Cloud-Storage Unit: Generative AI and its Environmental Impact

Introduction

Artificial intelligence and its closely related programs have occupied major conversations surrounding popular culture within the past year. The research and compartmentalization capabilities it offers have completely revolutionized operations in a large number of industries. This newfound concept has raised mixed feelings of apprehension and curiosity in relation to human function. In AI’s early stages, it is difficult to decipher what long-term impacts these programs will have on humans and the way they operate. As of now, the music industry is a particular epicenter of conversations surrounding the future of AI within art and creation, with no clear answer on the matter. What is more tangible however, are the environmental effects when these programs are in use. This article will delve into the usage and storage of the data produced by generative AI and the subsequent environmental impacts of such programs.

A Background on Data

To understand the environmental effects of AI, it is important to note exactly how the internet operates. The internet’s primary component is made up of data.^[1] Everytime the internet is used, whether that be a question in a search engine, or opening an application, data is retrieved to provide these results.^[2] Data is also produced when interacting with these websites and applications, such as uploading content.^[3] Although not a material item, this data needs to be stored. Typically, this data is stored on physical servers solely dedicated to hosting this information.^[4] Data centers act as warehouses for these physical servers, and are placed throughout the world, operating constantly to accommodate the widespread use of the internet.^[5] Reforms in data management are constant, but at a baseline, all data needs to be stored physically.^[6] Data and its relationship with generative AI spark a unique conservation surrounding the importance of storage.

Forms of Storage

The storage of generative AI encompasses the specific technology and systems used to store, manage, and retrieve data used within these programs.^[7] Unlike typical data frameworks, these systems need to have the capacity to handle the significant and unique workloads of AI. High-speed data access, large-scale storage capacity, and integration capabilities with AI frameworks are non-negotiable for these systems.^[8] A simple search on an engine will connect a person with articles and resources related to the keywords typed. A simple search using Generative AI programs, however, needs to have the capability to read and write, gather information over a multitude of different sources, and create a curated answer for one specific question or prompt – all while doing the research that a person would have had to do themselves.^[9] AI workloads commonly utilize and require simultaneous access to multiple streams of data, making their storage systems a unique process of data facilitation.^[10] As mentioned before, these hoards of data still require storage at physical data centers. While a similar base structure to typical storage locations, computing power and I.T. structures differ vastly in AI data centers.^[11] For example, CPU’s, or central processing units, are more commonly used within conventional data centers, whereas GPU, or graphics processing units, are used in AI data centers, and oftentimes require more square footage.^[12] These provide advanced storage, energy, and cooling capabilities, needed solely for the sheer velocity of data from AI programs. Due to uncertainties around the rising demand, creation of these warehouses, and future expansion plans, estimating the precise size of all data centers is difficult, However, CEO of Maryland based Ciena, Hannover’s Gary Smith, remarks on the size of data centers. In an interview with The Technology Letter in December 2024, he stated “You have data centers that are over two kilometers”, noting how some centers are multiple stories, and are much more significantly sized than he had known was possible.^[13] Conversations around the sheer size creates curiosity around the location of these data centers. These warehouses are strategically placed at many locations around the globe.^[14] Notable hubs include tech development epicenters such as Silicon Valley and Northern Virginia within the United States, with the nation holding the title for the most data centers in the world.^[15] Other notable locations include European countries such as Germany and the United Kingdom, along with China leading in Asia.^[16]

Environmental Effects

These data centers within the generative AI sphere specifically do a great deal to contribute to climate change. Primarily, these warehouses have massive energy demands. Since AI data storage operates at a higher velocity, their energy use and subsequent environmental impact left behind is significant.^[17] Typically cooling systems account for a majority of these centers energy use.^[18] According to the US Department of Energy, the largest data centers that operate up to tens of thousands of storage machines, utilize 100 megawatts of energy.^[19] This is enough power to sustain around 80,000 households within the United States.^[20] These cooling systems also use significant amounts of water. For example, a 1 megawatt data center is estimated per year to use up to 26 million liters of water. This is equivalent to the yearly water use of 62 families within the United States. Furthermore, as of December 2024, the current capacity for data center wattage within the United States is 40,000 megawatts.^[21] Water usage even within recreational use of chat-box AI programs is significant. A report by the Washington Post stated that a 100-word response from ChatGPT, a leading generative A.I chat-box tool, uses 519 milliliters of water and 0.4 kilowatt-hour of energy.^[22] While estimations of environmental impacts are constantly evolving, it strikes a chord surrounding the integration of such programs within the arts.

Music’s Involvement

What does this mean for music? Opinions on this matter are subjective, and the implications of generative artificial intelligence and its impact on the environment have created divided opinions. These conversations become particularly complicated, especially when considering the crossroads between art and AI. Many musical organizations are dedicating time and resources to researching the negative environmental impacts of AI. The New Interfaces for Musical Expression particularly has been facilitating research on new technologies for performance, those of which including AI-powered instruments, and in general how to build more sustainable practices within the music industry given the recent widespread use of these systems.^[23] Others seem to embrace these technologies, Imogen Heap notably being an open supporter in recent years. Throughout the beginnings of popular AI systems, Heap had been in the process of creating her own program named “Mogen”, a voice replicating model and information center for the artist. Initially, this was a premium application for dedicated fans, where “Mogen” could generate responses and opinions on certain topics fans requested, based on a large sum of data from Heap, including past interviews and quotes. Essentially, this was a generative AI search engine and chat-box, solely for Imogen Heap. In a 2024 interview with The Guardian, she stated “Anything I’ve ever said or done, I want Mogen to have access to that,”.^[24] Recently however, she has expanded upon her use of the program, utilizing it within her music for the first time in 2024 on a featured remix of “false god” by Karin Ann. While Heap still produced the record, her vocals are completely artificially generated, having marked her first musical pursuit with AI.^[25] While the conversation surrounding the ethics of art and generative AI are ongoing, it is clear that the use of these programs utilizes significant amounts of energy and resources, in turn negatively impacting the environment. It is unclear what the long term use of these programs could mean for climate change around the world, and the development of music creation within the industry.

^[1] “How Does the Internet Work?,” accessed April 22, 2025, https://web.stanford.edu/class/msande91si/www-spr04/readings/week1/InternetWhitepaper.htm.

^[2] “Information Retrieval: The Great Search for Knowledge,” IONOS Digital Guide, December 20, 2017, https://www.ionos.com/digitalguide/online-marketing/search-engine-marketing/information-retrieval-how-search-engines-retrieve-data/.

^[3] “What Is Uploading? | Definition from TechTarget,” WhatIs, accessed April 22, 2025, https://www.techtarget.com/whatis/definition/uploading.

^[4] Edward Mercer, “What Is an Internet Data Center Server & How Does It Work?,” Chron – Small Business, January 12, 2013, https://smallbusiness.chron.com/internet-data-center-server-work-60422.html.

^[5] “Four Primary Types of Data Centers | Data Center Virtualization – Alibaba Cloud,” accessed April 22, 2025, https://www.alibabacloud.com/en/knowledge/tech/four-primary-types-of-data-centers?_p_lc=1.

^[6] “Cloud Storage vs. On-Premises Storage: A Comparative Analysis,” Calsoft Blog (blog), March 29, 2024, https://www.calsoftinc.com/blogs/cloud-storage-vs-on-premises-storage-a-comparative-analysis.html.

^[7] “What Is AI Data Storage? | Supermicro,” accessed April 22, 2025, https://www.supermicro.com/en/glossary/ai-data-storage.

^[8] “What Is an AI Data Center? | IBM,” February 21, 2025, https://www.ibm.com/think/topics/ai-data-center.

^[9] Matthew Edgar, “Generative AI vs. Traditional Search: Technical Differences,” Matthew Edgar (blog), March 3, 2025, https://www.matthewedgar.net/generative-ai-vs-traditional-search-technical-differences/.

^[10] “What Is AI Data Storage?”

^[11] “Storage Technology Explained: AI and Data Storage | Computer Weekly,” ComputerWeekly.com, accessed April 22, 2025, https://www.computerweekly.com/feature/Storage-technology-explained-AI-and-the-data-storage-it-needs.

^[12] Ben Atherton, “The Square-Foot Race for AI Space: Physical Requirements for AI Data Centers,” AFL – Hyperscale solutions, October 21, 2024, https://www.aflhyperscale.com/articles/physical-requirements-for-ai-data-centers/.