Skip to content
Apixo
Blog
news· 3 min read· via The Guardian AI

IEA Warns AI Datacentre Emissions Could Surge 70% by 2030

A new IEA report reveals global datacentre carbon emissions will jump 70% within four years as the US and China lean on fossil fuels to fuel AI growth.

IEA Warns AI Datacentre Emissions Could Surge 70% by 2030

Global datacentre carbon emissions are on track to increase by 70% over the next four years, according to a recent warning from the International Energy Agency (IEA). The Paris-based energy watchdog stated that datacentres will double their electricity consumption by 2030, with artificial intelligence expansion driving three-quarters of that surge. The rapid pace of the sector means AI facilities will utilize an additional 350 terawatt-hours of electricity annually within four years—surpassing the total yearly electricity consumption of Great Britain.

The agency's annual report also highlighted that the global energy outlook faces heightened uncertainty following the impact of the Middle East war, which the IEA described as the largest energy market disruption in history. Fatih Birol, the IEA’s executive director, noted there is no credible pathway for the energy sector to return to its previous state. Instead, future energy markets will be dictated by government policies addressing a rising global population, increased demand for air-conditioning, and a booming AI economy.

Fossil Fuels Drive US and China AI Expansion

While datacentre developers have become the world's leading corporate purchasers of renewable energy to supply their grid-connected servers, reliance on fossil fuels is projected to grow at an even faster rate, particularly in the US and China. By the end of the decade, China is expected to draw roughly half of its AI datacentre electricity from coal plants, alongside renewable energy, smaller contributions from nuclear reactors, and minimal gas power.

Conversely, the US is projected to depend on gas for approximately 60% of its datacentre electricity by 2030. As of August, plans existed for about 180GW of new gas-fired power plants connected to US datacentres—a capacity nearly matching the total installed gas capacity across the entire European Union. This reliance contrasts sharply with Canada, where plans account for 11GW of gas generation, and the rest of the world, which totals just 4GW of planned gas capacity for AI developers.

Renewables will take a larger share outside the US and China. In the European Union, the IEA projects that datacentres will obtain about 70% of their electricity from renewables by 2030, supplemented by nuclear power, gas generation, and a minor share of coal.

Tracking the Scale of AI Infrastructure

The IEA formulated its projections by analyzing the notoriously opaque industry using figures on annual global shipments of AI servers, which have multiplied eighteen times since 2020. Researchers also utilized space-based imagery to monitor the construction of 93 large "AI factories" globally. This analysis indicates that global datacentre capacity has grown tenfold over the past two years to reach 17GW, and is expected to hit about 50GW upon completion of these facilities.

Current global datacentre capacity stands at 120GW, with 80% located across China, the US, and the EU. An additional 120GW is scheduled for construction by the end of the decade under the IEA's central base case, with three-quarters dedicated to AI datacentres.

What it means for developers

As infrastructure growth accelerates and energy demands shift regional power grids toward fossil fuels, the environmental footprint of large-scale AI deployment is coming under tighter scrutiny. Developers working with heavy computational workloads must increasingly navigate the energy constraints and policy changes surrounding major model providers. For those building applications and looking to experiment efficiently, developers can try top AI models cheaply through one API at https://apixoai.online.


Source: Datacentres to create 70% more carbon emissions by 2030, IEA report warns — The Guardian AI. Written by the Apixo team from that report.

#ai-news#artificial-intelligence#energy#iea#datacentres#carbon-emissions
Try it with your own tools

One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.

Get your API key

Keep reading