India Wants the World’s AI. Who Will Pay the Power Bill?
BIZ MASALA
| SIGNATURE STORY
The data-centre boom promises investment and digital
sovereignty. But electricity, clean power and water are becoming the real tests
of whether that opportunity can be built without shifting costs to households.
By Murthy ASR
| Biz Masala | 14
October 2026

A data-centre campus depends on more than servers: it needs
reliable power, cooling, water management and grid infrastructure. Illustrative
image created for Biz Masala.
THE BIG PICTURE A proposed
National Data Centre Policy could become the rulebook for who pays for new
electricity infrastructure, how clean power is sourced, and whether
water-stressed communities get a say. The announcement is a signal of
intent—not yet a set of binding safeguards.
A $200 billion opportunity with a very physical footprint
On 14 October, Union Minister of State for
Electronics and Information Technology Jitin Prasada said India’s green-power
generation was struggling to keep pace with the requirements expected from AI
and related technologies. At a Central Electricity Authority event in New
Delhi, he also announced that the government was preparing a dedicated policy
for data centres. He warned that India did not want a situation in which the
electricity appetite of data centres left ordinary consumers paying more.
The government sees data hosting as a $200
billion opportunity, tied to investment, digital infrastructure and the
ambition to keep more of the world’s data in India. That figure is an
opportunity estimate cited in reporting—not a guaranteed investment total, a
government budget, or money already committed.
The business case is easy to understand.
Data centres house the computing power behind cloud services, digital payments,
streaming, enterprise software and increasingly AI training and inference.
Their appeal depends on dependable electricity, fast connectivity, land,
predictable approvals and competitive operating costs. But a server farm is not
weightless simply because the product it sells is digital. It occupies land,
draws power around the clock and must remove heat continuously.
The capacity curve—and the next demand curve
Official figures from the Ministry of
Electronics and Information Technology put India’s installed data-centre
capacity at about 375 MW in 2020 and around 1,575 MW in 2026. A separate
government statement citing the Central Electricity Authority estimated
data-centre power demand could reach about 17 GW by 2031–32. An earlier
parliamentary answer cited 13.56 GW for the same horizon. These estimates
differ; they should not be silently blended into a single forecast. The policy
challenge is to make assumptions, scenarios and updates transparent.

Chart: reported installed capacity, not actual electricity
consumption. The 2025 figure is rounded (“around 1,500 MW”) in the official
release. Demand estimates for 2031–32 are discussed separately because they
measure a different concept and vary by source.
What households need the policy to guarantee
The minister’s consumer-protection warning
raises a practical question: when a large data centre arrives, who funds the
substations, transmission upgrades, backup capacity and grid balancing it
requires? The answer matters even if a data centre’s own electricity tariff is
negotiated separately. Network costs can be spread through regulated tariffs,
utility investment plans and public infrastructure decisions. The distribution
of those costs depends on tariff design and regulation; a project’s arrival does
not automatically mean household bills will rise.
A credible policy could require large new
facilities to disclose expected load profiles, ramp-up schedules and their
contribution to grid-upgrade costs. It could establish transparent connection
rules, protect reliability for existing consumers, and reward projects that
bring additional clean generation and storage rather than merely claiming
renewable-energy certificates. It should also distinguish a contractual
green-power purchase from the physical reality of delivering reliable
electricity every hour.
That distinction is crucial. Solar and wind
generation can be abundant at particular times and places but variable across
hours and seasons. Data centres, especially those running high-performance AI
workloads, need dependable power. Storage, transmission, flexible computing
loads, demand response and firm low-carbon generation can help bridge the gap.
The policy should measure actual performance—emissions, clean-energy matching
and peak demand—not only promises or annual averages.
Green power is not just a procurement checkbox
India is adding renewable capacity rapidly,
but generation growth, transmission capacity and round-the-clock supply are
different challenges. A data-centre operator may sign a renewable power
purchase agreement and still rely on grid power when its contracted source is
unavailable. A more meaningful framework would disclose how much electricity is
matched with clean generation by time and location, what backup systems are
used, and how carbon emissions are accounted for.
The CEA’s long-term planning and the
government’s existing green open-access and transmission initiatives provide
pieces of the framework. The new policy’s value will depend on whether it joins
these pieces into a clear, enforceable set of expectations for developers,
power utilities and state governments.
Water: the local licence to operate

The infrastructure debate is also a local one: communities
need clear information on water sourcing, groundwater permissions, noise,
construction impacts and the reliability of local utilities. Illustrative image
created for Biz Masala; not a photograph of a specific project or protest.
Data centres use water differently
depending on climate, cooling design and operating conditions. Evaporative and
adiabatic systems may consume water; direct-to-chip and immersion cooling move
heat through liquid systems but do not automatically mean every facility has
zero water use. Closed-loop systems recirculate coolant and can reduce ongoing
water demand, though facilities may still need water for heat rejection,
cleaning or other uses. The right question is not whether a technology is
labelled “advanced”; it is how much freshwater a specific facility withdraws
and consumes, where it comes from, and what happens during drought or local
scarcity.
In August, the government told Parliament
that Indian facilities were increasingly adopting direct-to-chip liquid
cooling, adiabatic cooling, immersion cooling and closed-loop systems. It also
said groundwater extraction is governed by Ministry of Jal Shakti guidelines.
Separately, the Environment Ministry stated that data centres do not
automatically require environmental clearance as a standalone category under
the EIA notification; some projects can require clearance when covered by the
building-and-construction or township provisions. That regulatory detail
matters: a national policy should not imply that every data centre currently
passes through one uniform, facility-specific environmental review.
Reports of local opposition near planned
projects in the Mumbai region and Visakhapatnam underline the importance of
early disclosure and consultation. The concerns reported include electricity,
water, noise, wildlife and land-use questions. Companies have disputed
allegations of wrongdoing and say they use technologies to reduce resource use.
Those are competing claims, not a substitute for publicly accessible data and
independent monitoring.

Closed-loop liquid cooling and other designs can improve
thermal management, but performance must be evaluated through measured water
use, energy use and operating conditions. Illustrative image created for Biz
Masala.
Six tests for the National Data Centre Policy
|
Policy test |
What should be disclosed or protected? |
|
1. Household bill firewall |
Clear rules on who pays for dedicated grid connections and
upgrades; transparent treatment in utility tariffs. |
|
2. Clean power with proof |
Time- and
location-aware reporting of renewable matching, emissions and backup power;
avoid relying only on annual certificates. |
|
3. Grid readiness |
Require credible load forecasts, phased commissioning plans and
coordination with transmission and distribution utilities. |
|
4. Water budget |
Facility-level
disclosure of freshwater withdrawal and consumption, source, recycling,
cooling design and drought contingencies. |
|
5. Local accountability |
Public access to approvals, monitoring results, complaint channels
and emergency plans; meaningful community consultation. |
|
6. Performance, not promises |
Publish
comparable indicators such as power usage effectiveness, water usage
effectiveness and carbon intensity, with independent verification. |
The unanswered questions
The minister did not specify the policy’s
release date or detailed provisions in the announcement reported on 14 October.
The central unanswered questions are therefore straightforward: Will it be
binding or advisory? Will it apply nationally or mainly guide states? Will it
set mandatory energy and water disclosures? Will large facilities pay a defined
share of grid upgrades? And will communities be able to inspect facility-level
resource-use data? Until the draft is published, these remain tests for the policy—not
features that can be assumed to exist.
The Biz Masala Take
India should not have to choose between AI
investment and consumer protection. But it cannot secure both through
investment announcements alone. Data sovereignty is meaningful only if the
country can host data reliably, affordably and responsibly. That requires power
planning, clean generation, storage, transmission, transparent tariffs and
water rules to move at the same speed as server construction.
The policy should make one principle
unmistakable: the commercial benefits of the AI economy should not be
privatised while its electricity and water costs are quietly socialised. Data
centres can bring investment and strategic capability. Their licence to grow,
however, must rest on evidence that households and local communities are not
being asked to subsidise the infrastructure behind someone else’s digital
future.
REPORTING NOTE This article
separates official figures from projections and reported local concerns. The
$200 billion figure is described as an opportunity estimate, not realised
investment. Official sources published different estimates of data-centre power
demand for 2031–32 (13.56 GW and about 17 GW); the discrepancy is disclosed
rather than reconciled without evidence. Images are illustrative, generated for
this story, and do not depict verified locations or events.
Sources and further reading
·
Reuters, 14 October 2026 —
India to roll out data centre policy as concerns grow over power use —
https://www.reuters.com/world/india/india-roll-out-data-centre-policy-concerns-grow-over-power-use-2026-10-14/
·
Moneycontrol, 14 October 2026 —
India to roll out data centre policy to shield consumers from higher power
bills —
https://www.moneycontrol.com/artificial-intelligence/india-to-roll-out-data-centre-policy-to-shield-consumers-from-higher-power-bills-article-14051903.html
·
Ministry of Electronics &
IT / PIB, 6 August 2026 — Government highlights rapid expansion of data centre
capacity —
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2295483&lang=2®=48
·
Ministry of Electronics &
IT / PIB, 12 August 2026 — Sustainable growth of data centres through
energy-efficient and clean-energy measures —
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2298108&lang=1®=48
·
Ministry of Environment, Forest
and Climate Change / PIB, 6 August 2026 — Effective use of water in AI data
centres —
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2295675&lang=1®=6
·
Ministry of Electronics &
IT / PIB, 13 March 2026 — Data-centre capacity and estimated electricity demand
—
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2239616&lang=2®=48
·
The Hindu BusinessLine — Green
power generation struggling to keep pace with AI growth, says Jitin Prasada — https://www.thehindubusinessline.com/info-tech/green-power-generation-struggling-to-keep-pace-with-ai-growth-says-jitin-prasada/article71582991.ece

THE BOOKS ARE OPEN. THE QUESTIONS ARE YOURS.
Comments
Post a Comment