- Wood Mackenzie says carbon capture can decarbonise gas powered data centres within 3 to 4 years.
- US data centre emissions are 48% higher than the national grid average amid accelerating AI driven power demand.
- Around 100 GW to 200 GW of new global data centre capacity will be required by 2030, with gas expected to dominate near term power additions.
Carbon capture technology is emerging as the most practical and scalable pathway to decarbonise the rapidly expanding global data centre sector, according to new analysis from Wood Mackenzie.
The report comes as global demand for artificial intelligence infrastructure accelerates the construction of energy intensive data centres, with an estimated 100 GW to 200 GW of new capacity expected to be added by 2030. Much of this growth is set to be powered by natural gas, particularly in the United States, where all three major combined cycle gas turbine manufacturers are reportedly operating at full capacity and expanding production.
Wood Mackenzie said carbon capture attached to gas fired power plants offers the fastest route to reducing emissions from the sector, particularly as alternative technologies such as advanced geothermal and small modular nuclear reactors remain years away from large scale deployment.
Related news: Wärtsilä Energy awarded 790MW order to power data centre in Texas
In 2025, global data centre electricity demand reached approximately 450 TWh, generating around 0.2 Btpa of CO2 emissions. While this represents just over 0.5% of total global emissions of 38 Btpa, the sector’s emissions are rising significantly faster than most heavy industries.
The analysis found that US data centres emit around 548 kg of CO2 per MWh, approximately 48% above the national grid average, placing increasing pressure on hyperscalers and colocation providers to implement credible decarbonisation strategies.
According to Peter Findlay, Director of CCUS Analytics at Wood Mackenzie, the central issue is no longer whether gas will power data centres, but how emissions from that gas generation will be addressed.
The report noted that approximately 58 GW of gas fired power projects are already under development in Texas alone, reinforcing the role of natural gas in supporting AI driven growth and grid reliability.
Wood Mackenzie estimates that integrating carbon capture into combined cycle gas plants would increase electricity costs by between US$15/MWh and US$45/MWh after accounting for federal 45Q tax incentives. Total power costs would rise to around US$115/MWh while capturing between 92% and 98% of flue gas emissions.
The company said carbon capture systems can be deployed within 3 to 4 years or retrofitted to existing plants within 3 to 5 years, making the technology substantially faster to implement than new nuclear projects, which often require more than a decade to develop.
The report also highlighted several longer-term technologies that could play a role in decarbonising data centre power supply beyond 2030.
Enhanced geothermal systems were identified as a potentially competitive solution by 2030 to 2035, with projected costs as low as US$61/MWh. However, only 1.5 GW of projects are currently in development, and the technology has yet to demonstrate scalability outside established geothermal regions.
Nuclear restart projects were identified as another low carbon option, although only 11.5 GW of retired nuclear capacity is currently available in the US. Small modular reactors also face deployment timelines of at least 5 to 8 years alongside ongoing cost uncertainties.
Long duration energy storage technologies could improve renewable energy integration in future, although current projected costs remain high at between US$100/MWh and US$300/MWh.
While renewables and battery storage will continue contributing to broader grid decarbonisation, Wood Mackenzie cautioned that solar and wind power alone cannot currently satisfy the constant uptime requirements of hyperscale data centres without significant gas backup generation.
The report also noted growing scrutiny around renewable energy credits, which has complicated the decarbonisation case for some renewable powered data centre developments.
Wood Mackenzie said hyperscalers are increasingly balancing competing priorities between rapid AI expansion, energy security and public sustainability commitments.
The company added that carbon capture provides operational flexibility because systems can either be installed during initial plant construction or retrofitted later as regulations, economics and corporate sustainability targets evolve.
Author: Bryan Groenendaal
