Digital Addiction’s Dirty Secret: How Your ‘Cloud’ Devours Earth’s Resources

Many people imagine their data simply floats weightlessly in a magical cloud. The truth is very different. “The Cloud” is actually someone else’s computer—hundreds of thousands of machines housed in vast warehouses that consume more electricity than small cities. When you upload a photo, document, or message, it travels through physical cables to these buildings, where it is stored on hard drives spinning continuously. The scale of this infrastructure is immense; for example, Google operates over 30 data centres worldwide, while Facebook maintains dozens, and Amazon Web Services powers roughly 30 per cent of the internet from server farms the size of football fields (Shift Project, 2025).

Traditional digital platforms such as YouTube, Facebook, Instagram, and TikTok involve constant data storage, transmission, and media streaming that consume large amounts of electricity. Overall, internet data centres consume approximately 1.5 to 2% of global electricity as of 2025, with projections to rise (Brightlio, 2025; IEA, 2025).

AI’s energy demands, although currently a smaller part of total consumption, are growing quickly because of increased use and the complexity of models. Studies estimate that AI query processing consumes more energy per request than traditional searches, but it is still less energy-intensive per unit time than video streaming platforms like Netflix or YouTube. Nonetheless, AI’s growth adds significant pressure on data centre energy consumption and is thus seen as an emerging major factor, or “the new kid,” in this landscape.​

Here are five key takeaways from the article:

  1. The cloud is not weightless or virtual—it consists of massive physical data centres worldwide consuming enormous amounts of electricity and resources to store and process data 24/7.
  2. Traditional digital platforms like social media, email, and streaming services reached critical mass in energy and storage demands during the 2010s to mid-2020s, significantly impacting natural resource balances and the environment.
  3. AI, described as “the new kid on the block,” is rapidly increasing data centre energy consumption, but also offers opportunities to optimise and improve energy efficiency through smarter resource management and hardware advances.
  4. Data permanence is real: personal data such as messages, photos, and emails are redundantly stored on multiple physical servers globally and often retained long after deletion, raising privacy and environmental footprints.
  5. Addressing the environmental challenges posed by digital infrastructure requires transparency, responsible governance, renewable energy adoption, innovation in AI and data centre efficiency, and a holistic view that includes water usage, land impact, e-waste, and grid stability.

These insights reflect current 2025 research on data centre electricity consumption, digital infrastructure sustainability, and the evolving role of AI in shaping the future energy landscape.

The article suitably positions AI within the wider context of digital energy use, recognising its unique and rapidly increasing demands while underscoring the ongoing impact of traditional internet and social media platforms that dominate current energy consumption.

This balance helps readers appreciate both the historic and emerging contributors to the digital infrastructure’s environmental footprint and the urgent need for sustainable innovation across all fronts.
It is accurate to claim that much of the concern and furore regarding the dramatic and potentially threatening use of power, water, and hardware manufacturing materials by digital technologies overlooks the fact that these traditional platforms have been consuming large amounts of such resources for at least the past 15 years.

Evidence shows that digital technology’s environmental footprint, including energy consumption, water use, and raw material demand, has been growing rapidly over the past decade and a half. By around 2010-2015, the scale of data centres, server farms, and network infrastructure required to sustain social media, streaming, and communication platforms began reaching levels where their environmental impacts became substantial and globally significant.​

For example, as of 2019, digital technologies accounted for about 4% of global greenhouse gas emissions—higher than the aviation industry’s share—and this share was projected to double to around 8% by 2025. The growth in digital footprint includes both the operational energy to run devices and data centres as well as the embodied carbon in manufacturing billions of user terminals and network components.​

Thus, while AI and emerging technologies are often spotlighted as new “energy hogs,” the underlying digital infrastructure and traditional platforms have already been using “riskily copious amounts” of resources for years. Detractors or critics sometimes fail to acknowledge this historical baseline, which is crucial for understanding the cumulative effects and for realistically addressing the sustainability challenge of the digital ecosystem.

In summary, the scale of resource use and environmental pressure imposed by legacy digital technology platforms has been critical and concerning for at least 15 years, making current debates on new technologies part of a longer continuum rather than a sudden shift.

Legacy technology platforms reached critical mass in terms of energy consumption and storage demands mainly during the 2010s and early 2020s. During this period:

  • The exponential growth of social media platforms like Facebook, Instagram, TikTok, and YouTube, combined with increasing user engagement, resulted in enormous data throughput and storage needs. For example, TikTok emerged rapidly in the late 2010s and by early 2020s was responsible for a 50 million ton annual CO2 footprint globally, comparable to the total emissions of some medium-sized countries.​
  • Digital content consumption involving video streaming, photos, messaging, and browsing escalated dramatically, driving up data centre energy use to around 1.5-2% of global electricity by the mid-2020s and endangering natural resource balances such as water and carbon budgets.​
  • The widespread use of mobile devices and laptops to access high-bandwidth content spurred greater electricity demand both at the network infrastructure and user device levels, compounding the environmental impact of internet platforms.​
  • Studies estimating environmental footprints of digital content consumption show that by around 2020-2025, internet usage accounted for a significant share of global resource consumption and carbon emissions, reaching a scale that posed substantial sustainability challenges.​

In summary, the period around the late 2010s to mid-2020s marks the critical mass era when traditional digital platforms became major global consumers of energy and resources, raising urgent environmental and sustainability concerns due to their scale and persistence.

It is feasible that AI, when developed and deployed with appropriate focus and controls, can significantly streamline energy use and improve efficiency in data centres, while also helping to optimise care for the natural environment.

Current research and industry practices in 2025 highlight multiple promising approaches:

  • AI can enable smarter resource management within data centres by dynamically scheduling workloads according to real-time electricity grid conditions. For example, non-urgent AI tasks may be delayed or shifted to times when renewable energy availability is higher, reducing reliance on fossil fuels and carbon emissions.​
  • Advanced AI-driven monitoring systems can coordinate and optimise cooling and hardware utilisation, avoiding wasted energy from servers running under low load or overheating.​
  • AI models can themselves be made more efficient using techniques like sparse modelling, data reduction, and hardware-aware optimisation, which reduce the compute demand required for training and inference without sacrificing performance.​
  • Integration of AI with energy intelligence systems in data centres supports sustainability goals, helping to “bend the curve” of energy consumption through continual learning and automated adjustments.​

Nevertheless, these benefits depend critically on intentional design and governance. Left entirely to market forces without oversight, unchecked AI growth risks increasing data centre energy consumption dramatically. But when combined with coordinated policy, renewable energy adoption, and technological innovation in AI and infrastructure, AI can be part of the solution to its own environmental impact.​

In summary, AI has the potential both to drive rising energy demands and to deliver breakthrough efficiencies and environmental gains if actively managed and integrated into sustainable data centre operations.

“Debunking the Cloud: The Myth of Virtual Infrastructure

Many people imagine that their data simply floats weightlessly in a magical cloud. The truth is very different. “The Cloud” is actually someone else’s computer—hundreds of thousands of machines housed in vast warehouses that consume more electricity than small cities. When you upload a photo, document, or message, it travels through physical cables to these buildings, where it is stored on hard drives spinning continuously.

The scale of this infrastructure is immense. Google operates over 30 data centres across the globe, while Facebook maintains dozens. Amazon Web Services powers roughly 30 per cent of the internet from server farms the size of football fields. That holiday photo you took? It might be stored on a hard disk in Oregon, replicated in Iowa, and backed up in North Carolina. Nothing is floating in thin air; everything exists somewhere concrete.

Inside the Data Centres: Real Fortresses

These data centres resemble fortresses. Rows of server racks stretch hundreds of feet, with thousands of computers stacked from floor to ceiling. Massive air conditioning systems—as large as trucks—keep the machines cool. Google alone pumps millions of gallons of water daily to maintain optimal temperatures.

Power infrastructure is critical and complex. Redundant power supplies, backup generators, and large battery arrays ensure that even one second of downtime could be catastrophic. Security is extremely tight: biometric scanners, man-traps, 24/7 guards, and comprehensive surveillance protect these facilities. The network equipment inside moves data at terabits per second through fibre-optic cabling. Gaining access to a data centre is harder than entering a bank vault.

The Energy and Scale Costs of Your Digital Footprint

The global impact of these data centres is staggering. Facebook stores over 100 petabytes of data (one petabyte equals one million gigabytes). Google processes 3.5 billion searches daily, all logged and stored. AWS supports 30 per cent of internet traffic. Data centres consume between 1 and 2 per cent of the world’s electricity—an amount equivalent to the energy usage of entire countries. A single email produces around 4 grams of carbon dioxide, and streaming one hour on Netflix uses 3.2 kilowatt-hours of electricity.

Your photos and files stored “in the cloud” draw constant power around the clock. This industry is far from weightless; it is among Earth’s heaviest consumers of resources.

How Artificial Intelligence Lives in the Cloud

When you interact with artificial intelligence, like a conversational assistant, your messages are processed on powerful GPU servers located in cloud data centres run by providers like AWS or Google Cloud. These GPUs, initially designed for gaming, now power complex neural networks with billions of parameters, producing responses in real time.

Your conversation travels through inference servers, is analysed by neural models, and a reply is generated and sent back—each step requiring significant computing power and energy. These conversations are typically stored temporarily for 30 to 90 days to ensure safety and quality. Though it feels instantaneous and intangible, AI depends on vast arrays of physical machines working tirelessly.

Gmail’s “Free” Storage Is Not Truly Free

Google offers free Gmail storage to billions of users worldwide, but nothing is truly free. Behind the scenes, over 30 massive data centres operate in locations such as Oregon, Iowa, Finland, Singapore, Taiwan, and Chile. Every email you send is routed, scanned for ads, encrypted, and then stored, replicated across three or more facilities.

Deleted emails do not immediately vanish; instead, they move to ‘trash’ and linger in backups for months. Google’s data centres consume enough electricity to power over 200,000 homes, demonstrating that the cost of “free” email storage is paid in energy and environmental impact.

Facebook (Meta) Platform and X Data Centres

Meta, the parent company of Facebook and X (formerly Twitter), operates a vast and rapidly expanding data centre infrastructure globally. These data centres support over 3 billion monthly active users on Facebook alone. The company is building new AI-optimised facilities, such as the upcoming Hyperion centre in Louisiana and Prometheus in Ohio, which will together provide gigawatts of computing power for AI workloads. Existing centres spread across the US (Iowa, North Carolina, Oregon), Sweden, and Singapore house thousands of servers, routers, and switches.

Meta leases much of its data centre space but also owns and builds massive campuses that consume enormous amounts of electricity and water. Security is stringent with advanced controls, and data is replicated across multiple sites for resilience. Services like Facebook messages and posts are stored indefinitely, analysed for ad targeting, and thematic insights. Deleted messages often remain in backups for months. The scale and power usage of Meta’s data centres are comparable to small cities due to their size and 24/7 operations.​

Facebook Messages: Stored Forever

Facebook messages are stored in Meta’s data centres situated across Iowa, North Carolina, Oregon, Sweden, and Singapore. These messages are replicated across multiple centres for reliability. Unlike some services, Facebook does not provide end-to-end encryption, so messages can be accessed and read by Meta.

The data is analysed for ad targeting, content moderation, and machine learning. Even messages you delete remain in backups for months. Your oldest embarrassing chats still exist somewhere, spinning on hard drives.

The Journey of Instagram Photos

When you post a photo on Instagram, it quickly travels through cables to the nearest Meta data centre. It is then processed—three or more copies are created (original, compressed, thumbnail), scanned for faces and locations, and saved on servers in multiple data centres globally.

Your photo is replicated across continents, stored on several hard drives, and delivered from the content delivery network closest to the viewer. Instagram photos are not floating in a cloud; they exist concretely, spinning physically on servers in places like Iowa.

TikTok Data Storage and Infrastructure

TikTok operates via a combination of data centres worldwide and third-party cloud providers to meet high traffic demands. The service stores user-generated content, including videos, comments, and messages, in multiple replicated locations to ensure availability and quick delivery. TikTok’s parent company, ByteDance, invests heavily in AI infrastructure housed in these data centres to support content recommendation algorithms and video processing. Data security is robust, with strong monitoring and encryption, but user data is kept in physical servers similarly distributed globally, not in an abstract “cloud” [general industry knowledge].

WhatsApp Data Centres and Message Storage

WhatsApp, owned by Meta, uses Meta’s global data centre infrastructure for message storage and delivery. Unlike the Facebook platform, WhatsApp employs end-to-end encryption, which means only the communicating users hold the encryption keys, providing a higher privacy level. Messages are stored transiently on servers only as long as needed for delivery and backup but can be encrypted backups stored locally by users or within Meta’s systems, depending on user settings.

WhatsApp backup data may reside in cloud storage providers like Google Drive or iCloud for users who enable backups, but the fundamental messaging operates on Meta’s physical servers worldwide. Despite encryption, metadata such as message timing, user contacts, and usage patterns are stored and analysed for operational and security purposes [industry practices consistent with Meta policies].

YouTube Data Centres and Infrastructure

YouTube, owned by Google, relies on Google’s global network of over 30 data centres distributed worldwide, including major facilities in the US, Finland, Taiwan, and Chile. These centres host billions of videos, user uploads, live streams, and comments. Google’s data centres use advanced server hardware and cooling technology to manage YouTube’s enormous data and bandwidth demands. Videos are stored redundantly across several centres to ensure availability and fast streaming performance globally. YouTube content contributes significantly to Google’s massive data footprint, consuming substantial electricity continually to serve billions of users every day.

Snapchat Data Infrastructure

Snapchat operates data centres and cloud resources to support its real-time multimedia messaging service. User snaps, stories, and chats are stored redundantly across multiple data centre locations to enable fast delivery and backup. Data is encrypted in transit and at rest, but like other platforms, it physically resides on hard drives operated continuously. Snapchat utilises AI for content moderation and ad targeting, increasing demand for computational power. Its data centres are located in strategic global hubs, sharing infrastructure with parent company Snap Inc. and cloud providers.

Telegram Data Storage and Hosting

Telegram uses a hybrid approach combining its own global data centres and cloud service providers to store and process messages, media, and files. Telegram emphasises security with encryption available for private chats and messages, yet stores non-secret messages on servers to allow fast syncing across devices. User data is distributed among servers located primarily in Europe and Asia. Telegram applies advanced compression algorithms, replication techniques, and data management policies to optimise storage and speed while maintaining operational scalability.

WeChat Data Centres and Operations

WeChat, operated by Tencent, depends on Tencent’s extensive data centre network in China and abroad to support over a billion monthly active users. These data centres store chat histories, multimedia content, transaction data for its integrated payments, and social media posts. Data replication across several data centres guarantees availability despite outages. Tencent invests heavily in AI and big data processing capabilities in its server farms for content recommendation, moderation, and advertising. Due to privacy regulations, WeChat data storage is predominantly located within China, but it also uses cloud providers for some international operations.

These platforms, like Facebook, Instagram, WhatsApp, Google, and Amazon services, maintain real, power-consuming physical infrastructures. They distribute data redundantly worldwide to balance speed, reliability, and disaster recovery. The commonly held myth that data exists in an intangible “cloud” is dispelled by the reality of vast, energy-intensive data centres running 24/7.

The Future of Data Storage and Energy Efficiency

Emerging technologies offer improvements, but do not change the physical reality of data storage. Concepts like underwater data centres use natural ocean cooling, AI-optimised chips reduce energy consumption, and experimental DNA storage promises incredibly dense data archives. Edge computing processes data locally for speed and privacy. Quantum storage could someday exponentially expand capacity.

Companies invest billions in green energy to power data centres from wind and solar farms, aiming for greater efficiency and reduced environmental impact. However, data will always need physical machines and power. Every post, email, message, and photo contributes to humanity’s largest physical archive. Your digital life has weight, and it lives somewhere real.

Data Scope

This article excludes detailed analyses of certain critical environmental and social issues related to data centres and digital infrastructure, which merit separate and focused attention:

  • Water Consumption and Resource Usage: Data centres consume vast quantities of water, often in water-scarce regions, primarily for cooling. This extensive usage can exacerbate local water stress and raises concerns about equitable resource allocation and long-term sustainability.
  • Land Use and Urban Impact: The physical footprint of large data centres requires substantial land, contributing to urban sprawl and possible ecosystem disruption. Careful planning is essential to mitigate these effects.
  • Climate and Carbon Emissions: Data centre operations contribute significantly to greenhouse gas emissions, particularly as demand grows with AI and digital service expansion. Addressing these emissions requires integration of renewable energy and improved efficiencies.
  • E-waste and Hardware Lifecycle: The frequent hardware upgrades generate growing amounts of electronic waste, a challenge for recycling and environmental health.
  • Grid Reliability and Energy Security: High and fluctuating energy demands from data centres stress power grids, creating challenges for infrastructure resilience and energy supply consistency.

Addressing these interrelated environmental, social, and infrastructural challenges involves embracing holistic strategies, including energy-efficient technologies, sustainable resource management, circular economy principles, and proactive urban development policies. Transparency and corporate responsibility in digital infrastructure growth will be critical for balancing technological progress with environmental stewardship.

This article’s scope is focused on energy consumption, data storage permanence, and user privacy within cloud computing and AI contexts and does not encompass the broader environmental and social impacts listed above.

Conclusion

The myth of “the cloud” as a weightless, intangible space conceals the very real, physical infrastructure powering our digital lives. Data centres spanning the globe consume vast amounts of energy and resources to store, process, and deliver every photo, message, email, and AI interaction. Traditional digital platforms like social media, streaming services, and communication networks have, over the past 15 years, reached critical mass in their energy demands, resource consumption, and environmental impact—often hidden beneath the surface of daily use.

Meanwhile, artificial intelligence emerges as “the new kid on the block,” rapidly increasing demand while holding promise for innovation that can optimise energy use and improve data centre sustainability. Achieving this balance requires intentional design, governance, and investment in renewable energy and efficiency technologies.

This article underscores the urgency to recognise the environmental footprint of digital infrastructure—its energy consumption, data permanence, and user privacy implications—while acknowledging broader challenges such as water use, land impact, e-waste, and grid demands. Sustainable progress depends on transparency, responsibility, and holistic strategies that integrate technology, policy, and resource stewardship.

Ultimately, the digital world is not weightless or virtual in the physical sense. It is grounded in massive, power-hungry installations that shape our environmental future. Our digital lives have weight, demand real resources, and call for mindful innovation to ensure that technology serves not just convenience and connectivity, but also the health of our planet.

References:

With sincere thanks to

https://www.instagram.com/entangletechh/p/DQLJQ_oEslc

4E Technology Collaboration Programme, International Energy Agency, 2025. Data Centre Energy Use: Critical Review of Models and Results. [online] Available at: https://www.iea-4e.org/wp-content/uploads/2025/05/Data-Centre-Energy-Use-Critical-Review-of-Models-and-Results.pdf [Accessed 6 November 2025].

International Energy Agency, 2025. Global Energy Review 2025. [online] Available at: https://iea.blob.core.windows.net/assets/5b169aa1-bc88-4c96-b828-aaa50406ba80/GlobalEnergyReview2025.pdf [Accessed 6 November 2025].

Carbon Brief, 2025. AI: Five charts that put data-centre energy use and emissions into context. 16 September. [online] Available at: https://www.carbonbrief.org/ai-five-charts-that-put-data-centre-energy-use-and-emissions-into-context/ [Accessed 6 November 2025].

Ember, 2025. Global Electricity Review 2025. April. [online] Available at: https://ember-energy.org/app/uploads/2025/04/Report-Global-Electricity-Review-2025.pdf [Accessed 6 November 2025].

Pew Research Center, 2025. What we know about energy use at US data centers amid the AI boom. 23 October. [online] Available at: https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom/ [Accessed 6 November 2025].

Surfshark Research, 2020. How much energy does generative AI use? 8 December. [online] Available at: https://surfshark.com/research/chart/chatbots-energy-consumption [Accessed 6 November 2025].

Pawprint, 2020. What’s the carbon footprint of an email? 4 March. [online] Available at: https://www.pawprint.eco/eco-blog/carbon-footprint-email [Accessed 6 November 2025].

Peak Nano, 2025. IEA Electricity Mid-Year Update 2025. 15 September. [online] Available at: https://www.peaknano.com/blog/iea-electricity-mid-year-update-2025 [Accessed 6 November 2025].

Epoch AI, 2025. How much energy does ChatGPT use? 6 February. [online] Available at: https://epoch.ai/gradient-updates/how-much-energy-does-chatgpt-use [Accessed 6 November 2025].

The Carbon Literacy Project, 2024. The Carbon Cost of an Email: Update! 7 March. [online] Available at: https://carbonliteracy.com/the-carbon-cost-of-an-email/ [Accessed 6 November 2025].