Anthropic, the AI startup that spent the last two years buying renewable energy and staying quiet about its climate footprint, just committed to buying carbon removal. Not pledging, not offsetting. Buying. On June 17, the company joined Frontier, the tech-industry carbon removal collective, and added its share to a fresh $915 million in advance purchase commitments, enough to nearly double the group's total pledges to $1.8 billion. The move marks something structural: the first time a pure AI company has walked into a carbon removal offtake market and signed up to pay for durable removal across decades.
Frontier was built to solve a specific problem that has haunted climate startups since 2020. Most carbon removal companies, direct air capture outfits, enhanced weathering plays, ocean alkalinity ventures, can build pilot plants for tens of millions. Scaling to hundreds of thousands of tonnes per year costs hundreds of millions more. But buyers have been reluctant to sign long-term contracts with companies that do not yet operate at scale. The chicken-and-egg gap is real. Frontier, founded by Stripe, Alphabet (Google), Shopify, Meta, and McKinsey Sustainability in April 2022, was designed to break it by making advance purchase commitments years before removal actually happens. Google is a founding member; Anthropic is the first AI lab to follow. This changes the market signal. When Google joined, it was a big-tech sustainability play. When Anthropic joins, it is a sector entering.
The numbers show Frontier is not ceremonial. Since launch, the collective has already contracted nearly $700 million across more than 50 projects to remove 1.8 million tonnes of carbon using multiple pathways: direct air capture, ocean alkalinity enhancement, enhanced rock weathering, biomass-based removal. Frontier vets each company before signing; it functions as procurement infrastructure, not a pledge vehicle. The new $915 million tranche includes offtake contracts that range from 8 to 10 years in length, with some extending as far as 2040, meaning Anthropic and other backers are betting that these startups will still be removing carbon at scale. That is not a risk-free bet, but it is a concrete one.
Here is what Anthropic's entry actually signals. The company has no published sustainability report. It has said it prefers an 'all of the above' approach to energy, meaning it will take renewable power where it can get it but will not rule out grid electricity from mixed sources. It has been building large language models and reasoning engines that consume significant power. Rather than defend that or offset it away, Anthropic is now tying itself to durable carbon removal at scale. This is not greenwashing theater, Frontier's contracts are enforceable, the tonnes are tracked, and the company's name is public. Other AI labs will see this move and face a choice: follow or explain why they are not.
The risk is velocity. Frontier has committed $1.8 billion, which sounds enormous until you place it against removal needs. Current global carbon removal capacity is roughly 10,000 tonnes per year across all methods combined. Climate models often assume removal rates in the gigatonne range by mid-century. Frontier's $1.8 billion will retire something like 1.8 million tonnes over time, useful for accountability, meaningful for the startups it backs, but a rounding error against what climate math requires. The real test is whether commitments from Anthropic, Google, and others create a new sector norm that spreads to Microsoft, Meta, and traditional corporations, turning carbon removal offtake from a niche collective into standard procurement practice.
Watch three things. First, whether Anthropic publishes its first sustainability report within the next 12 months and what energy and removal targets it commits to. Second, whether other major AI labs announce similar Frontier memberships, the competitive signaling effect matters more than any single company's decision. Third, whether Frontier's current $1.8 billion in commitments translates into actual tonnes removed on schedule, and whether the carbon removal startups it backs can hold down costs as they scale. If removal companies can build plants for under $200 per tonne by 2028, the economics of corporate offtake become sustainable. If costs stay above $300, Frontier's purchases remain a commitment device for early movers, not a market driver.