Available via license: CC BY 4.0
Content may be subject to copyright.
Page 1/11
E-waste Challenges of Generative Articial
Intelligence
Peng Wang
Institute of Urban Environment,Chinese Academy of Sciences https://orcid.org/0000-0001-7170-1494
Ling-Yu Zhang
Institute of Urban Environment, Chinese Academy of Sciences
Asaf Tzachor
Reichman University https://orcid.org/0000-0002-4032-4996
Eric Masanet
University of California, Santa Barbara
Wei-Qiang Chen
Institute of Urban Environment, Chinese Academy of Sciences https://orcid.org/0000-0002-7686-2331
Brief Communication
Keywords:
Posted Date: March 13th, 2024
DOI: https://doi.org/10.21203/rs.3.rs-3978528/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: There is NO Competing Interest.
Page 2/11
Abstract
Generative articial intelligence (GAI), a subset of articial intelligence, requires substantial
computational hardware resources for data processing and model training. However, the electronic-waste
(E-waste) toll of GAI remains underexplored and overlooked. Here, we propose a Computational Power-
driven Material Flow Analysis (CP-MFA) model to measure GAI-related E-waste generation, with a specic
focus on large language models. By quantifying server requirements and E-waste generation of GAI under
different scenarios, we nd that this emerging waste stream will grow at a rapid pace (16 million tons
cumulative waste by 2030) with deleterious environmental impacts. Accordingly, we call for an early
adoption of circular economy measures among server manufacturers and data center operators. This
study reveals signicant hardware-linked environmental implications in the context of GAI boom.
Introduction
Generative Articial Intelligence (GAI) represents a pivotal advancement in the eld of articial
intelligence (AI). GAI, including Large Language Models (LLMs), is capable of generating text, images,
videos, or other types of content in response to input prompts. LLMs, a category of GAI that leverages
Natural Language Processing (NLP)1, are often trained on vast datasets and can be ne-tuned on
domain-specic data to provide expert-level insights in certain elds2,3. This represents a signicant
advancement in NLP. LLMs require intensive computational power for training and inference,
necessitating advanced computing hardware and infrastructure4. This demand has considerable
environmental implications due to the energy consumption and carbon footprint associated with these
operations5,6. The development of LLMs (such as GPT-4, BERT, ERNIE and DeBERTa), alongside
successful GAI products in image and video generation, highlights the growing trend of global hardware
facilities expansion to support future computing needs, and emphasizes the timeliness and importance
of sustainable computing.
Previous studies of sustainable computing have focused on AI models’ energy use and carbon emission7-
10. However, the materials used during their lifecycle, and the waste stream of obsolete electronic
equipment – known as electronic waste (E-waste) – received little-to-no attention. Indeed, as the fastest-
growing waste stream11, E-waste, the materials it is composed of, and their potential impacts on public
and planetary health, have long been recognized as a pressing global concern11,12.
As a response, in this Brief Communication, we propose a Computational Power-driven Material Flow
Analysis model (CP-MFA, in Fig.1(a)) to quantify the prospective newly input, in-use stock, and end-of-
service number of designated AI servers installed in data centers aimed to support LLMs computational
requirements quarterly from 2020 to 2030. In our analysis, we focus on AI servers, leaving out ancillary
machinery such as cooling units. Details of our approach can be found inSection S1.1of Supplementary
Information. Notably, our purpose is not to precisely predict the amount of AI servers and their related
waste, but to provide initial gross estimates to appreciate the scale of the imminent E-waste challenge,
Page 3/11
and explore the possible solutions to address the rising waste challenges. Hence, we develop different
scenarios concerning different LLM development trends, consider feasible circular economy strategies,
and other factors including geopolitical technical barriers (SeeSection S3of Supplementary
Information). We emphasize that the following numerical results contain uncertainties, but can reect the
overall tread of the whole GAI eld.
We link the computational power with a gold standard benchmark server (an 8-units graphics processing
unit (GPU) server is chosen as the proxy, such as the state-of-the-art Nvidia DGX H100 system published
in 2023) to obtain conversion factor of computational power to physical infrastructure consisting GPUs,
central processing units (CPUs), storage and memory units, internet communication modules, and power
systems. Given future developments in digital electronics, we measure benchmark server’s development
by setting computational power intensity per server to exponential expansion according to Moore's Law.
The choice of 8-units server is a balance of all major AI server types (4-units, 8-units and 16-units, in
which 8-units are the most common). Although selecting 4-units or 16-units servers as the functional unit
may result in different conversion factors, the qualitative results of total waste generation will stay
unchanged.
To fully explore the possible futures of LLMs, we develop three scenarios where LLMs proliferate under
three representative patterns in terms of global congurations of acceptance and utilization (See Section
S3.1 of Supplementary Informationfor detailed hypotheses, congurations, and comparisons with
existing projections). Fig.1(c) depicts the global increase of AI server amount on a quarterly basis for the
three scenarios. These evaluations are traced to End-of-Service (EoS) E-waste generation, while their
further treatments (e.g., refurbishing or material recycling) are not considered. In the optimistic scenario
where LLMs become ubiquitous (i.e., everyone uses it on a daily basis such as social networks), the
results indicate that the EoS E-waste stream from designated data centers would rise to approximate 16
million tons (Mt) within a decade from 2020 to 2030.
In view of the current conventional large-equipment E-waste stream dened by the UN, our estimated
LLMs EoS E-waste will constitute roughly 11% of the global total during the same period11. The
compound annual growth rate (CAGR) of those LLMs-related E-waste will reach 110%, compared to the
2.8% of global conventional E-waste11 such as screens and washing machines. In moderate and
conservative scenarios (SeeSection S3.2of Supplementary Information), these numbers will decrease to
some extent but remain signicant, nonetheless. The most conservative estimation is 8 Mt E-waste in
cumulative, accounting for 6% of the global stream of E-waste with 90% CAGR. Given that data centers
are highly geographically clustered, these waste streams will be mainly located in Europe (14%), North
America (58%) and East Asia (25%) (see Fig.1(b)).
To address the impending E-waste challenge, circular economy strategies should be implemented10.
Accordingly, we develop and explore three strategic levers spanning different life cycle stages of servers
to mitigate the Out-of-system (OoS) E-waste. The rst strategic lever (C1) examines lifespan extension via
improved maintenance in the use phase. The second strategic lever (C2) examines stepwise server
Page 4/11
upgrading, indicating a conservative server purchase and deployment strategy. The third lever (C3)
explores module reuse considering the reuse of key modules in the (re)manufacturing phase. To compare
the potential ecacy of each strategy we examine, the optimistic LLM proliferation scenario is set as the
baseline scenario. The detailed settings, potentials, and real-world practices of each strategy can be
found in Section S3.3 of Supplementary Information. These strategies can alleviate OoS E-waste
environmental impacts by implementing circular processes on the upstream EoS waste (as depicted in
Fig.2(a)).
We nd that cumulative OoS E-waste between 2020-2030 can be reduced by -58% to 30% (-9.3~4.9 Mt)
depending on the chosen circular economy strategy, respectively. The most effective strategy is lifespan
extension (C1, which is actively implemented by global operators) on condition that the average lifespan
of servers is extended by 1 year indowncyclingusage (where in specic cases it can be extended to six
years, showing more signicant effects). Similarly, module reuse strategy (C3) reduces E-waste by 21%
(3.4 Mt). This measure refers to dismantling, renovation, and re-assemblage of critical modules of
obsolete servers (GPU, CPU, battery, etc.), before they can be reused indowncyclingcomputing. These
strategies (C1 and C3) are expected to have immediate results while, the presumable effect of stepwise
upgrading (C2) is uctuating and potentially counter-effective (with 9.3 Mt more E-waste compared to
baseline scenario, see more information in Section S3.3 in Supplementary Information). It indicates the
purchase and upgrade of servers in data centers are stepwise as the major GPU update period is
generally two-years, rather than the ideal continuous growth. Therefore, the actual updating strategy
should be approached to the ideal continuous upgrading while balancing other operational and nancial
factors such as electricity consumption.
Considering the regional concentration of data centers as well as the existence of ban imposed by
U.S.13,14, we further investigate how potential technical barrier can impact the scale of OoS E-waste ows.
Several scenarios are hypothesized: (T1) constant technical barrier that causes lagged computational
power, (T2) varying technical barrier due to regional unbalanced development, and (T3) technical barrier
with the above circular economy measures implemented (SeeSection S3.4of Supplementary
Information).
Our results indicate that technical barriers constitute a negative factor to E-waste treatment. Without
barriers (i.e., trade bans), data centers worldwide can freely purchase the latest model of servers to
provide more intensive computational powers. However, the highly concentrated supply of AI servers
(GPU chips) under geopolitical considerations can result in the loss of computational power eciency,
resulting in higher physical server demand (for instance, the eciency of Nvidia H800 is half that of H100
from the perspective of bandwidth, so it requires twice the amount to achieve the same effect. For
detailed information see Section S3.4.1of Supplementary Information). We nd that a quarter of LLMs
training and inference held by 1-year lag computational power will lead to 39% more EoS E-waste
generation, cumulating to 6.4 Mt from 2020 to 2030. This is comparable to the global screen and monitor
wastes in 2019 10. In a more optimistic case, if the regional semiconductor industry accelerates and
catches up through its self-development, only 25% more E-waste will still be generated compared to the
Page 5/11
baseline scenario. Adopting lifespan extension measure can compensate for the additional E-waste
brought by technical barriers, indicating the signicance of optimizing the usage of servers. The overall
result appeals for international cooperation to mutual advancement in GPU computational power.
We further quantify the valuable and hazardous metals associated with obsolete LLMs-related servers
inFig. 2(c). Unlike other major large equipment wastes, e.g., household appliance or professional heating
and ventilation, AI servers are metal enriched E-waste concentrated in three component categories:
printed circular boards (PCBs), batteries, and structural parts (as shown in Fig.1(a)). They contain toxic
metals including lead and chromium, and valuable metals such as gold, silver, platinum, nickel, and
palladium. The amount and values of these materials are demonstrated in Fig.2(c). Regarding toxicity,
waste servers can prove harmful to both the environment and public health. Polymers,which assumed to
constitute 30% of PCBsare not naturally degradable, and have been shown to damage soil fertility.
Concomitantly, E-waste burial (landlling) leads to contamination of soil and underground water.
Incomplete incineration could result in release of dioxins and furans, jeopardizing public health15. We nd
that 11.2 Mt OoS E-waste cumulated by 2030 under the lifespan scenario will generate a substantial
number of toxic substances, including lead (917 Kt), barium (6 Kt), cadmium (30 t), antimony (7 Kt),
chromium (907 t), arsenic (34 t) and hydrargyrum (7 t) (SeeSection S1.3of Supplementary Information).
While these gures are approximations (due to changes of material intensity within advanced computing
devices), the emission associated with these toxic materials at such scales, could contribute to increased
incidence of spontaneous abortions, abnormal fetal developments, mutations, abnormal thyroid function,
decreased lung function, or neurobehavioral disturbances16.
Conversely, if recycled properly, these materials can create economic gains, promote circularity in the
electronics industry, and thereby mitigate mining of raw materials (metal ores) while supporting
ecosystem conservation. Our quantication indicates that common recycling of materials (including
copper, lead, aluminum, iron, tin, nickel, gold, silver, platinum, and palladium) embedded in these
decommissioned servers would worth appropriately 70 billion dollars (SeeSection S1.3of Supplementary
Information). This amount will stand at 51 and 24 billion dollars under moderate and conservative
scenarios, respectively. In addition, metal recycling consumes only 5%-90% energy than equivalent
materials smelting from metal ore, which would result in signicantly reduced carbon emission17. For
example, mining gold from decommissioned and discarded electronics emits 80% less carbon dioxide
compared with mining from the ore18. Therefore, recycling technologies and sustainable dismantling
methodologies for LLM-related servers are essential if sustainable computing is to be realized. Currently,
various electronic component recycling technologies can reach a recycling rate of over 95% (SeeSection
S4of Supplementary Information).
In addition to data center-internal recycling and secondary raw material recycling, functional but outdated
AI server wastes may be repurposed and re-deployed for alternative uses. Indeed, workloads during
training and inference of LLMs are markedly more intensive than many common computer tasks such as
basic NLP tasks including sentiment analysis, named entity recognition, and language translation, small-
scale automation tasks including scripting, browsing, document editing, and basic oce applications. A
Page 6/11
functional yet obsolete AI server could therefore support educational activities in schools such as
experimenting with small-scale NLP projects and models, and in small- and medium-size enterprises,
such as serving as a media center PC and for various Linux-based projects. Taken together, these three
measures: extensive redistribution, close-loop reproduction, and material regeneration form the so-called
circular economy, which is gaining a greater foothold in the electronics industry.
Regarding the E-waste stream from data centers, grand operators declared their sustainable vision and
feasible approaches. Zero-waste or zero-landll-waste sustainability goals are already being implemented
and aimed to be realized over the next decade (SeeSection S5of Supplementary Information). One
notable example is the Circular Centers built by Microsoft. Served for its own Azure data centers, Circular
Centers in Europe and North America manage to process more than ten thousand waste servers monthly,
where 83% servers or their components are repurposed and 22% of their materials are recycled.
To prepare in advance for the surge of E-waste from LLMs deployment, circular strategies should be
encouraged and endorsed more rmly. For example, old recycling networks involve massive off-record
procedures, making the material ow nontransparent and dicult to trace. Advantageously, since LLMs-
related server farms – and their associated E-waste streams – are geographically concentrated,
traceability may be easier to achieve. To ensure traceability, the authenticity of data center operators,
manufacturing entities and recycling entities needs to be guaranteed. Policies for other waste-managing
entities such as battery can be referenced using sustainability certication or digital life-cycle
passports19. More transparent information and analytical methods on servers’ lifecycle status could
promote reuse, repurpose, and recycle. Hence, an industrial standard for servers’ status inspection and
labelling should be unitedly recognized, and more public databases containing data center operation
information is called for. In the perspective of hardware requirements, recycling entities should be
organized to provide abundant and professional services in waste collection, dismantling, transportation,
and recycling. Given clustering features of the AI server industry, an integrated circular center handled by
the manufacturers and operators is preferable in dealing with intensive server wastes. Future studies into
circularity strategies, recycling techniques and sustainable computing of GAI are direly required.
This Brief Communication studied the lifecycle of AI server from the perspective of computational power
demands. Compared with the exponential growth estimation indicated in previous research, the prediction
of LLM scale proliferation is more appropriate with the training-data-size constrain (See Methods). The
proposed CP-MFA model focuses on the AI-server-related E-waste stream, exploring both EoS E-waste
generation and OoS E-waste generation. Using several scenarios, the effectiveness of potential circular
strategies is analyzed and compared, alongside recommendations for industry and scholarship. This
model is not without limitations, including the hypothesis of constant GPU-server computational power
intensity and the rough estimation of parameter conguration. Future studies may require corrections and
modied assumptions to reveal more effective circular strategies. Nonetheless, our approximations call
for urgent action to mitigate the swell of GAI-related E-waste.
Declarations
Page 7/11
Acknowledgments
This research was nancially supported by the National Natural Science Foundation of China (72274187
to P.W., 71961147003 to W.-Q.C.), CAS IUE Research Program (no. IUE-JBGS-202202 to P.W.).
Author contributions
P. W. and Z. L. designed the research; Z. L., P.W., and A.T. led the drafting of the manuscript. P. W., Z. L.,
and W.-Q.C. contributed to the methodology; Z. L., E.M., P. W., and A.T. interpreted the results. All authors
contributed signicantly to the nal writing of the article.
Declaration of interests
The authors declare no competing interests.
Supplemental information
The detailed model information and additional results are provided in the supplemental information.
Data and code availability
This paper analyzes existing and publicly available data. Any additional information required to
reanalyze the data reported in this paper is available from the lead contact upon request.
References
1. Editorial. What’s the next word in large language models?
Nat. Mach. Intell.
5, 331–332 (2023).
2. Singhal, K.
et al.
Large language models encode clinical knowledge.
Nature
620, 172–180 (2023).
3. Grossmann, I.
et al.
AI and the transformation of social science research.
Science
380, 1108–1109
(2023).
4. Yin, J., Dash, S., Gounley, J., Wang, F. & Tourassi, G. Evaluation of pre-training large language models
on leadership-class supercomputers.
J. Supercomput.
79, 20747–20768 (2023).
5. Lannelongue, L.
et al.
GREENER principles for environmentally sustainable computational science.
Nat. Comput. Sci.
3, 514–521 (2023).
. Creutzig, F.
et al.
Digitalization and the Anthropocene.
Annu. Rev. Environ. Resour.
47, 479–509
(2022).
7. Mytton, D. & Ashtine, M. Sources of data center energy estimates: A comprehensive review.
Joule
6,
2032–2056 (2022).
. Debus, C., Piraud, M., Streit, A. et al. Reporting electricity consumption is essential for sustainable AI.
Nat Mach Intell.
5, 1176–1178 (2023).
9. Aldossary, M. & Alharbi H. A. An Eco-Friendly Approach for Reducing Carbon Emissions in Cloud Data
Centers,
Comput. Mater. Con.
72, 3175-3193 (2022).
Page 8/11
10. Kaack, L.H., Donti, P.L., Strubell, E. et al. Aligning articial intelligence with climate change mitigation.
Nat. Clim. Chang
. 12, 518–527 (2022).
11. Forti, V., Balde, C. P., Kuehr, R. & Bel, G.
The Global E-waste Monitor 2020: Quantities, ows and the
circular economy potential
. (2020).
12. Parvez, S. M.
et al.
Health consequences of exposure to E-waste: an updated systematic review.
Lancet Planet. Heal.
5, 905–920 (2021).
13. Aguilar-Hernandez, G. A., Singhvi, A., Böcher, C., & Zhong, X. Building resilience in high-tech supply
chains.
Nature Electronics
, 6(8), 546-548 (2023).
14. Luo, Y., & Van Assche, A. Techno-geopolitical uncertainty and international business: Implications of
the United States Chips and Science Act.
Journal of International Business Studies
(2023).
15. Kolias, K., Hahladakis, J. & Gidarakos, E. Assessment of toxic metals in waste personal computers.
Waste management
. 34, 1480-1487 (2014).
1. Lein, P. & Bruun, D. The toxicological implications of E-waste.
Open Access Government
. 39. 448-449
(2023).
17. Nuss, P. & Eckelman, M. Life Cycle Assessment of Metals: A Scientic Synthesis. PloS one. 9.
e101298 (2014).
1. Zhao, H.
et al.
A New Circular Vision for Electronics Time for a Global Reboot,
World Economic
Forum
. (2019).
19. Walden, J., Steinbrecher, A. & Marinkovic, M. Digital Product Passports as Enabler of the Circular
Economy.
Chemie Ingenieur Technik
. 93 (2021).
Methods
We develop a dynamic model to predict the amount of global GAI-related E-waste. We consider only the
servers used for LLMs computing inside date center, ignoring servers for other purposes and accessary
modules. The estimation of LLMs proliferation is the basis of model. Differs from the simple exponential
estimation, we use training-data-size limitation (i.e., the sentences for training the model are nite) as the
constrain to set limits for its development. Aiming at the server discarding amount, we rst predict the
global new server deployment amount and stock amount. Then, the E-waste amount can be predicted on
the hypothesis that the servers have a xed lifespan and are discarded at the end of this period. In the
baseline scenario, a 3-year lifespan is selected since the general lifespan of computing device is between
2-5 years. We evaluate the E-waste generated amount in 2030 and the cumulative amount between 2020
and 2030 at quarter intervals. This period is particularly appropriate for development of LLMs, as 2020
witness the rst grand commercial LLM – GPT3.0, and a decade period is remarkable for the expansion
of computer-related technology given to previous observations like cloud-services. Especially, 2030 is a
popular milestone among data center operators to realize their sustainability targets. A brief description
of the model components is given here with full details provided in theSupplementary Information.
Prediction model of LLMs-related GPU servers
Page 9/11
It is hard to gauge the accurate number of servers in data centers, as it is regarded as proprietary
information by operators. Here we use the dynamic demands of computational powers and the
performance of GPU servers to approximate the number of servers, by evaluating the evolution of LLMs
parameter scales, training dataset scale, number of LLMs model worldwide, number of daily active users,
and computational power of GPU servers. All the conguration of these factors is derived from
complementary researches and recognized theorems such as the Moore’s Law. ‘Dynamic’ means the
estimation is not done by simply dividing the total computational power demands with the individual
computational power per GPU server. The stock computational power should be considered in this
principle: [number of new servers = (new computational power demands – current computational power
stockage) / computational power per new servers]. The computational power is measured in
pfs-day
(24-
hours computing at 1 peta-ops/second). This unit is widely used to describe the scale of AI computing
tasks by OpenAI, Google, etc. For full details see Supplementary Information.
Regional distribution
AI data centers are geographically clustered. Here we assume the training and inference of LLMs will be
undertaken in the same region of model development. In this regard, three major LLMs regions are
selected: North America (USA and Canada), East Asia (China, South Korea, and Japan) and Europe
(European Union and the UK). The regional distribution proportion is calculated by counting the number
of existing LLMs in the three regions. Details of accounting are available inSection S2of Supplementary
Information.
Scenario development
We rst evaluate the baseline scenario to represent the radical case of LLMs-related E-wastes - LLMs will
be used by everyone with access to the internet (including via search engines and social platforms, e.g.,
Google, Bing, Baidu, and Facebook). Then, moderate and conservative scenarios are modeled with
assumption that LLMs has a specic yet wide-range target user (for example, Tiktok), and that LLMs
serves only to those who get used to this interaction (for example, voice assistant on smartphones). We
then explore 6 scenarios to examine to which extent the circularity strategies and technical barriers can
affect the E-waste amount. Details of all scenarios and interventions are available inSection S3of
Supplementary Information.
Figures
Page 10/11
Figure 1
The quantication framework and the Large Language Models (LLM)-related E-waste generation trends.
(a) Computational Power-driven Material Flow Analysis model, (b) Regional distribution of EoS E-waste
generation between 2020-2030, (c) LLM-related E-waste generation (in thousand tons) under optimistic,
moderate, and conservative LLM proliferation scenarios.
Page 11/11
Figure 2
LLMs-related server waste impact from different perspectives. (a) Circular economy strategies in different
lifecycle of servers, (b) Cumulative OoS E-waste amount for different scenarios, (c) Cumulative amount of
recycling materials (left) and toxic materials (right) within global obsolete servers during 2020-2030
Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.
SupplementaryInformation.docx
