Anthropic Is Buying Carbon Removal — But Not Clean Power

To the extent that any $965 billion artificial intelligence company built on pirated model training material can be “good-coded,” Anthropic has somehow managed to earn that reputation, at least relative to its peers. It’s somewhat surprising, then, that the company has been silent on climate change.

Until today. Sort of.

Frontier Climate, a corporate initiative to drive advances in carbon removal, announced a $915 million advance market commitment growth fund on Wednesday, naming Anthropic as one of the participating buyers.

Frontier supports projects that are capable of sucking large amounts of carbon out of the atmosphere, a solution scientists say is a critical supplement to reducing emissions in order to curb climate change. With the new fund, Frontier is shifting its focus from supporting early innovation to taking bigger swings on fewer, larger projects. Anthropic, alongside Google, Stripe, Shopify, and others, has committed to co-sign offtake agreements to buy the resulting carbon removal.

The news throws into relief Anthropic’s nearly complete absence from the clean energy development picture. The company’s primary contribution to climate change is its energy consumption, which is driving up coal and natural gas-fired power generation. According to data shared with Heatmap by the market intelligence company Cleanview, the average carbon intensity of Anthropic’s data centers is among the highest of its competitors, second only to xAI. Yet unlike many of peers, the company has not announced a single clean power purchase agreement to date.

Anthropic’s reputation as the ethical AI company traces back to its origin story, which begins with a guy leaving OpenAI to build a company more committed to AI safety. That guy, Anthropic CEO Dario Amodei, speaks and writes openly about the risks to humanity posed by powerful AI. Anthropic has also donated millions to support the development of AI regulations and prohibited the use of its models for mass surveillance or autonomous weapons, putting it at odds with the Trump administration. The company has focused on text-based products, in part to avoid the risk of users creating child sexual abuse material.

To date, however, the company has not publicized any sustainability strategy, nor has it published an annual sustainability report. It has not made any public commitments to use clean energy or reduce emissions. It is not a member of the Corporate Energy Buyers Association, a trade group representing companies that buy emissions-free energy. The only mention of any of the above themes in the company’s “Transparency Hub” is a note that many of its customers use Claude, Anthropic’s AI model, to “increase public health, education, environmental sustainability, and societal benefits.”

To be fair, it’s not that Anthropic has never discussed clean power. In a July 2025 report titled “Building AI in America,” the company made recommendations for ensuring the U.S. can support a competitive AI industry. It advocated for an “all of the above” approach to power generation to meet AI demand in the near term, which would “maximize opportunities for AI to catalyze emerging energy technologies, such as next-generation geothermal and advanced nuclear” down the line. It endorsed permitting reform to speed up transmission development and called for increased domestic production of electrical grid equipment.

In a section on the use of federal lands, the report also made a subtle dig at the Trump administration’s discriminatory policies against wind and solar. It noted that “solar, batteries, and geothermal may prove the most economically efficient choices before advanced nuclear power comes online,” and that “limiting developers’ opportunities to procure some power sources but not others” could make American AI “less competitive in a period of global competition.”

From one perspective, it makes sense that Anthropic hasn’t gone out of its way to procure clean power. To date, the company has mostly leased data center capacity from other providers that do have clean power commitments, including Amazon and Google. That will soon be the case no longer, however, as it is planning to both build its own data centers and rent capacity from xAI’s Colossus data centers, which rely heavily on power from on-site natural gas turbines. Colossus is currently the subject of a lawsuit filed by the NAACP over its air pollution.

Anthropic also doesn’t need to own and operate its own data centers to assume responsibility on climate change. Jane Flegal, a senior fellow at the think tank the Searchlight Institute, argued in a recent paper that companies should forget trying to minimize their individual carbon footprints and just make the most high-leverage investments they can, whether that’s helping to finance a geothermal power plant or a transmission line or a new transformer for the grid.

Anthropic did not respond to my inquiry for this story, but there’s some evidence to suggest that the company may be starting to take on climate and clean energy beyond the Frontier deal.

In March and April, Anthropic made three new hires to lead its energy strategy who all have a background in clean power. Ariel Horowitz is the company’s new data center energy lead. She previously spent five years at the Massachusetts Clean Energy Center before becoming the deputy director of grid modernization at the federal Department of Energy during the Biden administration. Sana Ouiji, who spent six years at Google working on data center clean energy strategy, is one of Anthropic’s new energy leads. Another new energy lead, Andrew Rudersdorf, came from roles sourcing energy for Meta’s data centers, including renewables.

The company is also currently hiring for a director of infrastructure and energy accounting, and looking for someone with “experience accounting for energy contracts — Power Purchase Agreements, Virtual PPAs, Renewable Energy Credits, or similar commodity arrangements,” according to the job listing.

Anthropic also appears to be preparing for mandatory emissions reporting rules that large companies will soon be subject to in California and the European Union. In April, the company hired Chris Power, who previously worked in sustainability reporting for Amazon and Salesforce, as its new head of non-financial reporting and strategy, according to LinkedIn. In a post announcing his new job, Power said part of his role would be building out the company’s sustainability reporting capabilities.

While funding carbon removal through Frontier is a major step forward for Anthropic on climate, the company is sure to face criticism over its order of operations. Scientists largely agree that carbon removal is an important solution for down the line, but only if the world also dramatically reduces the amount of carbon it emits in the first place — not least because doing so is less expensive and less resource-intensive than removing emissions in the future.

My colleague Robinson Meyer had Hannah Bebbington Valori, the head of Frontier, on his podcast Shift Key this morning, and asked her whether Anthropic is an example of the common concern that the potential to remove carbon from the atmosphere in the future could be used to delay cutting emissions today.

Bebbington Valori didn’t comment on Anthropic specifically. But she did say that most of the companies buying carbon removal with Frontier and otherwise do have broader climate programs. She also noted that buying carbon removal from Frontier is not a “get out jail free card,” since it costs hundreds of dollars per carbon credit, and that in general the world is spending a lot more money on decarbonization than carbon removal.

“And then, you know, the other way to answer this question,” she added, “is we should hold folks’ feet to the fire on this. People who buy carbon removal, people who don’t buy carbon removal, should be thinking about decarbonizing their emissions.”