The hidden environmental cost of AI: A drain on Vietnam's resources

Professionals across Vietnam increasingly use AI tools for emails, presentations, and content creation. Few realise each interaction carries an environmental cost that strains Vietnam's already challenged power grid and water resources.

The bottle of water in every AI-generated email

Research from the University of California, Riverside in 2023 revealed a startling fact: creating a single 100-word email using generative AI consumes approximately 519 millilitres of water – equivalent to a standard bottle of drinking water. This water primarily cools the massive data centres powering these AI systems.

Vietnam currently hosts a few dozen small-scale data centres with a total national output of 182 megawatts, which is far below the 870 megawatts target set for the year 2030. However, the sector is quickly picking up the pace with multi-billion dollar investments from global tech giants like Amazon, Microsoft, and Alphabet, as well as local companies. Viettel, for example, is building 11 large-scale data centres with a total capacity of 350 megawatts.

In addition, increasing AI adoption among Vietnam’s large tech-savvy population contributes to an even greater global demand, driving more data centre construction in neighbouring countries like Singapore and Malaysia.

This surge in data centres across the world is expected to put a strain on national power grids and water supplies.

"People see AI as this magical, cloud-based technology that doesn't have real-world impacts," says Dr Scott McDonald, a lecturer in Logistics and Supply Chain Management at RMIT University Vietnam. "But the reality is that every prompt draws physical resources from somewhere in the world."

The energy equation

Each AI interaction consumes substantial electricity. According to the Electric Power Research Institute, a single AI search request uses approximately 10 times more electricity than a traditional Google search—2.9 watt-hours compared to 0.3 watt-hours.

Generating a 100-word email using ChatGPT requires about 0.14 kilowatt-hours of electricity, equivalent to powering 14 LED bulbs for an hour and a bottle of water. While it may be considered a small output individually, the collective impact across the country is significant.

Not all AI applications have the same environmental footprint. Image generation consumes significantly more resources than text—up to five times more energy.

"Users should consider the necessity of each AI interaction," advises Dr McDonald. "Using AI for complex financial analysis that would take hours manually is a good trade-off. Asking it to write a simple greeting email or a LinkedIn post? Perhaps not."

"If one in ten Vietnamese workers used AI daily for only writing emails, the annual impact would be hundreds of thousands of cubic metres of water and enough electricity to power a small urban district," the RMIT academic adds. "But we use AI for more than just writing emails. So, what is the true impact we are having on our resources?"

Vietnam's vulnerable position

The International Energy Agency projects that globally, data centres’ electricity consumption in 2026 will double from 2022 levels, reaching 1,050 terawatts  – which is slightly more than Japan's entire electricity consumption.

Overall electricity demand in Vietnam is expected to grow at approximately 10-12% annually through to 2030, with digital technologies accounting for an increasingly significant portion.

This will pose a challenge as more local larger-scale data centres come online since Vietnam already experiences periodic power shortages, especially during summer months. Hydroelectric dams—responsible for about 28% of the nation's power—are increasingly vulnerable to drought conditions exacerbated by worsening climate change. With coal generating half of Vietnam's electricity and another 15% capacity expansion underway, the country's carbon emissions from coal-fired electricity production will continue rising with higher energy consumption demand due to AI.

Finding balance in the AI era

The solution is certainly not to abandon AI technology – which offers genuine productivity benefits – but to use it mindfully.

Recommendations include:

1. Reserving AI for complex or high-impact tasks where its capabilities truly add value

2. Drafting routine communications yourself rather than offloading everything to AI

3. Avoiding unnecessary regeneration of content

4. Being especially judicious with image generation features

Tech companies are working on solutions too. Microsoft, Google, and other major providers are investing in more efficient data centres and exploring alternative cooling methods that use less water.

A notable example in Vietnam is Viettel’s 30-megawatt data centre in Hoa Lac, which boasts of using green technologies providing 30% renewable energies for its operations.

"We need a digital sustainability movement in Vietnam that recognises the hidden environmental costs," says Dr McDonald. "Just as we became conscious about plastic waste and air pollution, we also need awareness about our digital footprint to maintain Vietnam’s commitment towards net-zero emissions by 2050."

As Vietnam continues rapid digitalisation, understanding the real-world impact of our virtual tools becomes increasingly crucial.

“The next time you consider asking AI to draft a simple email, remember there is a bottle of water and enough electricity to light your desk for hours behind that interaction,” says Dr McDonald. “The most sustainable AI prompt, after all, might be the one you never send.”

----

Masthead and thumbnail image: Gorodenkoff – stock.adobe.com