Anthropic signage during the company's initial public offering at the Nasdaq MarketSite in New York on Nov. 19. —John Lamparski/Bloomberg—Getty Images

From a business, economics, or technology angle, the record-breaking Anthropic IPO is one of the biggest stories of the year. Raising around $81 billion, it was the biggest IPO of all time. And it’s perhaps the cleanest signal of how interwoven the economy has become with AI. 

But the Anthropic IPO also provides an important lens into this moment for energy and climate. It’s only slightly hyperbolic to say that over the last few years AI has sucked up all the capital and all the attention, forcing everything else, including climate-linked ventures and initiatives, to ride the AI wave. 

So, it’s worth looking closely at Anthropic’s pre-IPO regulatory filing. Ahead of going public, the company outlined its plan to shift toward its own dedicated data center infrastructure. Like the other AI labs, Anthropic has thus far largely relied on data centers built by separate companies—leasing space from Amazon, Google, and others. But the tense competition has pushed the company to commission its own data centers and power supply. 

The AI push from many of the big tech firms has offered a lifeline to clean technology—paying to bring geothermal, battery storage, and other solutions to market. But to meet their data center power demand they have also increasingly relied on natural gas. It’s telling that even while Anthropic provides relatively limited information about where its power will come from, it acknowledges that natural gas will play a role despite the inherent fuel price volatility that accompanies it.  

This shouldn’t be entirely surprising. The International Energy Agency’s annual World Energy Outlook, released this week, identified 180 GW of new gas power capacity planned to power data centers in the U.S. with much of it coming online in the 2030s. That represents about 14% of today’s total utility-scale generation capacity. 

But it is striking that Anthropic left the issue relatively unaddressed. The company has built its entire image around using AI responsibly, and it was founded as a public-benefit corporation, which means that it considers how to better society alongside its profit motive. With that in mind, the company is careful in its filing to highlight its efforts to address public concerns about how its work may affect electricity prices. It lays out mechanisms to reduce strain on the grid. And the company notes the various regulatory pressures they face around the world related to climate change, such as disclosure laws in Europe and California. But it says little about how it might respond to concerns about the climate impact of its work, and it provides no insight into its emissions profile. The risks around gas are portrayed largely in terms of cost and availability.  

It’s possible that this proves to be a safe approach. After all, as the company notes, there is competing pressure around how to approach energy and climate (just think of the gap between governments in the U.S. and Europe). Moreover, at this moment, public attention is tuned more toward affordability than emissions. 

But the long-term bet on gas should at a minimum raise eyebrows for any company whose social license and, to a certain extent, competitive advantage relies on serving the public interest. As competition between providers of AI compute intensifies, companies will need to compete across a variety of fronts. In some jurisdictions, like Europe, emissions are increasingly a factor. In others, expect it to grow as a consideration.