Anthropic 上市文件披露 AI 数据中心对天然气的依赖

As it prepares to go public, Anthropic’s S-1 filing makes generative AI’s methane dependency glaringly obvious.

“Competition for access to electrical supply and data center locations with adequate power availability is intense.”

In documents filed with the US Securities and Exchange Commission Monday, Anthropic acknowledges that data centers are becoming increasingly reliant on “natural gas” — the fossil fuel industry’s marketing term for what is primarily methane — leaving the company’s AI data center buildout vulnerable to rising fuel costs and existing supply chain constraints. And in a possible catch-22, Anthropic admits that rising global temperatures — a symptom of burning fossil fuels — will increase the company’s operating costs.

“Competition for access to electrical supply and data center locations with adequate power availability is intense,” Anthropic says, adding that some of its planned data centers rely on electricity generated by natural gas turbines on-site. Anthropic leases capacity from hyperscalers, but it announced a $50 billion dollar plan last year to expand data centers in Texas and New York.

The US, the epicenter for data centers globally, is planning to build more fossil fuel generation capacity than it has in a decade as data centers demand more electricity. Donald Trump’s administration has sought to kill renewable energy projects while tilting the regulatory landscape in favor of fossil fuels, which Trump claims is a more important energy source for powering data centers around the clock than solar and wind farms that generate power intermittently.

Now, a shortage of gas turbines is slowing data center construction. Some developers are installing gas engines on-site as a way to avoid long wait times to connect to power grids. Developers are also turning to more simple turbine technology at these sites as they face backlogs from turbine manufacturers, making on-site power generation less efficient and more polluting as a result.

Data center electricity demand could wipe out decades of gains in air quality over the next five years, one recent Texas-based study found — raising levels of soot, smog-forming pollutants, and planet-heating carbon dioxide. Even as data centers create more greenhouse gas pollution, Anthropic’s S-1 points to climate change as a risk to its bottom line. It will have to pay more for electricity and water to cool servers at data centers as temperatures climb.

Nevertheless, Anthropic expects increased competition for natural gas and pipeline capacity from hyperscalers and cloud service providers rushing to build their own data centers. “This competition could drive up the price of natural gas,” the company notes.

Gas makes up the biggest chunk of the US electricity mix, 40 percent of the nation’s power generation. And Americans are already feeling the pain of rising costs as utilities scramble to upgrade energy infrastructure and meet rising power demand.

Pissed-off residents have successfully blocked dozens of proposed data centers across the US. Anthropic acknowledges that the specter of ballooning utility bills is enough to interfere with a data center project. “If our efforts to secure electrical power are perceived to affect other utility customers and the broader grid, regulators may impose conditions, standby charges, or other requirements intended to address such impacts,” it says.

If regulators push data center developers to shoulder more of the costs for new energy infrastructure, Anthropic says that could also raise its operating costs. And without a secure gas supply and the pipelines to transport the fuel, the company says, it might need to delay or abandon planned data center projects altogether.

Ultimately, these challenges could stunt Anthropic’s growth:

Any failure to secure access to adequate electrical power, any legal, regulatory, or other administrative action or change limiting the availability of electricity, or any significant increase in energy or compliance costs, could adversely affect our cost structure and limit our ability to scale our compute infrastructure, develop and deploy our models, and service our customers.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

• Justine Calma

原文版权归 The Verge · AI 及原作者所有，本站转载仅供阅读。如需删除，请联系 support@aisa.one。