Market Data 
- AI has touched off a data center development boom that will be largely powered by natural gas.
- Fearful of becoming big polluters, big tech has begun to explore the use of carbon capture.
- The technique of capturing and burying carbon emissions is still unproven and nascent.
Big tech companies like Amazon, Meta, Microsoft, and Google have promised to use renewable or low-carbon energy sources like solar, wind, and nuclear to power a record surge of data center development across the country.
Yet much of the around-the-clock electricity needed for the energy-hungry facilities has been — and will likely continue to be — generated by fossil fuels, including a fleet of new and existing natural gas plants.
To try to meet their environmental goals, these tech companies are considering the emerging business of capturing carbon at an industrial scale before it is released into the atmosphere and funneling it permanently underground.
Exxon Mobil, which recently announced an initiative to power data centers with plants whose carbon emissions it will sequester, has estimated that data centers will account for 20% of the carbon capture and storage market by 2050.
In December, Meta became the first big name in the industry to embrace the tech when the large southern utility Entergy announced that the company had agreed to fund a large carbon capture and storage project in Louisiana. The deal will help Meta offset emissions from a $10 billion data center campus it is planning in the state that will be powered by three new natural gas-fired plants.
Last month, Chevron announced that it would build natural gas plants using carbon capture, citing the soaring demand from data center customers.
The concept of piping carbon emissions underground is not new. Oil and gas extraction companies have pumped compressed, liquid carbon dioxide into wells for decades to help force out more fossil fuels. Using the technique at a scale large enough to begin significantly mitigating climate warming emissions, however, has been considered both enormously costly and complex.
A data center boom, generated by the race to develop and commercialize artificial intelligence, has suddenly brightened the outlook. Leading data center builders include the deepest-pocketed companies in the world. Some of them appear willing to pay the kind of hefty premiums for emission-free electricity that would justify the costs of capture and storage.
“We had an ‘aha’ moment that it’s actually Amazon, Microsoft, Meta that are very interested in these power plants,” said Jerry Ashcroft, the CEO of Crescent Midstream, which was selected to design and build the carbon capture system for the Louisiana plant that Meta will help pay for. The project, at the 1-GW Lake Charles power station, will cost around $1 billion to build, Ashcroft said, who noted that a final cost estimate is underway.
Crescent Midstream, meanwhile, is bidding to develop another utility-scale carbon capture system for a new gas plant that Entergy plans to build in Greenville, Mississippi, Ashcroft said. That station will help power a $10 billion data center campus nearby that Amazon announced in early 2024.
The investment firm Carlyle Group acquired a majority stake in Crescent Midstream, a pipeline company, in 2021, and expanded it into the business of capturing, transporting, and sequestering carbon.
Ashcroft says that he anticipates Entergy will pursue at least 10 carbon capture projects across its portfolio in the near term. The Houston-based executive, who has been in the fossil fuel industry for years, said he has been astounded carbon capture’s rapid recent advance.
“Going from 2020 to 2025, where now you’re actually doing a project, yeah, it’s pretty hard to imagine,” he said.
Soaring energy demands and few carbon-free solutions
Behind the interest in carbon capture lies an increasingly difficult array of choices in the nation’s constrained energy market.
Some forecasts estimate that more than 90 gigawatts of data center demand could be added to the grid nationally by the end of the decade—the equivalent of about nine New York City-size blocks of power.
Data center operators have sought out renewable sources of energy to curtail the enormous potential carbon footprint of all that capacity. Still, green options, including solar and wind, aren’t considered reliable or intense enough to readily match the enormous, steady strains that the data facilities place on the grid day and night.
Adding new nuclear power, meanwhile, is even more costly than capture and storage and fraught with labyrinthine regulatory hurdles.
To meet the growing shortage of power, utilities have turned to an old asset in their portfolios: natural gas.
Natural-gas-fired generation capacity in the US has grown to 79 gigawatts, an increase of 19% since 2014, according to the US Energy Information Administration. Even as more solar and wind is connected to the grid, surging energy demand from data centers is expected to drive more natural gas use. In Virginia, Georgia, North Carolina, and South Carolina, for instance, utilities are planning to build a combined total of 20 GW of new natural gas generation, primarily as a result of forecast data center demand.
“If the business model is we have to get the most power, quickest to stay out in front of the AI race as it were, gas must play a role in that,” said Erik Belz, the president and chief operating officer of the investment firm Engine No. 1.
Engine No. 1 is partnering with Chevron in its initiative to develop natural gas plants with a total capacity of about 4 GW for energy customers constrained by the power shortage, including data centers. The group also includes the energy equipment manufacturer GE Vernova, which is producing the turbines for the power plants. The venture plans to couple the facilities it builds with renewable energy sources and carbon capture to reduce their emissions. It expects to begin delivering the projects in 2027.
The group is close to announcing its first deal with a data center customer, and that project may seek to harness carbon capture, Belz said.
“This one has the capability to do it,” he said.
A web of technical hurdles
Carbon capture and storage come with myriad complications and uncertainties.
About 6% of the exhaust fumes from a modern natural gas plant are composed of carbon dioxide, according to Ashcroft, the Crescent Midstream CEO. Huge absorbers filter that gas out before compressors liquify it. Pipelines transport the carbon to wells where it is pumped thousands of feet below the surface.
The Lake Charles power station, where Crescent Midstream is planning its initial capture and storage project, is located on Louisiana’s Gulf Coast, the seat of America’s petroleum industry where there is an abundant network of pipelines and wells.
Ashcroft said the company would shove the gas as much as 10,000 feet below ground into saline beds beneath the gulf that lock in petroleum deposits and liquefied gas.
That geology and infrastructure don’t exist widely across the US. It’s not clear, for instance, whether carbon emissions could be as readily sequestered in northern Virginia, the country’s largest data center market. Researchers at Virginia Tech are currently conducting a feasibility study on industrial carbon capture in the region, funded by the US Department of Fossil Energy and Carbon Management.
There are also questions whether the gas, once funneled below the earth will reliably remain there forever or if it could eventually seep out into the atmosphere, negating the whole enterprise and causing environmental harm. A large leak of methane gas from a storage well in California in 2015 is considered the worst greenhouse gas disaster in US history.
There has never been a carbon capture and storage unit operating at a gas-fired power plant in North America at the size that Big Tech would need to capture emissions from an AI data center, said Dennis Wamsted, energy analyst at the Institute for Energy Economics and Financial Analysis, a nonprofit think tank that advocates for clean energy. “That doesn’t mean it can’t be done,” Wamstead noted, “but I would say it’s going to be significantly more expensive than anything you’ve seen in a press release.”
Difficult economics
Only about 51 million metric tons of carbon were captured in 2024 out of a total global emissions output of 40 billion metric tons — roughly a tenth of a percentage point, according to Pavel Molchanov, an analyst at Raymond James.
“Carbon capture was a tiny niche,” Molchanov said. “Sequestration will need to be available in a wider range of geographies, because guess what, not everybody lives next door to an oil field.”
Molchanov said that capture and storage could be expanded by adapting more geologies in areas that might not otherwise accommodate sequestration. Of course, that might also increase costs.
Capture and storage already faces difficult economics.
The parties who sequester the carbon can receive an $85 federal tax credit for every metric ton of carbon captured and stored, which can be monetized in the corporate tax credit market to produce revenue. Ashcroft said that questions have persisted in the capture and storage business whether the credit, which was roughly doubled per metric ton by the Biden administration, was sufficient to justify the steep costs.
“Many people say you need $120 to make this work,” he said.
Ashcroft said he had an epiphany when he saw Microsoft’s announcement last September that it had reached a deal to revive a decommissioned reactor at the Three Mile Island nuclear station in Pennsylvania to provide it with carbon-free power for its data center operations. Ashcroft said that the hefty electricity premiums Microsoft agreed to as part of that deal were a clear signal that data center operators were “willing to pay more than market for” decarbonized energy, including capture and storage.
If the tax credits alone aren’t enough to pay for the Lake Charles sequestration project, the deal will rely on Meta, which has agreed to pay a rider, up to a fixed amount, on its utility bill with Entergy in Louisiana to offset the costs. A spokeswoman for Meta declined to comment.
Molchanov agrees that capture and storage is poised to grow. He projects that the amount of carbon sequestered annually will quadruple by 2030, in part because of data center demand for it.
In its earnings call in late January, Exxon Mobil executives said that their carbon capture and storage operations and a related business producing low-emissions hydrogen with capture and storage technology could generate $2 billion of annual revenue by the end of the decade.
“We’ve got pretty aggressive growth plans in this space,”Exxon Mobil’s CEO Darren Woods said on the call, according to a transcript by The Motley Fool. “But again, all dependent upon customer interest and customers’ willingness to engage in long-term contracts.”
Betty Jiang, an analyst at Barclays who covers Exxon Mobil said that she believed the company would only trumpet the business of capture and storage if it had received interest from customers.
“We’ll stay tuned on exactly what these investments look like,” Jiang said. “But I don’t think they will be raising their capital in these areas and saying what they’re saying about the market opportunities if they are not having the conversation in the background to really give them the confidence that it is indeed an area of growth for Exxon.”
