Every AI-generated reply, image, or recommendation traces back to a physical machine that needs to stay cool. The cooling process usually needs water from a river, an aquifer, or a city pipe. Across Asia where data centers are booming, that fact is colliding with some of the world’s most water-insecure geographies.

The Cost of Ambition

Asia-Pacific attracted more data center investment in 2024 than any other region. Southeast Asia’s digital economy alone is expected to reach $600 billion by 2030. Much of that growth serves AI workloads, which generate more heat per server rack than older computing and demand far more cooling water to match.

The scale of that demand is no longer speculative. A  2023 study found that training GPT-3 in a US data center directly evaporated roughly 700,000 liters of clean freshwater. The study further warned that global AI demand could account for billions of cubic meters of water withdrawal within a few years, more than the annual water use of several mid-sized European countries combined.

Where the Boom Meets the Driest Grounds

Asia’s AI ambition is prominent in a way that goes beyond AI adoption. Across the continent, the boom is about processing power: data centers. 

India already extracts a quarter of the world’s pumped groundwater. Yet Indian hyperscale facilities are now clustering in Maharashtra, Karnataka, Tamil Nadu, and Telangana. Those are drought-prone states where the Krishna basin has dropped to a fifth of its usual level. 

The drought impacts are real and far-reaching. As of July 2026, 21 of Karnataka’s 31 districts and 19 of Maharashtra’s 36 were running rainfall deficits, threatening the season’s planting. One satellite study even found a statistical link between the shrinking Maharashtra groundwater and the state’s farmer suicide rate. Drought also strips topsoil, which drives crop failures and threatens food security and people’s livelihoods. 

For households, it means longer treks to fetch water for people and cattle. This particular burden falls largely on women and girls, who in some regions walk 5 to 20 kilometers a day for it. Meanwhile, unsafe substitute sources drive up diseases like diarrhoea, cholera, and typhoid.  NITI Aayog estimates roughly 200,000 Indians die yearly from inadequate safe water access even before the data center boom.

Timing That Exacerbates Inequalities 

While Asia carries an outsized share of the risk, this is not only an Asian story. A major 2026 assessment from the UN University Institute for Water, Environment, and Health estimates that global AI data centers could, by 2030, consume enough water to cover the basic domestic needs of every person in sub-Saharan Africa, over 1.3 billion people. 

The report’s authors also point to something sharper than raw volume: a pattern of extractivism. The communities living beside these facilities are almost never the ones using the AI they power. Left unaddressed, that gap just repeats an old pattern in a new form, where some places bear the cost and others collect the benefit. 

Separately, MSCI’s analysis of more than 13,000 data center assets worldwide found that roughly one in four areas could face significantly worse water scarcity by 2050 as construction keeps landing in already stressed basins. What sharpens Asia’s version of this problem is timing. Much of the continent’s data center expansion is arriving in the same decade as shrinking reservoirs and disrupted monsoon patterns. In some regions, it layers onto a groundwater deficit that will not simply refill when rain returns. 

When you add a single large facility capable of drawing millions of gallons of water a day to that condition, a data center stops being a neutral digital infrastructure. It becomes a direct competitor with nearby farms, hospitals, and households for the same limited supply.

Loopholes and Pushbacks

Once again, who profits and who bears the risk are two different entities. Singapore offers 10-year tax holidays to efficient data centers; India’s 2026 budget introduced one running to 2047 for foreign cloud providers. These are decades-long public commitments to already-profitable companies, with no equivalent commitment to nearby residents. 

Public pushback is already visible across Asia. In February 2026, more than fifty residents in Johor, Malaysia, protested a proposed data center facility. They were worried that it drawing from the same municipal supply as their homes would leave less water during the dry months. Regulators are beginning to respond too: Singapore now requires operators to gradually raise facility temperatures to 26 degrees Celsius or higher specifically to cut cooling demand. 

When Google broke ground on a gigawatt-scale Andhra Pradesh facility in April 2026, its environmental clearance did not disclose operational water use in a district with the state’s lowest groundwater levels. That kind of omission is standard, not exceptional, since most Indian operators are private companies and exempt from ESG disclosure rules.

China’s water-to-energy caps and Singapore’s Green Mark are the regional exceptions, not the norm. India, for instance, has no binding national framework for data center water use. There, 15 states compete on incentives instead. Even globally, disclosure is inconsistent. Meta reports only owned sites, Google excludes third-party facilities, and Microsoft skips site-level data. Meanwhile, Amazon publishes no total figure at all. 

A Way Forward

Governments have been far readier to guarantee water to data centers than information about that water to residents. Fixing these regulatory loopholes are, first and foremost, fundamental to any sort of just and sustainable development.

Innovations for a more environmentally friendly data center may help. Examples include closed-loop and air cooling in Europe and desalination piloted in Thailand’s Eastern Economic Corridor. Still, mandatory facility-level disclosure tied to enforceable water conditions is key to shaping a digital future that improves life on Earth instead of harms it. There might be a possibility that Asia’s AI ambitions and its water security are not inherently at odds. But that only holds if water is treated as a planning question now, while the infrastructure is still going up, rather than a crisis to be managed once it is already built.

Editor: Nazalea Kusuma

Co-create positive impact for people and the planet.

Amidst today’s increasingly complex global challenges, equipping yourself, team, and communities with interdisciplinary and cross-sectoral insights on sustainability-related issues and sustainable development is no longer optional — it is a strategic necessity to stay ahead and stay relevant.

Join Membership Partner with Us