With 100-200 GW of new data centre capacity needed by 2030, carbon capture on gas offers fastest route to decarbonisation

With natural gas dominating near-term data centre power additions − all three global combined-cycle gas turbine manufacturers are backlogged and expanding capacity − carbon capture offers the most practical path to decarbonization, according to new Wood Mackenzie analysis. While technologies like enhanced geothermal promise lower costs at scale, carbon capture on gas can be deployed within 3-4 years and retrofitted to the massive gas buildout already underway.

In 2025, approximately 450 TWh of data centre power demand generated 0.2 Btpa of CO2 emissions, representing over 0.5% of the world's total 38 Btpa. By comparison, the steel, chemical and cement industries emitted much more at 3.5 Btpa, 3.0 Btpa and 2.3 Btpa, respectively. Therefore, decarbonizing data centres is unlikely to itself substantially alleviate climate change.

However, data centre emissions are growing much faster than other sectors, with US data centres operating at 548 kg CO2 per MWh, 48% above the national grid average. As the market expands, hyperscalers and colocation developers face mounting pressure to pursue viable decarbonization pathways.

Wood Mackenzie's new report series, "Decarbonizing data centres amid the gold rush for power: viable pathways," finds that while multiple decarbonization technologies show promise, carbon capture offers the most immediate and scalable solution given current market realities.

"With gas power dominating buildouts right now and 58 GW already in development in Texas alone, the practical question isn't whether data centers will use gas, it's whether that gas will be decarbonized," said Peter Findlay, Director of CCUS Analytics at Wood Mackenzie. "Carbon capture offers the fastest, most scalable path to do that."

The analysis shows that adding carbon capture to combined-cycle gas plants raises costs by $15-$45 per MWh in the US, after accounting for federal 45Q tax incentives, bringing total power costs to approximately $115/MWh. The technology can capture 92-98% of flue gas emissions and can be deployed within 3-4 years—or retrofitted to existing plants within 3-5 years—making it significantly faster than alternatives like new nuclear reactors, which require more than a decade.

"At roughly $115 per MWh including capture, this represents a manageable premium for decarbonized power, especially for hyperscalers with the cash flow to absorb these costs," said Findlay. "The technology is proven, commercially available, and can scale immediately, but finding proximate transport and storage for the CO₂ may be challenging."

Looking beyond 2030, Wood Mackenzie's analysis identifies several emerging technologies that could offer competitive or superior economics, though each faces significant deployment challenges.

Enhanced geothermal systems show the most promise for cost-competitive decarbonization by 2030-2035, with projected costs as low as $61/MWh. However, the technology must still prove it can scale outside traditional geothermal regions, with only 1.5 GW currently in the development pipeline.

Nuclear restarts offer the lowest-cost decarbonized power at $155/MWh and have attracted major hyperscaler commitments. However, only 11.5 GW of shut-down capacity is available in the US, limiting scalability, and new small modular reactors face minimum 5-8-year timelines and significant cost uncertainty.

Long-duration energy storage technologies could eventually enable better integration of renewable power, but costs of $100-300/MWh at scale remain significantly higher than other options.

The report acknowledges that renewables and battery storage will continue playing an important role in grid decarbonization. However, the analysis finds that solar and wind cannot meet data centre uptime requirements without extensive gas firming—which can result in net lifecycle emissions higher than combined-cycle gas with carbon capture. Growing scrutiny of renewable energy credits further complicates the renewable pathway.

"This isn't an either-or situation," said Findlay. "Renewables will help decarbonize the overall grid, which benefits everyone. But for data centres specifically—with their massive, constant power demands—you need firm capacity. That's where carbon capture, geothermal, and nuclear come in."

The analysis also examines the competing pressures hyperscalers face as they balance rapid AI expansion with decarbonization commitments.

"The challenge for hyperscalers is managing multiple priorities simultaneously," said Findlay. "Securing power for AI growth is paramount; decarbonisation is secondary. Global data centre emissions will increase. But these companies also have exceptionally strong balance sheets and public sustainability commitments. How they address their emissions will shape the data centre power landscape for decades.”

"One advantage of carbon capture is optionality—it can be built in from day one or added later as circumstances evolve," Findlay added. "In an environment where energy policy, public expectations, and competitive dynamics remain volatile, that flexibility has strategic value."