The Digital Thirst: Unveiling the Environmental Cost of AI’s Hyper-Consumption Era

To go back to this article, go to My Profile and then click on View saved stories.

The Insatiable Energy Appetite of AI: Entering the Age of Excessive Internet Use

Currently, the presence of generative artificial intelligence on the internet is overwhelming. Whenever you perform a search on Google, you might find yourself faced with a summary crafted by AI at the very beginning of the results. Likewise, while navigating through Facebook, Meta’s AI feature could catch your attention. Moreover, that persistent sparkle emoji still lingers in my thoughts.

The eagerness to incorporate artificial intelligence into a wide array of digital communications has its roots in the groundbreaking launch of ChatGPT by OpenAI towards the end of 2022. This sparked a fascination with generative AI across Silicon Valley, and nearly two years on, tools driven by advanced language algorithms have become a staple in how users interact online.

A negative consequence of this expansion is that the computational tasks needed to operate generative AI systems demand significantly more resources. Consequently, we've entered a new phase of the internet characterized by the emergence of a computing style that requires vast quantities of electricity and water for both its construction and functioning.

"According to Sajjad Moazeni, a researcher in computer engineering at the University of Washington, the underlying mechanisms powering generative AI models significantly diverge from what we've seen with classic services like Google Search or email. He points out that the latter required comparatively minimal data exchanges between the processors. On the other hand, Moazeni believes that the computational demand for generative AI tasks is substantially higher, potentially 100 to 1,000 times greater."

The once-overlooked energy consumption required for developing and implementing generative AI has come to the forefront, with numerous specialists last year forecasting spikes in power usage at data centers involved in AI development. In a revealing move, Google has recently abandoned its claim of being carbon neutral, while Microsoft risks compromising its environmental objectives in the fervent pursuit of creating superior AI technologies.

"According to Junchen Jiang, a networked systems expert at the University of Chicago, there's a direct correlation between the computational workload and both the energy use and carbon emissions of data centers, as their operation scales with the computational tasks they perform. As AI models grow in size, they demand significantly more computational power, leading to these cutting-edge models reaching unprecedented scales."

Despite Google's energy use doubling between 2019 and 2023, Corina Standiford, a representative for Google, argued it wouldn't be accurate to claim that the company's energy usage surged in the midst of the AI competition. She explained in an email that it is significantly difficult to lower emissions from their suppliers, which constitute 75 percent of their environmental impact. Standiford pointed to the production companies of servers, network gear, and additional technological components for data centers as the primary sources of this issue. These elements are essential for developing the hardware needed for advanced AI models and are known for being highly energy-demanding.

Although the demand for energy in data centers is on the rise, it represents only a minor fraction of the total energy consumption by humans. Fengqi You, who specializes in energy systems engineering at Cornell, points out that currently, oil refineries, buildings, and the transportation sector have a significantly larger impact on energy usage. "At this moment, those areas consume a lot more energy than AI data centers," he notes. Nevertheless, as generative AI technologies become increasingly embedded across various online platforms and gain wider acceptance among internet users, the energy usage attributed to AI is expected to escalate in the foreseeable future.

Significant Water Consumption

Beyond their substantial energy consumption, the facilities responsible for developing and running generative AI algorithms utilize vast amounts of water, reaching into the millions of gallons.

"According to Shaolei Ren, a researcher focused on responsible AI at UC Riverside and one of the authors of 'Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models,' the quantity of water accessible for human use is quite scarce, encompassing merely the fresh water found on the surface and underground. Ren points out that the operation of data centers significantly contributes to water scarcity by releasing vast amounts of water vapor into the atmosphere."

At first glance, it might seem that the environmental toll from large-scale data centers and the daily habits of local residents, such as frequent bubble baths or leaving the tap on while brushing teeth, are on the same level. However, this is not the case. “Their usage patterns are quite distinct from those of typical household consumers. In a residential context, when we use water from the public supply and then immediately release it back into the sewer system, we are essentially just borrowing the water—we're not really using it up,” explains Ren. “On the other hand, data centers draw water from these utilities and then effectively vanish it into the air, contributing to atmospheric moisture.” According to Ren, the water utilized by data centers might not reappear on the ground for up to a year.

In an email communication, Alistair Speirs, who holds the position of senior director for Azure global infrastructure at Microsoft, highlights the role of AI in the expansion of data centers. He also emphasizes the shift towards cloud computing as a significant element to take into account. Speirs mentions, "This may cause the expansion to appear rapid, as it largely involves substituting hardware that was once managed in-house." He further notes that Microsoft is on track to achieve its environmental objectives of reducing its carbon footprint, enhancing water sustainability, and eliminating waste by the decade's close.

Cornell's own Fengqi You stresses the need for ongoing shifts towards renewable energy, while casting doubt on the effectiveness of carbon offset strategies adopted by corporations for sustainability. "Offset schemes serve as a stopgap measure – they're a step in the right direction, yet far from a lasting resolution," he states. His views extend to water conservation initiatives, acknowledging them as a positive, albeit limited, step. You advocates for a broader focus on the water usage throughout the supply chains of major businesses in addition to their immediate water use.

Certainly, Google and Microsoft aren't the sole giants competing in the artificial intelligence arena. When reached out to via email, Meta's representative, Melanie Roe, inquired for additional details regarding the story but did not follow up on subsequent communications. OpenAI, on the other hand, did not respond to inquiries for a statement.

Influential Forces

Instead of being detrimental to the environment, tech firms are increasingly framing the advancement of artificial intelligence (AI) as a key component in tackling climate change and driving innovation. To mitigate the direct environmental and financial costs associated with AI, experts are exploring novel methods to decrease the energy consumption required for AI development. This includes utilizing more efficient computing chips and investigating the feasibility of compact AI models that demand lesser computational resources.

Expanding beyond just ecological issues, these computing facilities could potentially strain and even overpower the energy supplies of local electricity networks due to their high demands. Moazeni points out, “Take, for example, the Microsoft data center in Quincy, Washington. It's widely acknowledged that the amount of electricity it consumes is monopolizing the region's energy supply.” Internationally, the massive complexes housing servers for AI development and operations might find themselves in competition with local communities and enterprises for electricity, which could result in power outages during times of high demand.

In an email statement, Microsoft's Vice President of Energy, Bobby Hollis, emphasizes that the company collaborates with relevant officials and utility providers to ensure local services remain unaffected. He asserts that Microsoft constructs additional facilities to prevent any decrease in utility services for the community.

Individuals aiming to be mindful of their energy usage might feel overwhelmed. Without actively searching for generative AI applications, avoiding them becomes a challenge, as they have become integrated as standard components within operating systems, web applications, and routine software utilities. Whether accessing a digital workplace or merely browsing the web to stay in touch with peers, navigating without encountering numerous chatbots that provide condensed information and assure boosts in efficiency is nearly unachievable.

Despite the pervasive presence of artificial intelligence in our daily digital interactions, its expansion into further aspects of our virtual existence seems inevitable. However, the maximum potential impact this growth may have on power and water resources is yet to be determined.

Suggested for You…

Delivered to your email: Will Knight delves into AI progress with Fast Forward.

Delving into the most extensive undercover FBI operation ever executed

The WIRED Artificial Intelligence Elections Initiative: Monitoring over 60 international electoral events

Ecuador finds itself completely at the mercy of drought conditions

Be confident: Here are the top mattresses available for online purchase

WIRED DISCOUNTS

Enjoy Savings of $7 to $23 on Turbo Tax Online

Receive a 20% Discount on All 2024 Tax Software from H&R Block with This Coupon

Get a $15 Discount on Your Purchases Using the Instacarter Promo Code

2024 Half-Price Student Offer on Door Dash Dash Pass

Special Offer: Kids' Hoodies & Joggers Two for $40 at Finish Line

Groupon Promo: Additional 15% Discount on Orders of Any Size

Additional Coverage from WIRED

Evaluations and Tutorials

© 2024 Condé Nast. Rights reserved. WIRED might receive a share of revenue from items bought via our website, thanks to our Affiliate Programs with retail partners. Content on this website cannot be copied, shared, broadcast, stored, or used in any form without explicit consent from Condé Nast. Advertising Choices

Choose a global website