The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2575
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T07:56:13.339982Z
status: published
lens: p_space
sector: climate
headline: The sustainable AI development contradiction: Chinese policy on sustainability and the costs of AI development
dek: 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 t…
gain_title: (none)
problem_title: A review paper argues that AI's emissions, water use and hardware waste fall especially hard on China, and that counting on AI to fix climate change could delay the emissions cuts needed now.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: A review paper argues that AI's emissions, water use and hardware waste fall especially hard on China, and that counting on AI to fix climate change could delay the emissions cuts needed now.
problem_evidence: treating AI as insurance against climate change may inhibit governments from taking necessary action now
quick_read: A review paper by Levi Checketts of Hong Kong Baptist University argues that relying on "AI for sustainability" is a moral hazard. Hope that AI will solve climate change, he writes, can excuse governments from cutting emissions now. He reviews three linked costs of AI development: carbon emissions, water use, and the damage from making and discarding hardware.

The paper cites estimates that AI's total carbon emissions are around 100 megatonnes a year, that data center electricity demand will more than double from 415 terawatt hours in 2024 to 945 by 2030, and that Google alone evaporated about 23 billion liters of water in 2023. It says China hosts about 25% of global data center consumption, and its energy mix is 56% renewable by the Communist Party's own count, against 21% for the US.

The author says China is especially exposed. Asia had up to 47% of natural disaster deaths worldwide from 1970 to 2021, data centers cluster near water-stressed Beijing and Shanghai, and China supplies 60% of traded rare earth elements, mostly from Bayan Obo in Inner Mongolia. He urges China to pursue narrow, targeted AI, but concedes it is uncertain whether the country will do so.
limitation: The paper is a review and argument grounded in cited secondary reporting; the author says no datasets were generated or analysed, and that it remains uncertain whether China will keep both its limited AI strategy and climate goals on track.
tag: Evidence-backed problem
key_points: The author calls "AI for sustainability" a moral hazard, comparing it to geoengineering, because the hope of future AI fixes can reduce pressure for immediate emission cuts. | AI's total carbon emissions are estimated at around 100 megatonnes per year; the paper says GPT-4 produced around 40 times the carbon of GPT-3, which was estimated at 550 tonnes for training. | Data center electricity use is projected to rise from 415 terawatt hours in 2024 to 945 by 2030; data centers took up to 4.4% of US electricity. | Google evaporated about 23 billion liters of water in 2023, and data centers are projected to use the equivalent of half of the UK's water use by 2027, according to cited estimates. | In 2022 the world produced about 63 million tonnes of e-waste, of which 13.8 million tonnes were recycled; data center components last 3 to 5 years. | China mines about 60% of traded rare earth elements, mostly at Bayan Obo, where soil contamination from open-pit mining extends up to 60 km from the site. | The author recommends China prioritize hardware efficiency and slimmer models over data center expansion, and cap AI use to specific goals such as agriculture, energy efficiency or healthcare.
rundown: The paper frames the issue through China's official "principal contradiction" between unbalanced development and the people's needs for a better life. China's Fourteenth Five-Year Plan sets goals both to lead on climate change and to advance AI, and the author says many Chinese leaders do not see these aims as in conflict. He compares this to the Hong Kong University Grants Committee's 2026-2027 funding call, which lists both sustainable environment and AI as areas of excellence.

On energy, the paper notes China accounts for 25% of global data center consumption versus 45% for the US, and that China's grid is cleaner. It calls this a free-rider problem, since the US relies on fossil fuels while the effects are felt more in countries like China. It also cites claims that corporate energy-credit swaps may let tech firms report emissions possibly seven times lower than actual, and that using AI for routine searches is between 5 and 25 times less efficient than simple search.

On water and hardware, the paper cites Ireland, where data centers could use up to 28% of local energy by 2031, and Chile, where activists blocked Google data centers in 2024. In China, it says Beijing and Shanghai have high water stress and each exceed 15 million people. Mining and refining rare earths uses acid baths and large amounts of water, and PM2.5 pollution from manufacturing is linked in the paper to up to 26% of deaths in South and East China.

The author also cites evidence on the other side: one study found a 2% energy reduction among Chinese firms adopting AI over 12 years, another found a pilot zone raised urban carbon emission efficiency, and a logistics study found a possible 30.3% emission cut. He argues none of this amounts to actively combating climate change. He notes DeepSeek reportedly needed far fewer resources, but also cites a study claiming it consumes the most energy in actual performance.
sources:
- peer_reviewed | AI & SOCIETY | https://doi.org/10.1007/s00146-026-03296-z | 2026-09-02
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

bd7d130e13e466bf6fc0ff2e3f8cb75f83a7d1e3761dcb253c98b796551c9e16

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000