Global Impact Report

Global Environmental Impact of AI

AI is reshaping global energy demand. Data centres consumed 415 TWh in 2024 -- projected to reach 945 TWh by 2030, equivalent to Japan's entire electricity consumption.

415 TWh
DC Electricity (2024)
945 TWh
Projected 2030
64B litres
US DC Water (2023)
600M
Daily AI Users

How AI Compares

AI's carbon footprint is growing faster than any other tech sector. Here's how it stacks up against familiar benchmarks.

vs Global Aviation

AI
120 Mt
Aviation
920 Mt

AI emissions are ~13% of global aviation (2024 est.), but growing much faster

vs Bitcoin Mining

AI
120 Mt
Bitcoin
70 Mt

AI estimated to emit ~1.7x more CO2 than Bitcoin mining (both figures carry uncertainty)

vs Video Streaming

AI
120 Mt
Streaming
55 Mt

AI inference estimated at ~2x more than all global video streaming

AI Emissions Over Time

AI workloads emitted an estimated 120 Mt CO₂ in 2024 — more than Belgium's entire annual emissions. By 2030, that could reach 300 Mt on current trajectory.

  • AI Emissions (Mt)
  • All Data Centers (Mt)
20182019202020212022202320242025202620272028202920300200400600800Mt CO2Now
2018: Pre-LLM era, AI mostly research workloads2020: COVID dip in global emissions; AI training scales up2022: ChatGPT launches; inference demand begins surge2024: AI inference dominates; 415 TWh data center total2027: Projected: efficiency gains slow growth rate2030: Projected: AI ~0.85% of global emissions

Source: IEA, Goldman Sachs, Ember Climate. Dashed line marks the present — everything after is projected.

Projected Energy Demand

Data center electricity is projected to reach 945 TWh by 2030 — a 165% increase in six years.

  • AI Share (TWh)
  • Total Data Center (TWh)
202420252026202720282029203002505007501000TWh

Source: IEA, Goldman Sachs projections

Where the Energy Goes

Inference — answering queries in real-time — dominates AI's energy footprint.

85%Inference
Inference — 85%
Training — 15%

Training gets the headlines, but inference is the iceberg beneath. Every question you ask an AI contributes to the 85% slice.

Efficiency gains make inference cheaper, driving more usage — total energy still rises (Jevons Paradox).

Country Carbon Intensity

The same AI query can produce up to 35x more CO₂ depending on data center location. Sorted cleanest to dirtiest.

NorwayDC↗ Greening
20 g CO₂/kWh
SwedenDC↗ Greening
25 g CO₂/kWh
FranceDC→ Stable
56 g CO₂/kWh
Brazil→ Stable
80 g CO₂/kWh
CanadaDC↗ Greening
120 g CO₂/kWh
United KingdomDC↗ Greening
170 g CO₂/kWh
GermanyDC↗ Greening
350 g CO₂/kWh
United StatesDC↗ Greening
380 g CO₂/kWh
South KoreaDC→ Stable
420 g CO₂/kWh
JapanDC→ Stable
450 g CO₂/kWh
AustraliaDC↗ Greening
500 g CO₂/kWh
ChinaDC↗ Greening
550 g CO₂/kWh
Saudi ArabiaDC→ Stable
600 g CO₂/kWh
IndiaDC↗ Greening
700 g CO₂/kWh
IndonesiaDC↘ Worsening
720 g CO₂/kWh

Future Paths

AI emissions through 2030 depend on policy, technology, and energy choices. Explore four possible scenarios.

20242025202620272028202920300150300450600Mt CO2

Current trajectory with moderate efficiency gains.

The Cost of Training

Training a single frontier model now produces as much CO₂ as thousands of transatlantic flights.

BERT2019
0.6t1 flights
GPT-22019
3t4 flights
GPT-32020
552t680 flights
PaLM2022
271t334 flights
GPT-42023
5,000t6,170 flights
Llama 3 405B2024
3,500t4,320 flights
GPT-5 (est.)2025
10,000t12,350 flights

Flights = NY-London return. Sources: Strubell et al., Patterson et al., industry estimates.

Provider Transparency

Corporate emissions may be 7.62x higher than reported due to market-based accounting.

Google

High
Renewable
100% matched
Emissions
+50% since 2019
Water
+88% since 2019

Published per-query energy data for Gemini. 33x efficiency gain in one year.

Microsoft

High
Renewable
34 GW contracted
Emissions
+23.4% since 2020
Water
WUE improved 39%

Carbon negative by 2030 goal. Launched zero-water cooling designs.

Amazon (AWS)

Medium
Renewable
100% matched (claimed)
Emissions
Rising in 2024
Water
Not disclosed in detail

Largest corporate renewable buyer 5 years running.

Meta

Medium
Renewable
Grid-based matching
Emissions
Growing with AI investment
Water
Not disclosed in detail

Pursuing nuclear power for data centers.

OpenAI

Low
Renewable
Not published
Emissions
Not published
Water
Not published

No dedicated sustainability report as of 2026.

Anthropic

Low
Renewable
Not published
Emissions
Not published
Water
Committed to water-efficient cooling

Claude scored highest eco-efficiency on AWS. No emissions data published.

What's Being Done

The picture isn't all bleak. Efficiency gains, clean energy deals, and regulation are shaping a better trajectory.

33x Efficiency Gain

Google achieved a 33x reduction in Gemini's per-query energy (May 2024 - May 2025) through hardware upgrades, distillation, and inference optimisation.

Mixture-of-Experts

MoE architectures like DeepSeek-V3 activate only ~5% of parameters per query, dramatically cutting computation per inference.

Nuclear Renaissance

Three Mile Island may reopen in 2028 for Microsoft. Amazon and Google are also buying dedicated nuclear capacity.

Immersion Cooling

Servers submerged in dielectric fluid achieve PUE of 1.03 vs the 1.3 industry average — cutting cooling energy by up to 95%.

EU Regulation

The EU Energy Efficiency Directive mandates PUE reporting for data centers >500 kW — the first major DC regulatory framework.

Cost Efficiency

Inference costs at GPT-3.5 level dropped 280x in two years (Stanford AI Index), making AI accessible while reducing per-query impact.

Infrastructure Evolution

Key hardware, grid, cooling, and policy milestones shaping AI's emissions trajectory.

2017mixed
NVIDIA V100 (Volta)

First tensor core GPU; enabled large-scale AI training. 300W TDP.

2020↓ emissions
NVIDIA A100 (Ampere)

2.5x training efficiency over V100. Became the workhorse of AI data centers.

2022mixed
NVIDIA H100 (Hopper)

3x inference throughput over A100 with transformer engine. 700W TDP.

2024mixed
NVIDIA B200 (Blackwell)

4x inference throughput over H100; ~2.5× better perf/watt. 1000W TDP (up from 700W).

2025↓ emissions
Custom AI chips (TPUs, Trainium, etc.)

Google TPU v6e, AWS Trainium2: 2–3x better perf/watt than general GPUs for transformer workloads.

2021↓ emissions
Corporate renewable PPAs surge

Tech companies become largest corporate buyers of renewable energy globally.

2023↑ emissions
Data centers strain local grids

Virginia (largest DC market) faces power shortages. Ireland DCs consume 21% of grid.

2024↓ emissions
Nuclear renaissance for AI

Microsoft signs Three Mile Island restart deal. Amazon/Google buy nuclear capacity.

2025↑ emissions
Natural gas buildout for DCs

Some providers building dedicated gas plants, undermining climate goals.

2026↓ emissions
SMR pilots for data centers

Small modular reactors targeted for DC-adjacent deployment by 2028–2030.

2022↓ emissions
Liquid cooling adoption begins

Direct-to-chip cooling reduces energy use 30–40% vs air cooling for high-density racks.

2024↓ emissions
Microsoft zero-water cooling

New designs eliminate evaporative cooling, drastically cutting water consumption.

2025↓ emissions
Immersion cooling scales

Servers submerged in dielectric fluid; PUE approaches 1.03 (vs 1.3 industry avg).

2023↓ emissions
EU Energy Efficiency Directive

Mandates PUE reporting for data centers >500 kW. First major DC regulation.

2024↓ emissions
US Executive Order on AI

Includes environmental reporting requirements for large AI training runs.

2025↑ emissions
IEA issues DC energy warning

Projects data centers could consume 3–4% of global electricity by 2030 (up to ~1,300 TWh).

Sources: Stanford AI Index, SemiAnalysis, Uptime Institute, IEA

Sources & References

24 sources cited on this page. All data points are linked to their original source.

ReportAcademicNewsDatasetProvider
[4]
[5]
AcademicPatterson et al. 2021 (2021)
[7]
DatasetEmber 2025 (2025)
[9]
ReportStanford AI Index 2025 (2025)
[11]
AcademicChowdhery et al. 2022 (2022)
[12]
ProviderGoogle Environmental Report 2024 (2024)
[13]
ProviderMicrosoft Sustainability Report 2024 (2024)
[14]
AcademicEpoch AI 2025 (2025)
[15]
ReportBloomberg NEF 2025 (2025)
[16]
DatasetIRENA 2024 (2024)
[17]
DatasetCambridge Bitcoin Index (2025)
[19]
DatasetATAG Aviation Facts (2024)
[20]
ReportEU Energy Efficiency Directive (2023)
[21]
[22]
AcademicMeta Llama 3 Report (2024)
[24]
ReportUptime Institute 2024 (2024)

Curious about your own impact?

Calculate the energy, carbon, and water cost of your AI usage.

Calculate My Compute