[DigitalToday reporter Jae-won Choi (최재원)] A projection showed that if AI data centre expansion in the United States leads to more fossil-fuel power generation, carbon dioxide emissions from the power sector in 2030 could rise 42 percent from 2025. It warned that efforts to reliably supply electricity to data centres could undo emissions cuts achieved over the past 15 years.

On Oct. 29 (local time), IT outlet The Verge reported that U.S. environmental group Sierra Club analysed that if planned AI data centres and related power plants are built on schedule, annual carbon dioxide emissions from electricity production could reach about 2.1 billion tonnes in 2030.

That would be 618 million tonnes more than in 2025, an increase of 42 percent. The increase alone would be comparable to emissions from the annual electricity use of about 129 million U.S. households. The figure is an estimate reflecting business plans and energy supply methods, not actual emissions.

Much of the increase is expected to come from self-generation facilities installed to supply power directly to data centres. As connecting to the power grid takes a long time, more operators are building separate natural gas power plants. Sierra Club estimated additional emissions from such facilities could exceed 460 million tonnes.

Earlier in June, an academic study estimated that in a scenario of rising demand from data centres and cryptocurrency mining, power-sector emissions in 2030 could be 28 percent higher than if there were no such demand increase. The two projections are difficult to compare directly by the numbers alone because their baselines differ.

The Environmental Defense Fund (EDF) also said recently that U.S. plans for new fossil-fuel power facilities have reached their highest level since 2016. By contrast, renewable energy projects are facing difficulties due to permitting delays and policy changes.

Experts stressed that the actual scale of emissions growth will vary depending on the scale of data centre construction, the choice of power sources and the pace of renewable energy deployment. Meeting rising electricity demand from the growth of the AI industry while curbing carbon emissions has emerged as a challenge for U.S. energy policy.