Do you know how much water the emails you sent or the AI questions you asked today consumed?

When I found out the answer, I honestly couldn't believe it. While the UK is struggling with a severe water shortage, one surprising conservation method they've urged citizens to adopt is to "delete old emails and photos."

National Drought Group meets to address “nationally significant” water shortfall

At first glance, this seemingly strange request shines a light on an unimaginable, hidden truth about our digital lives: the existence of "invisible water."

Is Your Digital Consumption an "Invisible Waste of Water?"

It's like someone who has dropped a water bottle in the desert, but instead of picking it up, they’re shouting at people far away to stop sprinkling water there. Or perhaps, our daily digital habits are what's expanding that very desert in the first place.

Why did the British government make such a bizarre request? And how will our future lives change? This time, we'll get to the heart of the hidden water problem in the digital age.

The Hidden Cost of Digitization: The Deep Connection Between AI and Water Resources

The AI services and cloud storage we use every day consume vast amounts of water in ways we can't see. This is the tip of the iceberg floating in the digital sea.

The "Invisible Water" Consumed to Cool AI

Training AI models and the "inference" process—where AI like ChatGPT responds to our questions—requires an incredible amount of electricity. This power runs servers, which generate enormous heat. To cool this heat, data centers use massive quantities of water. As an AI expert, I was astounded by this fact myself.

Data centers' water use is generally divided into three scopes. The first is "Scope 1," which is the direct consumption of cooling water within the data center itself. The heat from servers is absorbed by chilled water, which is then sent to a cooling tower. In the tower, some of the water evaporates, releasing heat into the atmosphere. This lost, evaporated water must be constantly replenished with fresh water. Since water is more efficient than air at removing heat, it's used in many data centers. Just imagine: every one of our digital activities is evaporating a portion of the Earth's precious freshwater.

Another often overlooked aspect is "Scope 2." This is the water consumed at power plants to supply electricity to data centers. Water consumption varies by power generation method, but in the US, for example, it's estimated that about 3.1 liters of water are consumed to generate 1 kWh of electricity. The demand for AI is accelerating data center power consumption at an unprecedented rate, which in turn is causing a sharp increase in its water footprint. It's predicted that global AI water consumption could reach nearly half of the UK's total annual water usage by 2027.

Finally, "Scope 3" is the water used in the manufacturing of AI chips and servers. Semiconductor factories require large amounts of ultra-pure water, and the wastewater can contain toxic chemicals. A report by Apple stated that 99% of its water footprint is due to its supply chain. The fact that the emails and photos we casually send and receive, and our questions to AI, are triggering massive water consumption somewhere on Earth is no longer a tale from a sci-fi novel.

The Reality of the UK's "Thirst": A Long-Standing Water Shortage Exacerbated by AI Demand

The UK has long suffered from water shortages due to climate change and aging infrastructure. AI demand is now compounding this situation.

The UK's Water Infrastructure is Leaking While AI Thrives

To ensure a stable water supply, UK water companies plan to build nine new desalination plants, ten new reservoirs, and seven new water recycling facilities by 2050. However, at the same time, Thames Water alone loses about 600 million liters of water a day to leaks—an amount that would nearly negate the storage capacity of the new planned reservoirs. The outdated infrastructure and massive leak problem are like pouring water into a bucket with a hole in the bottom.

The demand for AI and data centers is making it even harder to solve this water shortage. Data center electricity demand is projected to increase six-fold over the next decade, closely tied to the spread of AI. The reality of rapidly increasing AI demand in a country where water resources are already so strained poses a profound question for all of us.

The Controversy of the "Digital Water-Saving" Request: Responsibility Shifting or a Call for a New Awareness?

The UK government's request for citizens to delete old emails and photos caused a stir. Was it simply shifting responsibility, or was it the first step toward a new awareness?

The "Responsibility Shift" Backlash

This request sparked criticism that it was "shifting the responsibility for a massive infrastructure problem and the failures of water companies onto citizens." Many felt that asking individuals to delete digital data while large-scale leaks from water pipes were left unaddressed was "hypocritical." As a systems engineer, I strongly agree with the sentiment that companies and governments should focus on infrastructure improvements. We can't build a sustainable future by relying solely on individual effort without solving the root problem.

However, at the heart of this debate lies a lack of public awareness of the "true cost" of AI. Some have suggested that government officials may not fully grasp the environmental cost of AI. We need to be more aware that behind the convenience and innovation of AI services lies a hidden cost of immense energy and water consumption. The time has come for us to confront the question: What are we "taking" from the Earth in exchange for the benefits that AI provides?

The Overlooked AI Water Footprint: A Path to Transparency and Sustainability

The "water footprint" of AI has not received enough attention until now. But to build a sustainable digital society, we must shine a light on this overlooked issue.

Some researchers estimate that ChatGPT consumes about 0.5 liters of water for every 5 to 50 prompts (questions or instructions). It’s also said that the GPT-3 model consumes about 500ml of water to generate 10 to 50 average-length responses. It's as if every time we ask AI a question, a single 500ml plastic bottle of water is consumed. While I enjoy the convenience of digital services every day, I had no idea the environmental impact was this high.

This water consumption can vary greatly depending on the energy efficiency of the data center, the energy source used, and water management practices. It's also been pointed out that the water footprint of AI models can be significantly reduced by choosing "when" and "where" to perform the training. For example, running AI in water-efficient data centers or during cooler hours can reduce water consumption. This gives us a new perspective: as users, we can indirectly contribute to environmental protection by being mindful of the time and location when we use AI services.

Fortunately, technological innovation is progressing. Microsoft announced that starting in August 2024, they will begin a new "zero water consumption" data center design that uses a chip-level cooling solution, which will avoid using over 125 million liters of water per data center annually. Such technological advances offer a great beacon of hope for AI sustainability. In the projects we tackle every day, I’m feeling the importance of integrating this "water" perspective.

Our Choice in the Digital Native Age: A "Smart Thirst" for the Future

The UK government's seemingly strange request to "delete old emails" has given us a new perspective on our "digital footprint." This event powerfully demonstrates that water shortages are not just macro-level issues like climate change and infrastructure, but are also deeply connected to our micro-level, daily digital habits.

Awareness of "Invisible Water" Paves the Way for the Future

AI is a powerful tool that enriches our lives and solves societal problems. But behind it lies a huge environmental burden that has been overlooked—the "invisible water" of the "water footprint." This fact suggests that the time has come for us to adopt a deeper ethical and responsible perspective as we enjoy the benefits of the digital age.

Building a Sustainable Digital Society with Our Own Hands

Beyond the individual choice of whether or not to delete old emails, we can encourage companies and governments to take concrete actions to reduce the environmental impact of AI and data centers. We can demand greater AI transparency, support investment in water-efficient technologies, and most importantly, be aware of the impact our own digital actions have on the planet.

This is the first step toward the "smart thirst" required in the digital native era. Our future is shaped by the choices we make today. I believe that by paying attention to the "invisible water" hiding behind our digital lives and making smart choices, we can build a sustainable future.

Follow me!