Image source: unsplash
In an era when “deglobalization” is gaining momentum, if there are any issues that continue to attract global attention and debate, artificial intelligence and climate change undoubtedly rank among them. These topics touch on human societal development, great-power competition and cooperation, as well as people’s lifestyles and living conditions.
And now, the connection and contradictions between these two major issues are also being rapidly brought to the forefront.
On October 26, local time, U.S. modeling giant Anthropic filed its IPO application with the U.S. Securities and Exchange Commission. The prospectus noted that AI companies are facing increasingly fierce competition for electricity and data center site selection, with some planned data centers relying on on-site natural gas turbines for power. Although this statement did not attract much attention in the investment or technology sectors, it has once again become a hot topic in the climate and environmental arenas. Justine Calma, a senior science reporter at The Verge who has long tracked climate and energy issues, promptly wrote an article pointing out that the expansion of AI computing power is forming an “alliance” with fossil fuels, which will clearly exacerbate the climate crisis and lead to more severe air pollution locally.
In addition, Anthropic explicitly listed climate change and environmental issues as risk factors in its prospectus. This has sparked further controversy: Lionella Pezza, Research Director at Domini Impact Investments, notes that climate and environmental considerations are critical in today’s investment analysis and decision-making, yet Anthropic’s “general disclosures” on these matters are overly vague. Both investors and the public would prefer—and rightly expect—to see quantitative data on AI’s impact on greenhouse gas emissions, energy consumption, and water usage.
The climate and energy media outlet Heatmap News points out that, given their significant impact on the future of humanity, AI companies should not only disclose environmental risks but also inform the public whether they are fully leveraging the vast resources at their disposal to seize opportunities for mitigating climate change.
Notably, the aforementioned discussions took place at a time when AI development is accelerating, “alignment” risks have come into sharp focus, and market volatility is intensifying—just ahead of the most important annual event in the climate arena: the COP (United Nations Climate Change Conference). At COP31, scheduled to be held in Turkey from November 9 to 20, the environmental costs of AI are also being treated as a key issue. The host country, Turkey, has previously stated that it will work with all stakeholders to explore ways to advance “climate‑responsible AI.”
When Climate Change Meets AI Transformation: The “Net Effect” Is More Harmful Than Beneficial
“The Earth’s climate system is rapidly becoming unstable.”
On October 14, an international research consortium led by the University of Oregon stated in its latest annual climate report that seven of the 16 climate indicators it tracks have reached record-breaking extremes, while other indicators are also trending toward extreme levels. Meanwhile, the rate of global warming has accelerated from approximately 0.19°C per decade to about 0.36°C per decade.
On October 21, a report released by the United Nations Environment Programme (UNEP) also indicated that, based on current emissions reduction commitments and progress, global temperatures could rise by as much as 2.8°C by the end of this century—far exceeding the 2°C target set in the Paris Agreement adopted at the end of 2015. Achieving the more ambitious goal of limiting warming to 1.5°C would be even more challenging.
In recent years, numerous reports and statistical data in the environmental and climate fields have sounded the alarm, stating that climate change has reached its most critical juncture. Meanwhile, AI is seen as a new source of instability. The aforementioned UNEP report notes that although data centers currently account for less than 1% of global emissions, their electricity consumption is growing at an extremely rapid pace; by 2030, data center power usage is projected to double compared to this year. Consequently, relevant reports have identified the energy demand and associated emissions driven by AI growth as one of the most pressing trends requiring close attention in the fight against climate change.
Golestan Sally Radwan, UNEP’s Chief Digital Officer, stated that AI has the potential to drive human societal development, but its impact on global warming and environmental pollution “should also raise concerns.” She further emphasized that, before AI is deployed at scale, it is essential to ensure that its “net effect” on the planet is positive.
Meanwhile, in the highly contentious climate debate, “net effect” is emerging as the new “center of contention.”
Many perspectives note that AI is reshaping the paradigm of scientific research, with climate science and the development of clean energy set to benefit. Demis Hassabis, co-founder and CEO of Google DeepMind, views AI as a fundamental technological solution to the climate crisis, believing that with AI’s support, breakthroughs will be achieved in climate modeling, weather forecasting, and even in emerging energy technologies such as photovoltaics, hydrogen energy, and nuclear fusion.
However, the climate science community is far from optimistic about this.
According to renowned climatologist and foreign academician of the Chinese Academy of Sciences, Chen Deliang, AI has indeed brought benefits to climate science in research; his own team has also extensively applied AI in areas such as hydrological modeling, extreme-event and disaster early warning, and climate-model forecasting. However, overall, the carbon footprint of AI development is far greater, and at least for now, the research gains fall far short of offsetting its increased emissions. As a result, its net impact on the climate crisis remains more harmful than beneficial.
In his view, the core intersection between AI and climate change lies in energy. “AI consumes enormous amounts of power—it’s a real electricity hog—and at the current pace, this trend will only accelerate.” The issue of electricity shortages caused by AI in the United States has been on the table for some time: AI companies are building their own power plants and relying heavily on fossil fuels. Coupled with current policy directions, this is bound to hinder progress on carbon reduction. Even in China, where “electricity is not scarce,” the massive power demand generated by AI development will still put pressure on the deployment and integration of new energy sources.
Since September, the “alignment” issues and risks of runaway AI development have drawn widespread attention, with OpenAI, Anthropic, and others even proposing to temporarily halt AI progress to mitigate these risks. In the climate and environmental sectors, many voices have also pointed out that AI’s energy consumption and potential environmental impacts may be critical factors that have been overlooked amid the broader array of risks—and that these effects could emerge sooner than other risks and prove far more difficult to reverse. Yet, at present, both AI model research and development and investment in data center construction appear to be forging ahead unabated; many skeptics argue that the so‑called “slowdown” is likely nothing more than a strategic move in commercial competition.
Chen Deliang also believes that the business community is unlikely to voluntarily slow down its AI development, because if one company slows down, another will overtake it, potentially making the next round of the race even more intense. And when AI competition becomes unstoppable, the resulting problems could become very serious.
A researcher from Tsinghua University’s HPCA team also noted that the team has indeed achieved numerous breakthroughs in climate modeling by leveraging AI, including integrating AI with traditional numerical simulations to develop a kilometer-scale Earth system simulator. This has dramatically reduced the time required for weather–climate evolution simulations: natural-year simulations that once took weeks or even months can now be completed in just one day, while forecast accuracy continues to improve. However, he also pointed out that such research can help mitigate disaster losses and enable the public to more intuitively and quickly grasp the current state and trends of climate change. Nevertheless, when it comes to addressing climate change itself—particularly emergency responses to climate-related disasters—AI is currently unable to provide core solutions.
He also noted that as AI becomes increasingly widespread, its energy consumption has drawn growing attention in both daily life and the workplace. For many students and colleagues who are concerned about climate and environmental issues, using AI can evoke a sense of ambivalence—much like the trade-offs once faced when driving gasoline-powered cars or flying: balancing carbon reduction with progress and convenience.
When AI Development Meets the Climate Crisis: Risks Still Being Overlooked
As the scientific community begins to extensively explore how to address AI’s profound impact on the climate, the reverse effect—how climate influences AI—has also come into focus.
Anthropic explicitly stated in its prospectus that rising temperatures will increase the electricity and water consumption required to cool data center equipment, thereby driving up operating costs. Consequently, regulatory pressures and NIMBY movements are more likely to slow down data center construction and even harm the company’s reputation.
Many in the investment community have also weighed in on this issue. Blaine Townsend, Head of Investments at wealth management firm Bailard, pointed out that AI-driven investment risks have become a core financial and fiduciary concern—particularly climate and environmental risks, which have long been overlooked and treated as peripheral issues. This, he argues, is clearly detrimental to investment decision-making and the industry’s overall development.
Luo Nan, Head of China for the Principles for Responsible Investment (PRI), stated that the increasing frequency of extreme weather not only affects traditional sectors such as agriculture but also poses a direct challenge to AI, with its supply chains and energy systems facing ever‑more severe climate risks—risks that most investment frameworks have yet to account for or incorporate.
In fact, statistics released by Zurich Insurance in June this year showed that, over the past three years, extreme weather has become the leading source of losses in its U.S. data center construction insurance portfolio, accounting for one-third of all claims. A report published by First Street that same month further noted that 79% of global data center computing capacity is exposed to climate hazards such as flooding, extreme heat, high winds, and wildfires, while 54% of data centers face “chronic” climate stresses like extreme heat and drought—risks that continuously reduce energy efficiency and drive up operating costs. Moreover, climate risks overlap significantly with regions experiencing rapid industrial growth; in particular, chronic climate risk exposure in the Asia-Pacific region reaches 89%, the highest globally.
Meanwhile, with several strong El Niño years—including this one—on the horizon, these disruptions are becoming increasingly pronounced. Assessing the impacts of heavy rainfall, extreme heat, droughts, and other climate-related phenomena on data center operations, upstream manufacturing supply chains, energy supplies, and labor markets has emerged as a pressing new challenge. In the United States and elsewhere, protests by local communities over climate, environmental, and quality-of-life issues are further introducing unpredictable ripple effects. Many organizations have also warned that climate risks and their cascading consequences could slow the pace of AI data center construction and hinder the large-scale deployment of AI itself.
For investors, the aforementioned issues are particularly critical.
Currently, the global insurance industry faces inadequate coverage for climate-related losses and disaster payouts. Wang Zhongyao, General Manager of China Re P&C, has pointed out that natural disaster insurance coverage in the Asia-Pacific region stands at only about 10%, leaving a “massive protection gap” amid rising risks from severe typhoons and floods. Even in developed countries in Europe and North America, insurance coverage for natural disaster losses rarely exceeds 40%. Meanwhile, in the AI sector—where development is faster and market influence is concentrated among major corporations—investment in risk mitigation remains conspicuously insufficient. Florian Krampe, Director of Peace and Development Studies and Director of the Climate Change and Risk Program at the Stockholm International Peace Research Institute, has noted that markets lack proper pricing for climate risks, and throughout the lengthy lifecycle of data centers, risks stemming from natural disasters are largely borne by investors—posing a potential hidden threat to the future of AI development.
AI’s energy consumption will exacerbate the climate crisis, while climate issues, in turn, are likely to impact AI’s own development.
As Chen Deliang noted in his research on the two-way interactions between climate and energy, the relationship between climate and industry is far from unidirectional; rather, it forms a closed-loop of mutual influence, with each element’s feasibility, vulnerability, development trajectory, and scale continuously shaped—and even determined—by the other. In the face of such complexity, humanity must shift its response from post-crisis “passive recovery” to proactive “active transformation.”
He also noted that the scientific community is increasingly hosting research studies and conferences focused on the impact of AI on the climate. With AI’s ability to actively slow down being severely limited and its long-term climatic effects growing ever more urgent, the immediate priority is to forge a consensus on managing climate risks and advancing renewable energy development. (This article was first published on the TMTPost app; author: Feixiang Tai Kong.)
For more exciting content, follow TMTPost on WeChat (ID: taimeiti), or download the TMTPost App.