The environmental price of the AI boom

MANILA, Philippines — During a recent trip to Baguio, a poster inside a convenience store caught my attention. Its bubbly lettering, pastel colors and cartoon-like illustrations gave it the polished, almost too-perfect look of an AI-generated image.

“Hydration is self-care,” it said. Below it, another poster read: “Thousands have lived without love, not one without water.”

The irony was hard to miss. A reminder of water’s importance sat beside rows of bottled water, delivered through imagery that appeared artificially generated.

Similar images follow us elsewhere. At night markets, glossy burgers and spaghetti with strange textures can look more synthetic than appetizing.

Scroll through a business’s social media feed and AI-generated models, product shots and ads are becoming harder to distinguish from reality, often without the need for photographers or costly shoots. But the convenience comes with a physical footprint: the infrastructure behind AI consumes real resources.

The infrastructure behind the illusion

AI needs electricity, and electricity can mean emissions. The International Energy Agency (IEA) estimates that data centers consumed 415 terawatt-hours of electricity worldwide in 2024, about 1.5 percent of global consumption, and generated around 180 million metric tons of indirect carbon dioxide emissions from that electricity use.

Demand could reach 945 TWh by 2030, driven partly by AI. These figures cover all data center workloads, not AI alone. When electricity comes from fossil fuels, it releases greenhouse gases that contribute to global warming. Growing demand can therefore increase AI’s climate footprint unless cleaner power and energy efficiency keep pace.

Water is another hidden cost. Data centers need cooling to prevent servers from overheating, and some cooling systems consume substantial amounts of water. The IEA estimates that a 100-megawatt data center in the United States can consume around two million litres of water daily, including indirect use for electricity generation.

Bloomberg found that about two-thirds of new US data centers built or under development since 2022 were in areas experiencing high water stress. In Louisiana, Meta estimates that its planned Hyperion data center could consume more than 23 million gallons of water a day for cooling when completed around 2030, more than the daily water use of a town of 17,000 people. The concern is that industrial demand can put additional pressure on local supplies, particularly where households and agriculture also depend on them.

The footprint extends beyond power and water. Producing AI chips and servers requires raw materials, energy and water, while replacing equipment can add to electronic waste. Mining, manufacturing and disposal all carry environmental costs. AI’s impact therefore begins before a model answers a prompt and continues after the hardware that runs it reaches the end of its useful life.

Not just a Silicon Valley problem

Pax Silica puts the question on the table. The US-led initiative seeks to strengthen supply chains for semiconductors, critical minerals and technologies supporting AI. Its proposed industrial hub in New Clark City, Tarlac, has raised questions about electricity, water, land and local livelihoods.

In August 2026, the Department of Environment and Natural Resources said no project-specific environmental impact assessment had been processed because project details were still unclear. This does not establish that environmental harm has occurred, but it underscores the need for transparent planning before projects move forward.

The timing matters. As the Philippines braces for a potentially very strong El Niño, with drier months threatening water supplies, the country is also courting industries that will need electricity and water to operate. PAGASA has warned that the phenomenon could persist through the first half of 2027. El Niño is not caused by AI, but it sharpens a question policymakers cannot afford to ignore: when water becomes scarce, how will communities and industries share it?

The answer is not to abandon AI or assume every prompt is environmentally damaging. Companies can choose more efficient systems, use AI when it serves a clear purpose, disclose synthetic imagery where appropriate and avoid passing generated visuals off as accurate representations of products.

Data center developers, meanwhile, should disclose expected energy and water use, assess local resource constraints and invest in cleaner power and more sustainable cooling. Regulators must ensure environmental reviews keep pace with the industries they are meant to govern.

Back in that Baguio convenience store, the words “Thousands have lived without love, not one without water” seemed like an ordinary reminder to stay hydrated. Beside the AI-generated hydration poster, they take on another meaning. Water sustains life, but the technology increasingly woven into our lives also depends on it.

The next time we joke about skipping a bath because we have used too much AI, perhaps we should look beyond the punchline. AI does not explain every dry tap, and one prompt will not drain a reservoir. But the infrastructure powering the technology is not resource-free. As the Philippines pursues digital investment, the challenge is to enjoy its benefits without overlooking what it takes to sustain them.