Should insurers insure something that is certain to worsen their claims?
Every time you ask an AI assistant a question, somewhere in a building you will never visit, a server is working on your behalf. That building is a data centre. And India is about to build a great many of them.
India’s data centre capacity has been growing at over 20% annually and is expected to reach 2,000 MW by 2027, up from around 900 MW in 2023, according to JLL’s 2026 Global Data Centre Outlook. The government’s Digital India mission, the rapid spread of fintech, healthtech, and edtech, and now the AI boom have all created voracious demand for physical computing infrastructure. Global technology giants like Google, Microsoft, Amazon and Meta have collectively pledged over $30 billion in Indian data centre investments in the last two years alone.
This is, by most measures, a story of economic progress. Data centres create jobs(?), attract foreign investment, underpin digital services that millions of Indians depend on daily, and position India as a serious player in the global technology economy.
But data centres are not weightless. They are among the most resource-intensive structures ever built, and as they scale to handle AI workloads, their appetite for electricity and water is growing faster than the infrastructure built to supply either.
The UN University Institute for Water, Environment and Health has quantified what this means globally: training a single large AI model requires as much energy as five cars over their entire lifetimes. The carbon, water, and land footprint of generative AI, across data collection, model training, and daily inferencing, is already measurable at a national scale, and growing at 36% annually as hyperscaler capital expenditure on AI infrastructure is forecast to exceed USD 600 billion in 2026 alone.
The question India needs to ask, before the infrastructure is locked in for decades, is whether the benefits of this buildout are being weighed honestly against its costs.
The environmental cost
The consequences are already visible. On 21 May 2026, US Congresswoman Alexandria Ocasio-Cortez walked into a congressional hearing carrying two glass jars of thick, visibly brown water, drawn from the taps of residents in Newton County, Georgia, living next to Meta’s Stanton Springs AI data centre. That single facility consumes approximately 10% of the county’s total daily water use, and the area is projected to be in water deficit by 2030.
Loudoun County, Virginia, the world’s most concentrated cluster of data centres, faced months-long drought conditions in 2024 triggering mandatory water-use restrictions. Data centres’ share of total water consumption in the Washington DC metro area is projected to grow from 8% in 2025 to 25% by 2035.
These are not the problems of a distant country. India is now building at the same scale, and the local conditions are, in many ways, more fragile.
Construction of Google’s planned $15 billion data centre hub in Andhra Pradesh is progressing rapidly, but growing opposition from environmental groups over water supplies and wildlife is creating hurdles for the US technology giant’s biggest-ever investment in India. A hillside above the site has already been stripped to red earth and terraced into steps.
The project faces several legal challenges over its impact on water supplies and its proximity to the Kambalakonda Wildlife Sanctuary, located just 860 metres from the construction site.
The Andhra Pradesh government, which has called the project “historic and transformational,” denies that rural or residential water will be used. But the objections reveal that Visakhapatnam already faces water shortages, rationing is common in a city of 2.5 million people and environmental group Green Visakha has questioned how the government can promise “guaranteed” water supplies to the project for 20 years in a city that already rations its water.
Three cases have been filed at India’s environmental court by the Human Rights Forum, demanding a halt. The Andhra Pradesh High Court is hearing a separate public interest litigation.
India has fewer water resources per capita than China, fewer than the US, and significantly fewer than Europe. The concentration of AI data infrastructure in water-stressed zones is a risk multiplier, not just for communities and ecosystems, but for the insurers ultimately called upon to cover the consequences.
What the insurance industry stands to lose
The insurance angle on data centres is typically framed around physical risk: fire, flooding, power failure, cyber attack. And those risks are genuinely severe. According to Swiss Re’s sigma insights 07/2026: “Insuring AI: Data Centre Value Accumulation Risks”, single data centre campuses now cost up to USD 20 billion to build, with construction costs doubling again once GPUs and equipment are installed. Over a quarter of US data centre capacity sits in locations experiencing three or more large-hail days per year. Around 40% of US capacity sits in significant-to-very-high tornado zones. Fire, driven by new lithium-ion battery backup units embedded in server racks, accounted for 42.3% of total data centre loss costs in a 15-year FM Global study, despite representing only 10.9% of loss events. Similar growth will be seen in India as the boom materialises.
Global insurance premiums tied to data centres are expected to rise from USD 10.6 billion to USD 24.2 billion by 2030.
There is, in other words, a genuine and growing insurance market here. And that is precisely what makes the strategic question interesting.
Insuring AI data centres: Hypocrisy or Wisdom?
Well, you might think, business is business. As long as the premium is paid, an insurer will insure a risk. But what if the very risk being insured is a contributing factor to the worsening claims outlook in the same region, and around the world?
Consider the logic. Insurers backing out of fossil fuel risks, as major European and Lloyd’s market underwriters have done progressively since 2018, did so not only to protect their own exposure to transition risk and stranded assets, but to accelerate the decarbonisation of the energy system. The withdrawal of insurance capacity from new coal mines and oil projects made financing those projects more expensive, and in some cases impossible. Insurance leverage, applied systemically, contributed to structural change.
Data centres present a structurally similar problem. They consume enormous quantities of water and electricity, increasingly in water-stressed geographies. They generate heat that intensifies urban heat islands. They drive electricity demand that, in markets with carbon-heavy grids like India, translates directly into additional emissions. The Yale School of the Environment estimates that data centres already contribute meaningfully to urban heat stress, the same heat events that are, according to Swiss Re’s NatCat sigma 1/2026, driving insured natural catastrophe losses up by 5–7% annually in real terms.
The climate losses insurers are absorbing today, from floods, wildfires, heatwaves, and extreme precipitation, are not disconnected from the infrastructure they are simultaneously underwriting. Water-stressed communities in Newton County, Georgia, and Visakhapatnam, Andhra Pradesh, are experiencing today what insurance loss models are increasingly pricing for tomorrow.
This is not a call to refuse coverage. It is a call for conditions.
The fossil fuel parallel is instructive because it was not binary. Insurers did not simply refuse all coverage and walk away. They imposed requirements: environmental impact assessments, disclosure of emissions targets, transition plans as underwriting conditions. The same framework could logically apply to data centres, requiring credible water usage efficiency targets, cooling technology standards, renewable energy commitments, and meaningful community impact assessments as conditions of coverage.
Some underwriters in the Lloyd’s market are already beginning to ask these questions.
The Visakhapatnam project is the largest single AI infrastructure investment on Indian soil. It will not be the last.
The questions this raises for India’s insurance industry are the same ones global underwriters are now facing. Does insuring an AI data centre in a water-stressed region contribute to the very climate and resource conditions that worsen the claims outlook? Should insurance capacity be a tool for requiring better environmental standards like with the fossil fuel transition? Is the Indian regulatory environment equipped to ask these questions yet?
References:
- Reuters – “Google’s $15 Billion India Data Centre Project Battles Water, Wildlife Concerns” (6 August 2026): reuters.com
- Yahoo Finance – “AOC Holds Jar of Brown Water, Calls for EPA Investigation into Data Center Impact on Drinking Supplies” (21 May 2026): finance.yahoo.com
- UN University Institute for Water, Environment and Health – “Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints”: inweh.unu.edu
- Swiss Re sigma insights 07/2026 – “Insuring AI: Data Centre Value Accumulation Risks” (27 March 2026): swissre.com
- Yale School of the Environment – “Data Centres, Urban Heat and AI Growth”: environment.yale.edu
- MIT News – “Explained: Generative AI’s Environmental Impact” (January 2025): news.mit.edu
- World Resources Institute – “US Data Centre Growth Impacts”: wri.org
This blog post is brought to you by the minds at insurancepe!
Got questions or doubts about anyone insurance?
Need advice or help understanding your insurance needs?
Want the best bang for your buck when buying insurance?
We got you!
Reach out to us at:
Whatsapp/Phone: 89779 18030
E-mail: contact@insurancepe.com
Visit us at www.insurancepe.com
