The sustainable AI development contradiction: Chinese policy on sustainability and the costs of AI development

Abstract The ecological cost of developing artificial intelligence, particularly the cost caused by data centers, has been detailed by many scholars, but few have outlined what this implies for specific countries. This review paper brings together reporting on the forecast of ecological harm to China from various forms of industrialization, including climate change, and the ecological impact of AI development. The author suggests that relying on “AI for sustainability” as a solution to ecological problems is a type of “moral hazard,” borrowing a metaphor from economics, because treating AI as insurance against climate change may inhibit governments from taking necessary action now to move toward negative carbon emissions. Considering the situation of China, the author contends that the actual environmental toll of AI presents a significant ecological problem. Three interconnected ecological problems are reviewed: AI’s carbon emissions, AI’s water costs, and ecological devastation tied to AI hardware production and disposal. Together with forecasts from the World Meteorological Organization’s, this bodes poorly for the future of China. Using the political philosophy of the Communist Party of China, the author argues that it is in the stated interest of the People’s Republic of China to pursue limited AI with caution, fully attuned to the increased harm that industrialism poses to the People, and oriented toward a truly ecologically harmonious future. 1 Introduction: the contradiction of AI sustainability China, like many countries, simultaneously wants to have its cake and eat it, when it comes to environmental policy. Perhaps this is inevitable for the communist country, which defines its planned direction in terms of “principal contradictions.” A Marxist dialectical view of history entails that social progress involves overcoming such contradictions. Mao’s economic plan for the country initially involved reconciling the “contradiction between the people's demands for an advanced industrial country and the reality of a backward agrarian country, and between the people's need for rapid economic and cultural development and the current situation in which the economy and culture cannot meet the needs of the people” (Zhou 2022, p. 147). After decades of industrialization, the contradiction has evolved. Under Xi Jinping, the principal contradiction facing China is articulated as the “contradiction between unbalanced and inadequate development and the people’s ever-growing needs for a better life” (Xi 2017, p. 9–10). Resolving this contradiction is essential to China’s planned advancement, and the worsening environmental crisis highlights perfectly this tension. As outlined in the Fourteenth Five-Year Plan, the goals of the nation include both leading the world in combating climate change and in advancing artificial intelligence. Under “Priority Actions,” the Communist Party emphasizes “Actions for Constructing a Green and Smart Ecological Civilization” and promoting “the profound transformation of ways of production and life, and [assisting] the realization of carbon peaking and carbon neutrality objectives” as major tasks (CPC 2021, pp. 51, 32). At the same time, the document is oriented entirely around large-scale construction of digital infrastructure and an overhaul of the economy to be primarily “digital,” with artificial intelligence as a key technology for the Communist Party. The two goals are directly put into conversation as they seek to “promote the profound convergence of the Internet, big data, artificial intelligence, etc., with all industries [and] forcefully advance the coordinated transformation of industrial digitization and greening” (29).Footnote 1 But can these contradictions be resolved without compromise? From my experience working in the Hong Kong Special Administrative Region of China, I perceive that many Chinese leaders do not even see these aims as a contradiction. The Research Grants Committee of the University Grants Committee of Hong Kong, for example, which takes much of its lead from national policies set in Beijing, awards grant money for “Areas of Excellence” related to both aims. For their 2026–2027 call for applications, the RGC lists “Developing a Sustainable Environment” and “Big Data and Artificial Intelligence” as areas of special interest for research funding (RGC 2025). None of these are articulated as oppositional, but rather as twin aims for the same broad goals. Many other countries are pursuing these aims, albeit without acknowledging the dialectical contradiction to be resolved. The European Union wishes to promote AI development while simultaneously combating climate change. Korea and Japan wish to be among the AI elite while trying to keep up with IPCC goals. Perhaps the US stands alone as a country single-mindedly pursuing one without the other (although, as the past decade should remind us, this all depends on the caprice of whomever sits in the Oval Office). Perhaps a part of the reason why these contradicting goals are so often pursued together is because of the ideology of sustainable development. The United Nations’ Sustainable Development Goal 9 to “build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation” is meant to be seen as complementary to SDG 13 to “take urgent action to combat climate change and its impacts.” Noble as the United Nations’ SDGs are, they reinforce the belief that not only can industrialism be not ecologically harmful, but even that it can be somehow commensurate with ecological conservation. Our institutions aim at both increased industry and ecological harmony. A further layer is added to this collective confusion in the concept of “AI for sustainability” (van Wynsberghe 2021). In this view, not only is AI not bad for the environment, but it is possibly good. Indeed, according to the most enthusiastic advocates of this model, such as former Google CEO Eric Schmidt, it is the only solution to climate change (Niemeyer and Varanasi 2024). Indeed, many AI leaders are plainly demanding an increase in electricity consumption and trying to convince the general public that this is good (Dastin 2024; Anthropic 2025). The claim, then, is somewhat akin to chemotherapy: current worsening of the environment (e.g., through increased carbon emissions for energy production) for the pursuit of AI is necessary so that AI can solve the climate crisis. We make the planet sicker so that it can heal. I contend that this is not only a fantasy, but a potentially suicidal one at that. Worse than most corporate efforts at “greenwashing,” this approach makes under supported promises while dismissing the real problems inherent in the growth mindset. In truth, this problem ought to be understood and approached as a problem of dialectics. The contradictions inherent in industrialization and ecological care are not easily solved without significant tension. While some American AI proponents may feel no qualms about championing AI for ecological problems, Chinese leaders have more at stake in this matter and ought to practice greater wisdom. The “unbalanced” or “inadequate” development that results from inattention to environmental needs ought to be kept firmly in full view as AI is pursued for “the people’s ever-growing needs for a better life.” Thus, I argue in this paper that AI cannot be seen as a miracle cure for China’s ecological future. To begin, I address the problem of AI as a moral hazard: hope that AI will fix the climate crisis functions to excuse not taking actual efforts to combat climate change. Following this, I note why AI’s impact on the environment is a special concern for China, as well as Asia more broadly. Then, attending to the official political ideology of the People’s Republic of China, I assess the contradiction between developing advanced AI and preserving the environment, especially as it impacts the people. Part of the challenge of resolving this contradiction, I contend, entails avoiding the ideological pitfall of the Western techno-capitalism that treats AI as the only solution to climate change. 2 AI for sustainability as a moral hazard It is at this point well-known that AI has a significant ecological impact. Some specifics of this impact are detailed in the following section and sub-sections. However, there are advocates who justify AI development on the premise that somehow, AI could eventually lead to breakthroughs in addressing the climate crisis. For example, Aylak (2025) demonstrates how agentic AI could be applied to logistics with a possible 30.3% reduction of carbon emissions in the global supply chain. There is also evidence of a 2% energy reduction by corporate firms that have adopted AI in China over a period of 12 years (Zhou and Bu 2026). AI is also improving renewable energy availability in remote areas through microgrid systems (Vermaa et al. 2025). At the same time, the gap between the potential for AI to help on climate change and the harm entailed in its development should be noted. We may think of the hope that AI will make up the difference here as a moral hazard. As noted by Corner and Pigeon, moral hazard is a concept derived from economics wherein it refers to a situation when “an individual or party is insured against a particular undesired outcome, Fernandez (2025) so they may feel protected against the undesired outcome and therefore take greater risks” (2014, p. 2). As a public policy concern, this means a government may be disincentivized to attend to urgent national concerns if they believe some other policy provides an insurance against the urgent problem. To illustrate this problem, consider the parallel case of geoengineering, which Corner and Pidgeon examine. Geoengineering is any large-scale interference with the makeup of the planet, whether that entails transforming terrains or interfering with the ocean or atmosphere. In general, discussion of geoengineering for climate change today focuses primarily on manipulating the atmosphere. Two major proposals for geoengineering have been suggested on this front: deploying particulates in the upper atmosphere to reflect sunlight back into space, and removing and sequestering carbon dioxide on a large scale (Lawrence et al. 2018). Both require significant efforts to be effective, and both have anticipated harmful side effects. The first option—deploying particulates in the atmosphere Braakmann-Folgmann (2023)—is usually seen as a stopgap measure to cool the Earth’s surface immediately, while the second solution—sequestering carbon dioxide—has a slower impact but addresses the root problem of global warming, namely the accumulation of greenhouse gases (GHGs) in the atmosphere. There are a number of moral arguments against the use of geoengineering, including uncertainty and disproportionate costs and benefits (Geoengineering Monitor 2018), but one of the most common arguments is the moral hazard argument (Corner and Pidgeon 2014). The climate crisis requires us to reorient our way of living and drastically reduce our consumption of Earth’s resources. The problem of climate change is a problem of overexploitation. Geoengineering appears to be an insurance policy against the ecological repercussions of industrialism, allowing us to avoid the difficult work of combating carbon emissions. Critics of geoengineering, such as Geoengineering Monitor, worry that rich nations will deploy geoengineering, offload the costs onto poorer nations, and decide not to address the problems of consumption and industrialization. This need not be the case; so, some advocates have contended that geoengineering should be part of the solution and not the entire solution. However, because it offers a stopgap solution, it presents itself as a moral hazard insofar as it could be “perceived by the public as offering a license to continue carbon-intensive lifestyles [thus resulting in] the ‘rebound’ associated with geoengineering technologies [to be] even more significant” (Corner and Pidgeon 2014, pp. 2–3). Let us consider AI for sustainability as a moral hazard. As noted above, AI proponents suggest that the technology could open new solutions to fighting climate change. The techno-solutionism we see in geoengineering is paralleled in AI for sustainability and coupled with a techno-optimism that assumes all current and future problems will be solved by newer technologies. This is morally hazardous insofar as (1) AI’s development is currently exacerbating the climate problem, and (2) the reason to suppose AI will find a solution more efficiently to fix climate change than we already have is not substantiated. The perception that AI may potentially help us (or worse, the fantasy that AI alone will) address problems of climate change may lead us to not carry out the difficult work of actively combating climate change because of the perception that AI will solve it for us. An attitude of techno-solutionism allows the “AI for sustainability” narrative to obscure the actual various impacts of AI on the environment (Fernandez and Wiese 2025, Brevini 2023). While there are many promises about how AI can be part of national policies for sustainability, much of this has not been concretely implemented (O’Connor et al. 2024). More alarming, the convergence of transhumanism, effective altruism, and longtermism embraced by many tech leaders promotes the development of AI above fighting climate change as a more important “existential risk” (Gebru and Torres 2024). To wit, the problem of climate change is as straightforward as: emissions of carbon dioxide and other GHGs are heating the planet. The only real, long-term solution amounts to lowering the amount of GHGs in the atmosphere. Even the overly optimistic goals of the Intergovernmental Panel on Climate Change reflect this: the nations of the world should attain carbon neutrality by 2050 and negative carbon emissions (i.e., active removal of carbon dioxide) by 2070 (Rogelj et al. 2018). However, implementing this simple solution is not so easy in a world of advanced industry in some countries and national efforts to industrialize in others. To reach its own goals, China is pursuing a “dual carbon” strategy of improving energy efficiency across the country while simultaneously de-carbonizing its energy production (Xinhua 2025a). Although the need to de-carbonize is the only real solution for combating climate change, many countries see AI as part of the solution. Proposed applications at present include: tracking weather patterns to prepare populations for disaster (Cirri 2023), simulations of climate change (Tripathi et al. 2024), better crop yields with changing climates (NIFA 2025; Rowe 2025), forecasting arctic ice changes (European Space Agency 2023), reducing carbon emissions for transportation (Iyer 2021; Aylak 2025) and energy (TEC 2024; Vermaa et al. 2025), beter materials construction (Martínez et al. 2026), and so forth. The range of these proposals run from helping us prepare for disaster to potentially slowing the disaster as it comes. Undoubtedly, these improvements can be an important part of our sustainability solution. However, none of these proposals amounts to positively combating climate change. None of these solutions address the problems of increased industrialization. To say that AI in these cases is “tackling climate change” is misleading, as what can best be said is that it has the potential to lessen the expected impact of anthropogenic climate change or slow our present course. Much like automobile companies deceptively market their cars as “green” because they emit less carbon dioxide than other cars, AI for sustainability currently seems to rely on confusion of what it means to actively combat climate change and what it means to take a less destructive path toward development. In addition, and key to the subject of this paper, the aims of “AI for Sustainability” tend to separate the real hardware costs of AI development from the potential benefits of the software. As Kate Crawford notes (2021), AI is a material technology, the development of which has far-reaching repercussions for the environment. Thus, a secondary moral hazard entails taking the potential benefits of AI software on the environment to be insurance against the harm caused by AI infrastructure itself. A report from Friends of the Earth (2024) notes, for example, that even if AI makes data center electricity consumption 10% more efficient, doubling the number of data centers still amounts to a massive increase in electricity consumption. Thus, even within the context of China and its AI sustainability goals, an awareness of this tension is becoming more apparent (Mao 2026). If AI were not an ecologically harmful technology, AI for sustainability may come off merely as optimistic projection. Devoting resources to AI programs that may achieve these ends poses at least a moral hazard against positive action that will combat climate change. But, as I note below, AI is ecologically harmful, and so the proposal to go all in on AI for sustainability constitutes a genuine risk. The contradiction of AI development and ecological preservation is flattened out in sustainable AI resulting in the moral hazard; the (bad faith) hope that AI insures us against the climate crisis leads to the “unbalanced and inadequate development” the Communist Party is tasked to resolve. Thus, to address the contradiction and its significance for policy, I analyze the environmental problems of AI below through the political philosophy of the Communist Party of China. As Checketts et al. argue, it is important for Asian societies to apply their own frameworks to problems vexing them in order to best respond (2025; Jackson 2022; Tremayne-Pengelly 2024). While China has important moral resources in Confucianism, Daoism, and Mahayana Buddhism, among other traditions (see: Wong 2021; Hershock 2021; Teschner and Tomasi 2009; Song 2023), the CPC’s model of Chinese Marxism has more obvious implications for public policy than do other traditions. Thus, key to this study are the explicit aims and goals of the CPC as they relate to technological development and sustainability. 3 AI development’s impact on China and Asia It is an interesting fact that the US has suffered the most economically destructive natural disasters of the past 50 years while experiencing relatively mild human effects to these. In the period from 1970 to 2021 inclusive, the US had seven of the top ten most economically destructive natural disasters, and the country’s economic losses counted for a total of 38% of all global losses, with tropical cyclones (hurricanes, typhoons, cyclones) having the greatest impact (WMO 2023b). However, the US’s infrastructure has historically meant that droughts, hurricanes, blizzards, and floods can be responded to quickly without much loss of life. There is evidence this is changing, in no small part due to massive budget cuts to key agencies like the Federal Emergency Management Agency and the National Oceanic and Atmospheric Agency (Frazin 2025). Although a white paper from the Pentagon during the Obama administration warns that climate change is the key national security problem of the twenty-first century (DoD 2015), it is historically undeniable that the US has until now had little human reason to be concerned about climate change, and this is reflected in its policy. Compare the US to China. China’s population is four times that of the US, but the land area is roughly equivalent (also note that India, whose population is larger than China’s, has around one-third of China’s area). Most of China’s population lives on the East Coast, while the US has its population across the East and West Coasts. The US is relatively isolated geographically, only bordering Canada to the North and Mexico to the South and straddling two oceans, while China borders 14 states and faces only the Pacific. Taken together, China’s ecological concerns entail more interconnectedness with the rest of Asia, greater concerns about the inland, and a larger population (domestically and regionally) than the US faces. According to the World Meteorological Association, Asia is the most disaster-prone continent in the world (2023a). In contrast to the high-cost, low-casualty history of the US, Asia experienced up to 47% of all natural disaster-related deaths worldwide from 1970 to 2021 and 33% of all economic losses (WMO 2023b). While flooding caused the greatest amount of economic loss, the greatest number of human deaths are attributed to tropical cyclones. Although the US has had more economic devastation from cyclones, Asia hasexperienced them worse in both frequency and mortality, experiencing 38% of all cyclones and suff ering hundreds ofthousands of deaths from these storms. In addition to cyclones, other meteorological problems tied to global warming abound in Asia, including severe floods, droughts, extreme winds, heat waves, loss of glaciers, and rising ocean tides. China, in particular, is noted as a country facing water scarcity and drought. Agriculture is significantly affected by these impacts, with less stable water supplies resulting in challenges for farming and increased temperatures causing increase of crop-consuming pests (Shaw et al. 2022). In considering the impact of monsoon seasons, long droughts, crop failure, super typhoons and heat waves, China’s stakes in combating climate change are possibly higher than any other country, and certainly higher than any other developed nation dictating AI standards. As a result, China has a vested interest in combating carbon emissions even if this slows industrial processes. While the US has, in Trump’s second term, rescinded nearly all of its plans to respond to climate change, China’s political philosophy and most recent 5-year plan outline specific policy directives to reduce their own contribution to this global problem. China expects to reduce from carbon peak to neutrality within 30 years, achieving “the largest reduction in carbon dioxide emissions per unit of GDP (“carbon intensity”) in the world” (Xinhua 2021). Some specifics of this plan include reducing carbon emissions by 18% between 2021 and 2025, setting carbon emission caps on each province according to their development, lowering their “carbon intensity by over 65 percent by 2030 from the 2005 level,” and, most pertinent to this discussion, “curbing the haphazard development of energy-intensive and high-emission projects” (Xinhua 2021). The impact of climate change on China and the country’s own ecological plan and the goal of balanced and adequate development make environmentally harmful development of special concern. Because of these goals, the measured impacts of AI development constitute significant points of concern both globally and specifically to China. 3.1 Climate change and energy consumption AI is a massively energy-consuming technology. Up to 4.4% of all electricity consumption in the US at present is consumed by data centers (Shehabi et al. 2024). China’s electricity consumption on this front is lower than that of the US, with the US’s data center consumption being 45% of the global total and China merely at 25% of total (IEA 2024). However, the electricity demand for data centers is also expanding, and the total usage is expected to more than double from 415 terawatt hours in 2024 to 945 by 2030. This places incredible strain on existing electricity grids, so the proposed solution is to massively increase the number of power stations globally. Carbon emissions due to data centers depend upon the energy grids these draw from. Places with more green energy infrastructure have less carbon emissions than places still relying more fossil fuels such as coal and natural gas. On this front, China is certainly excelling Western countries. According to the Communist Party, the People’s Republic of China has now achieved 56% total renewable sources for their energy consumption (Communist Party of China (CPC) 2025), while the US Energy Information Administration (2024) counts only 21% of their sources to be renewable. One must therefore note the unfortunate free-rider problem tied to AI carbon emissions; the US currently relies too heavily on fossil fuel-based power plants while contributing substantially more per capita in AI training and development than countries with more sustainable electricity generation. Thus, the US is causing disproportionately more ecological burden, while the consequences, as noted above, are experienced worse in countries like China. This is not just the tragedy of the commons; it is a genuine problem of unfair costs and benefits which China’s own green policy cannot alone fix. Nonetheless, the slowness of some countries to adopt more carbon-friendly policies is not justification for other countries to follow suit. The urgency of global warming demands all countries with an eye to the prosperity of the future do their utmost, even if their chief economic competitors pursue annihilistic energy policies. If this truly is the Chinese Century, China should take the lead in demonstrating ecological prioritization. At present, because of massive energy needs, AI’s total carbon emissions are estimated to be around 100 megatonnes per year (Yu et al. 2024). Some commentators compare the total emissions of data centers to the global air transit industry (One Nine Nine 2024, Jackson and Hodgkinson 2022). The amount of carbon emissions for each new training process of a new AI program increases geometrically as the parameters of the program expand, so that GPT-4 produced around 40 times as much carbon as its predecessor, GPT-3 which was estimated at 550 tonnes for training or around the annual consumption of 120 American homes (Ji and Jiang 2026; Patterson et al. 2021). The use of AI for routine tasks like searching for information is between 5 and 25 times more inefficient than simple search algorithms (FoE; Brussels Times 2024), and the general use of LLMs requires more energy than training due to “unpredictable inference requests” (Ji and Jiang 2026). The plan to increase data centers also entails a need for more energy, and while some tech companies are looking at nuclear power for this end (St. John and McDermott 2024), others are investing in emission-heavy fossil fuels (Halper 2024). This data is also based on the self-reporting of these companies, so it is rather doubtful whether it can be trusted, as swapping energy credits has allowed tech companies to report emission rates possibly seven times lower than they actually are (O’Brien 2024). At a time when it is imperative for us to reduce electricity consumption, the continuation of training and using AI with no restrictions is directly antithetical to efforts to combat climate change. Even if China only uses renewable energy sources for its data centers, there remains the problem that creation of renewable energy generators (e.g., solar panels, wind turbines, hydroelectric dams) also has a direct impact on the environment with the consumption of resources or disruption of local ecosystems. Even though this impact could be less than the carbon emissions of fossil fuels, we should aim to limit it as much as possible. Massive energy consumption for both training and use of AI does not support this goal. As Friends of the Earth points out, even if AI makes data centers more energy efficient, this achievement is undermined by the further proliferation and use of these same data centers. Recalling the sharp impact that climate change has on Asia as a whole and China specifically, with increased floods, droughts, heat waves, tropical cyclones, rising sea levels, and more, it is doubtful that expanding AI training will have a net positive impact on China. Further plans for building more data centers is troubling insofar as this will only add to the climate crisis. However, this is only one aspect (albeit a significant one) of the ecological problem AI development poses for China. 3.2 Water consumption It is increasingly apparent that another cost of AI development is water, especially to cool down these massive data centers (the heat produced by which further adds to the heat of the Earth). Water is pulled from existing water sources where the data centers are located. While the water is not in any sense “destroyed,” its removal from the water table has repercussions for human populations dependent on water sources and for the ecosystems. Karen Hao (2024) points out, for example, that data centers in Arizona already exist in a hot environment and require larger amounts of water, a crucial and scarce resource in the Sonoran Desert. On the other hand, data centers have proliferated across Ireland to take advantage of the lower temperatures, resulting in a huge drain on local energy resources, up to 28% by 2031 (Brodie 2023). Despite this variability in consumption, it is increasingly clear that data centers use huge amounts of water. In 2023, Google alone was responsible for evaporating around 23 billion liters of water, using more in the year than PepsiCo, one of the largest global beverage producers (Li et al. 2025). Current projections estimate that by 2027, data centers will use the equivalent water of half the entire use of the UK (ibid.). Use of this water is implicated obviously in training AI models, but just like carbon emissions, actual use of the AI program also entails large consumptions of water, so that Li et al. estimate “10–50 medium-length responses” elicited through GPT-3 consume around one-half liter of water. Because water withdrawn from the local water table may be evaporated or wasted, the use of water for data centers has significant implications for water-scarce regions. This is of great concern for many other countries beside China, such as Chile where water activists successfully blocked Google’s development of large data centers in 2024 (Lehuedé 2025). In the Chinese context, data centers cluster around major cities such as Shanghai, Beijing, Guangzhou, and Shenzhen. These are also among the most densely populated cities on the planet. While the Guangdong region, which includes Guangzhou, Shenzhen, and Hong Kong, has high annual rainfall, Beijing and Shanghai both have high levels of water stress (Wang et al. 2016). Each of these major Chinese cities numbers over 15 million people, all of whom are dependent upon limited local water supplies. Placing additional strain on local water supplies amounts to placing stress on local populations. A notable effect of climate change is increased droughts and more unstable weather patterns. At present, 75% of Asia is water insecure “with countries hosting more than 90% of the region’s population already confronting an imminent water crisis” (Yunus 2024). Current population sizes, urbanization, and industrialization are already placing stress on water, and as global warming becomes worse, the security of water becomes even more unstable. Adding “thirsty” data centers to this stress certainly will not have the effect of stabilizing water in the region, and China must take the tens or hundreds of millions of its citizens affected by water scarcity seriously. 3.3 Hardware costs Finally, AI requires hardware to develop, with further interconnected ecological impacts beyond further exacerbated costs related to water consumption and carbon emissions, namely extraction of rare Earth elements (REEs), manufacturing impacts, and production of e-waste. REEs are a critical part of high-end computing hardware. Elements such as molybdenum, strontium, and lithium are integral parts of the creation of graphic processing units and other hardware used in these large data centers where thousands of computers are networked together. As such, extraction and refinement of REEs is critical for AI development. However, as the name suggests, REEs are not easily obtainable nor ubiquitous. Sixty percent of all REEs traded globally come from China, most of which come from a large mining and refining site in Bayan Obo, Inner Mongolia. The impact of this mining and refining to the environment is significant. Continual open-pit mining leaves higher amounts of REEs and heavy metals in the soil up to 60 km from the mining site (Han et al. 2024). The process of refining these minerals requires huge amounts of water and large open-air baths of either sulfuric or hydrochloric acid (Hao and Nakano 2011). The water necessary for this places significant strain on a geographically arid region as well (Wang et al. 2023). The result is devastating to both the local ecosystem and the human comunity. As Julie Klinger notes about the cost of REE extraction, “the resulting radioactive rivers, cancer villages, acute chronic arsenic toxicity, and long tooth disease constitute an environmental and epidemiological crisis so grave and expansive that addressing it is now viewed as a matter of national security and territorial integrity” (Klinger 2018, 132). Those who live nearby experience significant health decline due to the process of resource extraction. REEs are not present in high concentrations, so the task of mining requires huge movement of the earth to yield small amounts. There are few other places in the world that produce REEs on a large scale. Nevada, in the US, does have a lot of lithium, but the costs of American mining and (at least past) environmental policies make foreign sources like Chilean lithium more desirable for computer manufacturers. Other REEs, such as cobalt, may be primarily found in conflict regions like the Democratic Republic of the Congo (Prause 2020), and Vietnam is beginning to compete with China in the REE market (O’Connor 2025). The impact of REEs on these countries is significant, but as the world’s primary exporter of REEs, the environmental and human impact of REE mining and refining is more present in China than anywhere else. China is sometimes called “the world’s factory” because it manufactures huge amounts and varieties of goods. Electronics manufacturing in particular has a strong base in southern China, with Foxconn gaining notoriety for its employment conditions in Shenzhen 20 years ago (Dean and Tsai 2010). While hardware manufacturing companies have been considering moving away from China due to increased labor wages, the integrated systems necessary for electronics manufacturing make relocation to a cheaper economy like Vietnam somewhat prohibitive (Nguyen 2025; Dunwoody 2024). Although semiconductor manufacturing has a strong presence in Taiwan, mainland China also has its own role in large-scale chip manufacturing. Taken together, East Asia, especially the region around and including South China, is deeply affected by environmental impacts of electronics manufacturing. These impacts include toxic fumes, soil pollution, and groundwater poisoning. Air pollution in the form of PM2.5, particulates of 2.5 µm or less typically found in smoke and exhaust, contributes to a total of one-fifth of all premature deaths worldwide, with the highest rate of these deaths being in South China and East China (Wan et al. 2025), where it has been as high as 26% (IHME 2019). Manufacturing AI hardware is literally killing people. Finally, the extraction of REEs and the manufacturing of electronic components remain high because computer hardware has a short shelf life and is poorly recycled. The components used in data centers last on average between 3 and 5 years (Maguire 2025). Once a piece has worn out, it should be replaced, requiring more manufacturing. But even though these components are packed with REEs that require difficult processes to extract and refine, only around 20% of electronics are properly recycled. For the year 2022, there was an estimated 63 million tonnes of e-waste globally, of which only 13.8 million tonnes were recycled (Baldé et al. 2024). The rest ends up as e-waste, a problem not unique to AI, but certainly exacerbated by it. While there is little good data on where e-waste ends up, existing scholarship suggests a great amount of it has often ended up in Southern China, in places like Guangzhou and Hong Kong (Campbell and Christensen 2016), though today much may end up in other countries in South Asia like Pakistan (Kazim et al. 2024). The same REEs and other toxic components in the hardware can then seep into the ground, where they sow further chaos with local ecosystems and populations. The World Health Organization (2024) notes that this has harmful effects particularly on children who may be involved in processing waste or may live in areas where industrial runoff affects the atmosphere, soil, or water supply. The problems of carbon emissions and water consumption are not unique in any way to China, or even Asia, but they will be felt more harshly in this region than in Europe and North America. However, the often overlooked hardware impact of AI hasmanifold effects on China compared with other countries. Every step of the process, from extracting and refining REEs to assembling electronics to their decomposition to e-waste, entails environmental harm to China and significant health impacts to Chinese workers and those living near industrial sites. The relatively low rate of e-waste recycling, short lifespans of AI components, and the increasing demand for more data centers feed into a vicious cycle of extraction, manufacturing, waste and extraction again, and the costs are borne by the environment and the people connected to it. 4 Discussion: Chinese Communist political philosophy and the AI moral hazard None of the above information on its own provides any normative indication of how Chinese officials ought to consider AI development. As Hume famously noted, description of how things are is no indication of what they ought to be (1739). Noting that AI’s development has a high ecological toll and that this toll has significant impacts on human populations is not the same as noting that we must not pursue AI, nor is it to say why we should or should not. Normative evaluation is necessary on this front. Indeed, even my above claim that sustainable AI constitutes a type of moral hazard is contingent upon a normative conclusion that it is important to protect the environment and the people dependent upon this environment. Traditional Western philosophical approaches, such as utilitarianism, human rights or natural law, or specific approaches to environmental or technological ethics, such as Hans Jonas’s “new categorical imperative” or Aldo Leopold’s “land ethic,” may be utilized to assess the descriptive information above and to provide clarification on the conflict between AI development and environmental sustainability. Indeed, I recommend that relevant insights from above be considered for different countries and different moral contexts. However, as noted by Asian-based AI ethicists (e.g., Checketts et al. 2025; Checketts and Chan 2024; Wang 2026; Song 2023), Asian societies should use their own cultural resources for addressing AI concerns. Thus, I turn to the political philosophy of the Communist Party of China as a framework for addressing the above considerations. Here, a normative approach native to one country significantly affected by AI’s environmental impact is applied to the case at hand. I also suggest that as far as other countries such as India, Vietnam, and South Korea are devising their own AI future goals, they must also use their own culturally rooted frameworks to weigh environmental costs against their technological futures. The current political ideology of the People’s Republic of China is shaped around the tenets of Lenin-Marxism, Mao Zedong Thought, Deng Xiaoping Thought, and now Xi Jinping Thought. Under premier Jiang Zemin, the Communist Party adopted the “Three Represents,” committing the party to incorporate “the most advanced productive forces, the most advanced culture, and the fundamental interests of the broad masses of the Chinese people” into their strategy, positioning China as a global industrial power (Bo 2004, p. 35). Under Hu Jintao, the Communist Party also added the “Scientific Outlook on Development” which “summarizes the experiences and lessons learned from development issues both domestically and internationally, absorbs the new achievements of human civilization, and, standing at the height of history and the times, further clarifies the major issues of why the country should develop, and how it should develop in the new century and new stage” (Wen 2004). The most recent stage of development of political doctrine includes the Fourteen Commitments (China Central TV-1 (Beijing) 2017) and the Ten Definites articulated by General Secretary Xi Jinping at the 19th National Congress of the CPC in October 2017 (Chinese Media Project 2022). These core points take priority in national strategy, and so examining those relevant for the contradiction of AI development and ecology is important for Chinese national strategy, and are at the heart of the Fifteenth Five-Year Plan (CPC 2025). Definite 3 states: “The main contradiction in China’s society in the New Era is the contradiction between the people’s growing need for a better life and unbalanced and inadequate development, and that it is necessary to adhere to the people-centered development ideology.” The push for AI dominance at the cost of the environment should probably be seen as “unbalanced” or even “inadequate” development. This Definite insists that development is not for development’s sake but rather must be centred on the people. The ecological impacts of AI, including human harms from climate change, water scarcity, hardware production, and industrial waste, are significant. Recall that air pollution is the single greatest killer in South China, and China’s largest metropoles are already water stressed. The doctrine of Scientific Outlook on Development warns against this, noting that “the prominent problems in the country's economic construction are an irrational structure, extensive management methods, and economic growth mainly relying on increased investment and expanded investment scale, resulting in excessive resource and environmental costs.” The path to industrialization must therefore be “characterized by high technological content, good economic benefits, low resource consumption, minimal environmental pollution, and full utilization of human resources” (Wen 2004). This problem is further drawn out through the fourth Definite, which includes the “Challenge of the “five in one” (五位一体"总体布局) approach—meaning economic, political, cultural, social and ecological development together.” Clearly, the belief in AI for sustainability ties into this goal, that economic, social, and ecological development can be tied into a single strategy. Such is further supported in the Fifteenth Five-Year Plan, which aims to “promote technological transformation and upgrading to shift toward digital and intelligent development in the manufacturing sector, develop smart, green, and service-oriented manufacturing, and work faster to transform industrial models and enterprises’ organizational forms” (CPC 2025, p. 7). However, we see above that this does not always seem to be the case. Ecological development entails that AI growth cannot be pursued at all costs. AI will have to be restricted and the proliferation of data centers will constitute a backwards development on this front. But the problem remains challenging insofar as it is also a question of orienting public policy away from dangerous notions of progress. As for the Fourteen Commitments, two stand out as significant for this contradiction. Commitment eight is directed to “improving people’s well-being and livelihoods [as] the primary goal of development” (cf CPC 2025, p. 5). Certainly, the pursuit of AI is meant to achieve this commitment, but the commitment is prior to the specific developments entailed. Indeed, there are many ways in which AI can be positively detrimental to the goals of “long-term peace, order, and stability in the country, and [ensuring] that people are content with their lives and jobs.” The CPC does see many specific applications on this front, such as applying “AI tools in industrial development, cultural advancement, public wellbeing initiatives, and social governance” (CPC 2025, p. 14). On the other hand, environmental degradation certainly will lead to the reverse of this goal, as the people’s well-being, security and stability are threatened by worsening environmental disasters. What can be said with some clarity is that China has prioritized reduction of carbon emissions in its AI strategy, including an international policy to “support the continuous exploration of innovative, resource-saving, and environmentally friendly AI development models, jointly formulate AI energy and water efficiency standards, and promote green computing technologies such as low-power chips and efficient algorithms” (Xinhua 2025b, p. 7). Current research suggests the implementation of the National New Generation Artificial Intelligence Innovation and Development Pilot Zone has actually led to an increase in urban carbon emission efficiency (Wen and Yang 2026). At the same time, observers note that China needs to ramp up further its carbon reductions to achieve targets set in the 2015 Paris Agreement (Myllyvirta et al. 2025). Thus, some commentators note that within the People’s Republic, the goal of “dual carbon” reduction and monitoring that AI provides is at odds with the real carbon footprint AI development creates (Mao 2026). Resolving the contradiction remains paramount. Commitment nine, however, is most explicit about the environment. In full, it reads, “We must establish and practise the philosophy that lucid waters and lush mountains are invaluable assets, uphold the basic national policy for energy conservation and environmental protection, treat the ecological environment as we treat life, coordinate the systematic management of mountains, waters, forests, fields, lakes, and prairies… build a beautiful China, create a good production and living environment for the people, and contribute to global ecological safety." Furthermore, Part XII of the Fifteenth Five-Year Plan, comprising sections 45–48 inclusive, is focused on “accelerating the green transition across the board and building a beautiful China” (CPC 2025). With this commitment fully in mind, the way development is undertaken must not be at the cost of the environment. Rather, it must be oriented toward both the well-being of the people and toward the preservation of nature. It is my suggestion, then, that China carefully consider the real purpose of developing AI. Definite 1 states that “the most essential feature of Socialism with Chinese Characteristics is the leadership of the CPC…the Party being the supreme political force.” This insight aligns with Lukács’ (1971) own view that true consciousness must come be given to the people from the Party. This means the CPC has the opportunity, but also the burden, to cut through false ideologies. If the PRC’s promotion of AI is the same as that of the US and other Western capitalist nations with no nuance or distinctiveness, then Socialism with Chinese Characteristics in the New Era, especially the digital society, is nearly identical to Western capitalism. More, uncritically following the path of Western capitalist models betrays the aims of ecological preservation while moving away from resolving the “principal contradiction” of the current phase of Chinse development. A suggestion, then, is that as China seeks “economic, political, cultural, social and ecological development together,” AI development must be reigned in and directed to specific purposes to these ends. Data centers cannot be constructed ad infinitum. The destruction to Bayan Obo through REE mining and refining, the water stress on cities like Beijing and Shanghai, the carbon emissions from electricity generation and construction, and expansion of pollution through e-waste, manufacturing and chemical refining are not indicative of “ecological development.” Insofar as China controls REEs and hardware production, it can and should use its power to curb large-scale impacts to the environment for AI development. Insofar as AI training is deeply resource intensive, China should prioritize hardware efficiency and slimmer AI models over data center proliferation. Most of all, China should avoid the moral hazard of trusting the insurance of “AI for sustainability” to the detriment of direct effort toward negative carbon emissions. China’s report that DeepSeek required far less resources than comparable LLMs is significant on this front (Calma 2025).Footnote 2 At the same time, it is not enough; the usefulness of LLMs has been exaggerated and their overuse for mundane purposes constitutes significant waste. It is necessary to assess what uses of AI will really “promote the creation of production methods, ways of life, and consumption patterns that are green and low-carbon, civilized, healthy, and create customs that are favorable for all” (CPC 2021, 16), and which are either entirely antithetical to this or at least obstacles on this path. AI development for the “five in one” approach would mean specific applications, a decided push to Artificial Narrow Intelligence directed at CPC identified goals rather than a blanket promotion of a technology known to be ecologically harmful. Rather than developing AI within the context of current global capitalism, China must develop AI for specific Marxist ends. Rather than an unquenchable thirst for greater AI development, targeted applications like agriculture, energy efficiency or healthcare would be better Marxist approaches. AI should be developed within specific parameters, with limitations on its consumption and use, and not for the sake of unchecked profit. This is how to properly navigate the contradiction toward dialectical synthesis. 5 Conclusion: rhetoric and moral hazard Combating the moral hazard of AI is difficult because the rhetoric surrounding it remains powerful. Governments, schools, entertainment, marketing, economists, and, of course, the tech industry with one voice have declared AI to be the future we must face. As the ecological crisis daily worsens, the headlong rush into technological apocalypse is only possible with a population already surrendered to this rhetoric. Even the Chinese government faces this pressure as their envisioning of the future of the country connects to the path of the US and the struggle for AI dominance. The aim to be powerful in AI development and deployment has led to a headlong rush into devoting huge portions of economies and workforces to this end, without considering broader ramifications of the AI race. Unfortunately, as countries aim for further AI growth, they also seem to be lagging behind in their goals combating climate change. UN reporting has shown that most countries and large corporations are lagging far behind their promises from the 2015 Paris Agreement (Warren et al. 2024). The United Nations has warned, furthermore, that the Asia Pacific region has undergone “alarming regression” in recent years on its goals for sustainable development toward climate change (Mishra 2025). Worse, of course, is how countries like the US have taken deliberately regressive stances with their political shifts, prioritizing free-market consumption over pursuing green policies. At some point, it must be accepted that the climate change problem is not fundamentally a technological problem but a social problem. Insofar as our economies prioritize consumption over balance, technological fixes may only pave the way for new demands, as Herbert Marcuse prophetically suggested 60 years ago (1991). What is truly needed is a different orientation toward material production and consumption. On this front, China has the real opportunity to make a difference: as a communist power, its founding tenets prioritize economic development for the people rather than for the sake of growth. This potential, however, remains uncertain. In the first place, the current funding and economic structure of the People’s Republic of China would have to undergo stronger control than already present. This is unlikely to happen without a great deal of domestic and international friction. In the second place, China would have to engage critically with the meaning of “AI for sustainability” and avoid the temptation of trusting that a possibility for AI to help combat climate change is a guarantee that AI will combat it. This leads to a third problem, namely that many researchers and tech leaders within China (as is true everywhere) lack the critical thought to consider whether their projects are actually yielding good results against climate change. With all the pressure to produce AI projects that could be used to address climate problems, little thought among researchers in this region has been given to whether these solutions are not themselves exacerbating the problem. The hope that AI may be able to address issues of climate change should not be taken as an insurance that our uncritical pursuit of greater AI development will yield positive ecological results. Rather, the actual impact of AI development on the climate and the vulnerability that China and its neighbors face with worsening environmental effects must be an important check against any move toward unwarranted techno-optimism. Some recent insights provide at least a glimmer of hope for the People’s Republic of China. While the CPC has no intentions to scale down its development of AI, President Xi Jinping has promised to the United Nations to expand renewable energy sources sixfold by 2035 (Sengupta 2025), and achieving peak carbon emissions before 2030 (CPC 2025). More significantly, the Communist Party seems more focused on developing “practical AI” and applying it directly into national solutions rather than the promethean Artificial General Intelligence project being pursued by American companies (Hamid 2025). Of course, it remains to be seen to what degree China remains on track with both their limited AI development strategies and their climate change goals, but if recent news is any indication, China could pave the way out of the AI climate change moral hazard and toward a true model of both sustainable AI and AI for sustainability. Data availability No datasets were generated or analysed during the current study. Notes Similar emphases are detailed in the more recent Fifteenth Five-Year Plan, which Sect. 4 discusses. However, this more recent document (at least the current English translation) emphasizes AI less than the Fourteenth does. NB: Jegham et al. (2025) claim that in actual performance, DeepSeek consumes the most energy, making it the least efficient comparatively. References Anthropic (2025) Build in America. https://www.anthropic.com/news/build-ai-in-america. Accessed 12 Dec 2025 Aylak BL (2025) SustAI-SCM: intelligent supply chain process automation with agentic AI for sustainability and cost efficiency. Sustainability 17(6):2453. https://doi.org/10.3390/su17062453 Baldé CP, Kuehr R, Yamamoto T et al (2024) Global e-waste monitor. United Nations Institute for Training and Research (UNITAR), Geneva Batin LA (2025) SustAI-SCM: intelligent supply chain process automation with agentic AI for sustainability and cost efficiency. Sustainability 17:2453. https://doi.org/10.3390/su17062453 Bo Z (2004) Hu Jintao and the CCP’s ideology: a historical perspective. J Chin Polit Sci 9:27–45 Braakmann-Folgmann A, Shepherd A, Hogg D, Redmond E (2023) Mapping the extent of giant Antarctic icebergs with deep learning. Eur Geosci Union. https://doi.org/10.5194/tc-17-4675-2023 Brevini B (2023) Artificial intelligence, artificial solutions: placing the climate emergency at the center of AI developments. In: Sætra HS (ed) Technology and sustainable development the promise and pitfalls of techno-solutionism. Routledge, New York, pp 23–33 Brodie P (2023) Data infrastructure studies on an unequal planet. Big Data Soc. https://doi.org/10.1177/20539517231182402 Calma J (2025) AI is ‘an energy hog,’ but DeepSeek could change that. The Verge. https://www.theverge.com/climate-change/603622/deepseek-ai-environment-energy-climate. Accessed 12 Dec 2025 Campbell K, Christensen K (2016) Where does America’s e-waste end up? GPS tracker tells all. Public Broadcasting System. https://www.pbs.org/newshour/science/america-e-waste-gps-tracker-tells-all-earthfix. Accessed 12 Dec 2025 Checketts L, Boyles RJ, Chan BSB et al (2025) Ethical and social concerns of artificial intelligence in Asia. NanoEthics 19:12 Checketts L, Chan BSB (eds) (2024) Social and ethical considerations of AI in east Asia and beyond. Springer, Cham China Central TV-1 (Beijing) (2017) Xi Jinping thought on socialism with Chinese characteristics for a new era. BBC. https://monitoring.bbc.co.uk/product/c1dmwn4r. Accessed 12 Dec 2025 Chinese Media Project (2022) Ten Definites. https://chinamediaproject.org/the_ccp_dictionary/ten-definites/. Accessed 15 Dec 2025 Cirri G (2023) AI for climate change: AI for flood adaptation plans and disaster relief. Oxford Insights. https://oxfordinsights.com/insights/managing-floods-ai-for-adaptation-plans-and-disaster-relief/. Accessed 15 Dec 2025 Communist Party of China (CPC) (2021) The outline of the People's Republic of China 14th five-year plan (2021-2025) for national economic and social development and long-range objectives for 2035. Center for Security and EmergingTechnology, May 12 https://cset.georgetown.edu/wp-content/uploads/t0284_14th_Five_Year_Plan_EN.pdf. Accessed August 10, 2026. Communist Party of China (CPC) (2025) Recommendations of the central committee of the communist party of China for formulating the 15th five-year plan for national economic and social development. The State Council, The People’s Republic of China, October 28. https://english.www.gov.cn/news/202510/28/content_WS6900adb9c6d00ca5f9a07216.html. Accessed August 10, 2026. Corner A, Pidgeon N (2014) Geoengineering, climate change scepticism and the ‘moral hazard’ argument: an experimental study of UK public perceptions. Philos Trans R Soc A Math Phys Eng Sci 372:20140063 https://doi.org/10.1098/rsta.2014.0063 Crawford K (2021) Atlas of AI. New Haven: Yale University Press. Dastin J (2024) OpenAI CEO Altman says at Davos future AI depends on energy breakthrough. Reuters. https://www.reuters.com/technology/openai-ceo-altman-says-davos-future-ai-depends-energy-breakthrough-2024-01-16/. Accessed 12 Dec 2025 Dean J, Tsai T-I (2010) Suicides spark inquiries: apple, H-P to examine Asian supplier after string of deaths at factory. The Wall Street J, May 27. Department of Defense (DoD) (2015) National security implications of climate-related risks and a changing climate. State & Federal Reports and Publications Dunwoody R (2024) The beginning of the end: why electronics manufacturing is leaving China. LinkedIn. https://www.linkedin.com/pulse/beginning-end-why-electronics-manufacturing-leaving-china-dunwoody-rx1zc. Accessed 12 Dec 2025 European Space Agency (2023) AI Maps Icebergs 10,000 Times Faster than Humans. ESA, November 9. https://www.esa.int/Applications/Observing_the_Earth/Copernicus/Sentinel-1/AI_maps_icebergs_10_000_times_faster_than_humans/. Accessed August 10, 2026 Fernandez C, Wiese K (2025) The mirage of EU techno-solutionism to the climate crisis. EU Observer. https://euobserver.com/digital/ar125f5e3f. Accessed 12 Dec 2025 Frazin R (2025) Texas floods shine spotlight on Trump’s weather and disaster cuts. The Hill. https://thehill.com/policy/energy-environment/5388436-texas-flooding-flash-flood-camp-mystic-trump-cuts-noaa-nws-fema-national-weather-service/. Accessed 12 Dec 2025 Friends of the Earth (2024) Artificial Intelligence threats to climate change. https://foe.org/wp-content/uploads/2024/03/AI_Climate_Disinfo_v6_031224.pdf. Accessed 12 Dec 2025 Gebru T, Torres ÉP (2024) The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence. First Monday. https://doi.org/10.5210/fm.v29i4.13636 Geoengineering Monitor (2018) Manifesto against geoengineering. http://www.geoengineeringmonitor.org/wp-content/uploads/2018/10/home-new-EN-dec4.pdf. Accessed 12 Dec 2025 Halper E (2024) A utility promised to stop burning coal; then Google and Meta came to town. The Washington Post Hamid J (2025) China’s AI playbook is nothing like America’s and that might be the whole point. Cryptopolitan. https://www.cryptopolitan.com/chinas-ai-playbook-nothing-like-americas/. Accessed 12 Dec 2025 Han X, Xiong Z, Wang S et al (2024) Long-term open-pit mining activities at the world’s largest light rare earth mine significantly affect light rare earth elements in road dust over long distances. J Hazard Mater. https://doi.org/10.1016/j.jhazmat.2024.136287 Hao K (2024) AI is taking water from the desert. The Atlantic. https://www.theatlantic.com/technology/archive/2024/03/ai-water-climate-microsoft/677602/. Accessed 12 Dec 2025 Hao Y, Nakano J (2011) Rare earth elements, Asia’s resource nationalism, and Sino-Japanese relations. National Bureau of Asian Research Hershock P (2021) Buddhism and intelligent technology: toward a more humane future. Bloomsbury Academic, London Hume D (1739) A treatise of human nature: being an attempt to introduce the experimental method of reasoning into human subjects. John Noon, London IHME (2019) Global burden of disease 2019. The Institute for Health Metrics and Evaluation. https://vizhub.healthdata.org/gbd-compare/#. Accessed 12 Dec 2025 International Energy Agency (IEA) (2024) Energy and AI. World Energy Outlook Special Report Iyer LS (2021) AI enabled applications towards intelligent transportation. Transp Eng 5:100083. https://doi.org/10.1016/j.treng.2021.100083 Jackson T, Hodgkinson IR (2022) What is ‘dark data’ and how is it adding to all of our carbon footprints? World Economic Forum. https://www.weforum.org/stories/2022/10/dark-data-is-killing-the-planet-we-need-digital-decarbonisation/. Accessed 12 Dec 2025 Jegha N, Abdelatti M, Elmoubarki L, Hendawi A (2025) How hungry is AI? Benchmarking energy, water, and carbon footprint of LLM inference. Arxiv. https://doi.org/10.48550/arXiv.2505.09598 Ji Z, Jiang M (2026) A systematic review of electricity demand for large language models: evaluations, challenges, and solutions. Renew Sustain Energy Rev 225:116159. https://doi.org/10.1016/j.rser.2025.116159. Kazim M, Syed JH, Saqib Z et al (2024) Informal e-waste recycling in nine cities of Pakistan reveals significant impacts on local air and soil quality and associated health risks. Environ Pollut 355:124259. https://doi.org/10.1016/j.envpol.2024.124259 Klinger JM (2018) Rare earth frontiers: from terrestrial subsoils to lunar landscapes. Cornell University Press, Ithaca Lawrence MG, Schäfer S, Muri H et al (2018) Evaluating climate geoengineering proposals in the context of the Paris agreement temperature goals. Nat Commun 9:3734 Lehuedé S (2025) An elemental ethics for artificial intelligence: water as resistance within AI’s value chain. AI Soc 40:1761–1774. https://doi.org/10.1007/s00146-024-01922-2 Li P, Yang J, Islam MA, Ren S (2025) Making AI less ‘thirsty’: uncovering and addressing the secret water footprint of AI models. https://doi.org/10.48550/arXiv.2304.03271 Lukács G (1971) History and class consciousness. Livingstone (trans). MIT Press, Cambridge Maguire N (2025) How long do data center servers last? Procurri. April 24. https://www.procurri.com/2025/04/24/how-long-do-data-center-servers-last/. Accessed 12 Dec 2025 Mao V (2026) China: climate change and Artificial Intelligence. Climate scorecard. https://www.climatescorecard.org/2026/04/china-climate-change-and-artificial-intelligence/. Accessed 10 Jun 2026 Mao Z (1966) Quotations from Chairman Mao, 1st pocket ed. Foreign Languages Press, Beijing Marcuse H (1991) One-dimensional man: studies in the ideology of the advanced industrial society. Beacon Press, Boston Martínez A, Fan J, Miller SA (2026) Machine learning in life cycle assessment and low carbon material discovery: challenges and pathways forward for the construction industry. Resour Conserv Recyc. https://doi.org/10.1016/j.resconrec.2025.108567 Mishra, V (2025) Asia-Pacific falling behind on sustainable development and climate targets. UN News. https://news.un.org/en/story/2025/02/1160216. Accessed 12 Dec 2025 Myllyvirta L, Schäpe B, Qin Q, Qiu C, Shen X (2025) China’s climate transition: outlook 2025. Centre for Research on Energy and Clean Air. https://energyandcleanair.org/wp/wp-content/uploads/2025/12/CREA_China-Climate-Transition-Outlook-2025.pdf. Accessed 10 Jun 2026 National Institute for Food and Agriculture (NIFA) (2025) Artificial intelligence. US Department of Agriculture. https://www.nifa.usda.gov/grants/programs/data-science-food-agricultural-systems-dsfas/artificial-intelligence. Accessed 12 Dec 2025 Nguyen T (2025) Vietnam’s electronics sector rising fast, but structural hurdles remain Vietnam Investment Review. https://vir.com.vn/vietnams-electronics-sector-rising-fast-but-structural-hurdles-remain-131838.html. Accessed 12 Dec 2025 Niemeyer K, Varanasi L (2024) Former Google CEO Eric Schmidt says we should go all in on building AI data centers because 'we are never going to meet our climate goals anyway. Business Insider. https://www.businessinsider.com/eric-schmidt-google-ai-data-centers-energy-climate-goals-2024-10. Accessed 12 Dec 2025 O’Brien I (2024) Data center emissions probably 662% higher than big tech claims; can it keep up the ruse? The Guardian. https://www.theguardian.com/technology/2024/sep/15/data-center-gas-emissions-tech. Accessed 12 Dec 2025 O’Connor D (2025) Vietnam’s rare earth awakening: from sleeping giant to global contender. Rare Earth Exchanges. https://rareearthexchanges.com/news/vietnams-rare-earth-awakening-from-sleeping-giant-to-global-contender/. Accessed 12 Dec 2025 O’Connor R, Bolton M, Saeri AK, Chan T, Pearson R (2024) Artificial intelligence and complex sustainability policy problems: translating promise into practice. Policy des Pract 7:308–323. https://doi.org/10.1080/25741292.2024.2348834 One Nine Nine (2024) The cloud now has a greater carbon footprint than the airline industry. One Nine Nine Agency, April 30 https://www.oneninenine.agency/articles/the-cloud-now-has-a-greater-carbon-footprint-than-the-airline-industry Accessed August 10, 2026 Patterson D, Gonzalez J, Le Q et al (2021) Carbon emissions and large neural network training. https://doi.org/10.48550/arXiv.2104.10350 Prause L (2020) Chapter 10—Conflicts related to resources: the case of cobalt mining in the Democratic Republic of Congo. In: Bleicher P (ed) The material basis of energy transitions. Academic Press, Cambridge, pp 153–167. https://doi.org/10.1016/B978-0-12-819534-5.00010-6 Research Grants Committee (2025) Call for preliminary proposals: areas of excellence scheme 2026/27 (fourteenth round), theme-based research scheme 2026/2027 (sixteenth round). University Grants Committee. https://www.ugc.edu.hk/eng/rgc/funding_opport/aoe/call_letter.html. Accessed 12 Dec 2025 Rogelj J, Shindell D, Jiang K et al (2018) Mitigation Pathways compatible with 1.5°C in the context of sustainable development. In: Masson-Delmotte V et al. (ed) Global warming of 1.5°C. An IPCC special report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. Cambridge University Press, Cambridge, pp 93–174. https://doi.org/10.1017/9781009157940.004 Rowe J (2025) Delivering regenerative agriculture through digitalization and AI. World Economic Forum. https://www.weforum.org/stories/2025/01/delivering-regenerative-agriculture-through-digitalization-and-ai/. Accessed 12 Dec 2025 Sengupta S (2025) At U.N. Summit, China promises to reduce greenhouse gas emissions. New York Times. September 24 Shaw R, Luo Y, Cheong TS et al (2022) Asia. In: Pörtner R (ed) Climate change 2022: impacts, adaptation and vulnerability: contribution of Working Group II to the sixth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, pp 1457–1579. https://doi.org/10.1017/9781009325844.012 Shehabi A, Newkirk A, Smith SJ et al (2024) 2024 United States data center energy usage report. Lawrence Berkeley National Laboratory. Report #: LBNL-2001637. https://doi.org/10.71468/P1WC7Q Song B (2023) How Chinese philosophy impacts AI narratives and imagined AI futures. In: Cave S, Dihal K (eds) Imagining AI: how the world sees intelligent machines. Oxford Academic, Oxford, pp 338–352. https://doi.org/10.1093/oso/9780192865366.003.0021 St. John A, McDermott J (2024) Amazon, Google make dueling nuclear investments to power data centers with clean energy. AP News Teschner G, Tomasi A (2009) Technological paradigm in ancient Taoism. Techne Res Philos Technol 13:190–205 The Brussels Times News Staff (2024) ChatGPT consumes 25 times more energy than Google. The Brussels Times. https://www.brusselstimes.com/1042696/chatgpt-consumes-25-times-more-energy-than-google?pStoreID=hpepp. Accessed 10 August 2026. The Technology Executive Committee (TEC) (2024) Artificial intelligence for climate action in developing countries: opportunities, challenges and risks. United Nations Climate Change Tremayne-Pengelly A (2024) A.I. data centers are emitting nearly as much greenhouse gases as commercial airlines. Observer. https://observer.com/2024/12/ai-data-center-carbon-emission/. Accessed 12 Dec 2025 Tripathi G, Shakya A, Kanga S et al (eds) (2024) Big data, artificial intelligence, and data analytics in climate change research: for sustainable development goals. Springer, Singapore US Energy Information Administration (2024) What is U.S. electricity generation by energy source? https://www.eia.gov/tools/faqs/faq.php?id=427&t=3. Accessed 12 Dec 2025 van Wynsberghe A (2021) Sustainable AI: AI for sustainability and the sustainability of AI. AI Ethics 1:213–218. https://doi.org/10.1007/s43681-021-00043-6 Vermaa AK, Prakash M, Shakila B et al (2025) Optimal dispatch of distributed renewable energy sources in isolated microgrid system exploiting metaheuristic optimization algorithms. In: Satpathy S (ed) AI for climate change and environmental sustainability. CRC Press, Boca Raton, pp 98–108 Wan Z, Ma Y, Feng F et al (2025) Health risk assessment of atmospheric fine particulate matter in China. Atmos Environ 362:121550. https://doi.org/10.1016/j.atmosenv.2025.121550 Wang J, Zhong L, Long Y (2016) Baseline water stress: China. World Resources Institute Wang QQ, Geng CX, Wang L et al (2023) Water conservation and ecological water requirement prediction of mining area in arid region based on RS-GIS and InVEST: a case study of Bayan Obo Mine in Baotou, China. Sustainability 15(5):4238. https://doi.org/10.3390/su15054238 Wang Z (2026) Plural intelligence: Chinese philosophy, intercultural semantics, and the future of AI. Humanit Soc Sci Commun. https://doi.org/10.1057/s41599-026-07830-8 Warren H, Stringer D, Janicki J, Clark A (2024) The world’s best hope to beat climate change is vanishing. Bloomberg. https://www.bloomberg.com/graphics/2024-cop29-climate-change-targets/. Accessed 12 Dec 2025 Wen H, Yang W (2026) Artificial intelligence applications and the improvement of carbon emission efficiency from the perspective of sustainable development: empirical evidence from China. Carbon Balance Manag. https://doi.org/10.1186/s13021-026-00402-y Wen J (2004) Firmly establish and conscientiously implement the scientific outlook on development. Xinhua. http://news.sina.com.cn/c/2004-02-29/11221926301s.shtml. Accessed 12 Dec 2025 Wong P-H (2021) Artificial intelligence, personal decisions, consent, and the Confucian idea of oneness. In: Wong P-H, Wang TX (eds) Harmonious technology: a Confucian ethics of technology. Routledge, London, pp 79–94 World Health Organization (2024) Fact sheet: electronic waste. https://www.who.int/news-room/fact-sheets/detail/electronic-waste-%28e-waste%29. Accessed 12 Dec 2025 World Meteorological Organization (WMO) (2023a) State of the climate in Asia 2022, WMO-No. 1321, Geneva, World Meteorological Organization World Meteorological Organization (WMO) (2023b) WMO atlas of mortality and economic losses from weather, climate, and water extremes (1970–2019), WMO-No. 1267, 2023 update Xi J (2017) Secure a decisive victory in building a moderately prosperous society in all respects and strive for the great success of socialism with Chinese characteristics for a new era. Address to the Nineteenth National Congress of the Communist Party of China Xinhua (2021) Full text: responding to climate change: China’s policies and actions. The State Council of The People’s Republic of China Xinhua (2025a) Full text: carbon peaking and carbon neutrality China’s plans and solutions. The State Council of the People’s Republic of China Xinhua (2025b) Full text: global action plan for Artificial Intelligence governance. The State Council of the People’s Republic of China Yu Y, Wang J, Liu Y et al (2024) Revisit the environmental impact of artificial intelligence: the overlooked carbon emission source? Front Environ Sci Eng 18:158. https://doi.org/10.1007/s11783-024-1918-y Yunus M (2024) Asia’s looming thirst: can we avoid a water crisis? Geography Directions https://doi.org/10.55203/UJWA9156 Zhou Y (2022) Correctly understand the changes in China’s principal social contradictions. Int J Educ Humanit 4(3):146–149 Zhou Y, Bu W (2026) Artificial intelligence adoption, energy management, and corporate energy transition: evidence from energy consumption, energy intensity, and carbon emission intensity. Energies 19:821. https://doi.org/10.3390/en19030821 Acknowledgements I wish to acknowledge the Asian AI Research Group for the Centre for Digital Culture in the Dicastery for Culture and Education for feedback on earlier versions of this paper. Funding Open access funding provided by Hong Kong Baptist University Library. Author information Authors and Affiliations Contributions LC wrote the entire manuscript, proofread, edited, included the references, and submitted the work. Corresponding author Additional information Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Rights and permissions Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. About this article Cite this article Checketts, L. The sustainable AI development contradiction: Chinese policy on sustainability and the costs of AI development. AI & Soc (2026). https://doi.org/10.1007/s00146-026-03296-z Received: Accepted: Published: Version of record: DOI: https://doi.org/10.1007/s00146-026-03296-z