dgx-gb-rack-hardware-guide.** [NVIDIA — DGX GB Rack Scale Systems Hardware Guide](https://docs.nvidia.com/dgx/dgxgb200-user-guide/hardware.html). NVIDIA. Accessed 2026-07-24.
1. <a id="src-uptime-global-datacenter-survey-2025"></a>**src-uptime-global-datacenter-survey-2025.** [Uptime Institute — 2025 Global Data Center Survey](https://uptimeinstitute.com/about-ui/press-releases/uptimes-15th-annual-global-data-center-survey-results-shows-both-commitment-and-hesitancy). Uptime Institute. Accessed 2026-07-24.
1. <a id="src-nvidia-h100-specifications"></a>**src-nvidia-h100-specifications.** [NVIDIA — H100 Tensor Core GPU specifications](https://www.nvidia.com/en-us/data-center/h100/). NVIDIA. Accessed 2026-07-22.
1. <a id="src-mlcommons-mlperf-inference"></a>**src-mlcommons-mlperf-inference.** [MLCommons — MLPerf Inference benchmark documentation](https://docs.mlcommons.org/inference/). MLCommons. Accessed 2026-07-24.
1. <a id="src-mlcommons-mlperf-power"></a>**src-mlcommons-mlperf-power.** [MLCommons — MLPerf Inference power measurement methodology](https://docs.mlcommons.org/inference/power/). MLCommons. Accessed 2026-07-24.
1. <a id="src-lbnl-water-use-workloads-2025"></a>**src-lbnl-water-use-workloads-2025.** [LBNL — The water use of data center workloads (2025)](https://eta-publications.lbl.gov/publications/water-use-data-center-workloads). Lawrence Berkeley National Laboratory. Accessed 2026-07-24.
1. <a id="src-ghg-protocol-scope2-guidance-2015"></a>**src-ghg-protocol-scope2-guidance-2015.** [GHG Protocol — Scope 2 Guidance](https://ghgprotocol.org/scope-2-guidance). GHG Protocol / WRI. Accessed 2026-07-24.
1. <a id="src-jlarc-data-centers-virginia-2024"></a>**src-jlarc-data-centers-virginia-2024.** [JLARC — Data Centers in Virginia (2024)](https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp). Virginia JLARC. Accessed 2026-07-24.
1. <a id="src-wa-jlarc-urban-data-centers-tax-preference-2026"></a>**src-wa-jlarc-urban-data-centers-tax-preference-2026.** [Washington JLARC — Data Centers in Urban Counties: 2026 Tax Preference Review](https://leg.wa.gov/jlarc/taxReports/2026/data/1/report.html). Washington JLARC. Accessed 2026-07-24.
1. <a id="src-itu-submarine-cable-resilience-2026"></a>**src-itu-submarine-cable-resilience-2026.** [ITU — International Advisory Body on Submarine Cable Resilience: Working Groups Report 2026](https://www.itu.int/hub/publication/s-iab-wg-2026/). ITU. Accessed 2026-07-24.
1. <a id="src-oecd-data-localisation-2020"></a>**src-oecd-data-localisation-2020.** [OECD — Data localisation trends and challenges](https://www.oecd.org/content/dam/oecd/en/publications/reports/2020/12/data-localisation-trends-and-challenges_d775fe8a/7fbaed62-en.pdf). OECD. Accessed 2026-07-24.