Investment analysis on India is not hard to find. Analysis that treats water as an actual variable in the model is rare. Not a sustainability section, not an ESG checkbox, but a genuine attempt to ask whether the places where capital is going have enough water to support what is being built there. That gap is what this piece is about.

What the reservoir data shows

As of May 14, 2026, India's 166 monitored reservoirs held water at 34.45% of total capacity. Two weeks earlier they were at 38.72%. Eight billion cubic metres disappeared in fourteen days, in summer, before the monsoon. Three dams reached zero storage in that period:

  • Chandan, Bihar
  • Bhima Ujjaini, Maharashtra
  • Maudaha, Uttar Pradesh

Not low. Zero.

About 70% of India received below-normal rainfall in the first quarter of 2026. The IMD's monsoon forecast for this year is 92% of the long-period average. The thing India is counting on to refill all of this is arriving below par.

This is not a new trend. India extracts roughly 250 billion cubic metres of groundwater annually, around 25% of global extraction, from a country holding 4% of the world's freshwater and 18% of its population. The Central Ground Water Board classifies 26% of India's groundwater blocks as over-exploited, critical, or semi-critical, across parts of Punjab, Haryana, Rajasthan, Karnataka, Tamil Nadu, and Delhi. The places where the economic activity is concentrated.

The data centre problem in Bengaluru

The National Compilation on Dynamic Ground Water Resources of India 2025 records groundwater extraction in Bengaluru's most economically active districts at 177% of annual replenishable limits. Net availability for future use in those districts is effectively zero.

The city has 31 operational data centres consuming around 20 million litres of water per day. Most use evaporative cooling systems where roughly 80% of the water used evaporates rather than being recycled. The clusters are concentrated in Whitefield and Devanahalli. Devanahalli has no perennial water source, runs entirely on groundwater extraction at 169% of replenishable limits, and Karnataka approved at least eight data centres there between 2013 and 2025.

The policy record:

  • In 2022, Karnataka introduced a Data Centre Policy promising investors uninterrupted water supply, with no drought contingency clause
  • In March 2026, IT Minister Priyank Kharge told the state assembly that hyperscale data centres may not be suitable for Bengaluru given water constraints

Significant capital was committed in the gap between those two positions. The 2026 statement is not a policy revision and is not binding on any approved project.

Meanwhile the central government moved the other way. The Union Budget 2026 extended a tax holiday until 2047 for foreign companies building data centres in India. India's data centre water consumption is projected to grow from 150 billion litres in 2025 to 358 billion litres by 2030, with most capacity being added in Mumbai, Hyderabad, Chennai, and Bengaluru.

  • Hyderabad has a projected water deficit of 870 million litres per day by 2027
  • Chennai's four major reservoirs fell below 1% capacity in 2019

The buildout is concentrated in the four cities most exposed to water stress.

The manufacturing exposure

The same constraint runs through manufacturing less visibly. Textiles are the clearest case. Growing one kilogram of cotton requires around 22,500 litres of water. Dyeing and processing consume 1.6 million litres per tonne of output. During the Bengaluru water crisis in 2024, production costs for manufacturers rose by at least 10% as borewells ran dry and companies shifted to commercial tankers. The Karnataka Small Scale Industries Association warned that small textile processing units, each needing around 100,000 litres per day, might have to refuse new orders.

Steel and thermal power depend on river water for cooling. The Council on Energy, Environment and Water found that 11 of India's 15 major river basins will experience water stress. Five, including the Ganga, are already in water scarcity. Three, including the Cauvery, are in absolute water scarcity, below 500 cubic metres per capita per year. Most thermal plants sit along these basins.

In December 2025, Moody's flagged water stress as a risk to India's sovereign credit strength, specifically naming coal power generators and steel manufacturers. When water stress appears in the same paragraph as sovereign credit, the point being made is financial.

Why it stays invisible in underwriting

Water stress does not arrive as a single identifiable event. What happens is a slow accumulation that looks, quarter by quarter, like ordinary operational variation. A company shifts from borewell extraction to commercial water procurement. Costs rise 8 to 10%. Management calls it temporary. A year later the borewells have not recovered and the tanker arrangement is the permanent cost structure. Water recycling infrastructure gets added to the next capex plan that did not originally budget for it.

None of this triggers a formal write-down. It aggregates into margin compression over the holding period, easy to attribute to other causes because other causes are always available. This is why the risk stays invisible in underwriting. It will not surface in the deal until the deal is already done.

The overlap problem

India's macro growth story is told in national aggregates: GDP growth, demographic dividend, domestic consumption, export opportunity. These numbers are real. They also obscure a geographic concentration problem. The places attractive for investment tend to be exactly the places where water has been extracted most aggressively over decades of growth. Karnataka has no net groundwater available for future use in its most productive districts. Punjab has been classified as critically over-exploited for over a decade. Tamil Nadu, Haryana, and large parts of Rajasthan and Uttar Pradesh are in the same category.

Estimates of the long-run GDP impact from water scarcity range from 6% to over 14% by 2050. The range is too wide to treat as precise. The direction is not in dispute.

The policy lag problem

The obvious counterargument is that policy responds to crises, that India has a Jal Shakti Ministry, that desalination and recycling technology is improving. Probably true over a long enough horizon. Not a reassuring argument for capital allocated in 2026 on a five-to-seven-year hold.

The Karnataka sequence is the illustration. A 2022 policy promising uninterrupted water with no contingency clause. A 2026 assembly statement walking back the premise. Capital committed in between, based on the 2022 signals. The governance response lagged deployment by four years, arrived as an informal statement, and is binding on nothing. That is roughly how regulatory responses to resource constraints work: reactive, slower than the capital that precedes them, triggered by consequences already visible.

A five-year horizon from 2026 exits around 2031. The water data for that period is not a speculative forecast. It is the forward projection of trends measurable today.

The point

I grew up in Rajasthan. Water scarcity there is not an environmental concern in the abstract. It is a daily logistics problem. The idea that sophisticated capital allocation frameworks are underwriting multi-year holds in water-stressed geographies without a water line in the model is, from that vantage point, genuinely strange.

This is not an argument against investing in India. The market size is real, the growth runway is real, and the digital infrastructure buildout has no parallel in comparable economies at this stage. The geographic and sectoral overlap between where capital is going and where water stress is most acute is not being modelled. That gap will close, either because someone starts modelling it proactively or because the operating cost data makes it impossible to ignore.

Three dams hit zero storage in May 2026. The data centres in Devanahalli are becoming operational the same year. At some point those two facts need to appear in the same analysis.