India's semiconductor build-out × water exposure: a watershed playbook for fabs

1. The 2026 setup, in numbers

Dholera (Tata–PSMC)

~$11 B

66.2 ha · first silicon target late 2026

Sanand (Micron ATMP)

Operational

Inaugurated Feb 28, 2026 — first ISM-cycle facility online

Dholera output target

50,000 wph

Wafers per month at full ramp (commercial node)

ISM 2.0 budget

FY26-27

Announced in the Union Budget; ecosystem-scope expansion

The headline numbers do not yet circulate alongside the corresponding water-balance numbers, but the water demand of advanced fabs is in the same engineering literature as the cleanroom and equipment specs. A 300 mm fab at a mid-node delivers wafer volumes that translate to tens of thousands of cubic metres per day of ultrapure water (UPW), produced from a much larger feed of municipal- or canal-grade intake. Recycling rates of 70–90% are achievable but require capital and operational discipline.

Place that demand inside an Indian watershed and the second-order question — "where does that water actually come from, and how does the upstream forest cover behave during a drier monsoon?" — becomes the question lenders and ESG investors increasingly want answered before they sign.

2. Why semiconductor × India is uniquely water-exposed

Water demand per unit output is large by industrial standards. A logic fab at a sub-10nm node uses thousands of litres of UPW per wafer pass. Indian fab volumes are still ramping, but the per-site demand is comparable to a small city.
The grades matter. UPW production needs source water of consistent quality. Seasonal turbidity spikes from monsoon runoff create real operational risk — pre-treatment systems can absorb a lot, but not unlimited.
India's chosen fab geography is not coastal-coincidence. Dholera SIR, Sanand GIDC, Hosur, and Mysuru all sit on industrial parks chosen for reasons that did not centre on water surplus. Gujarat in particular is a moderately water-stressed state by national rankings.
Industrial water pricing is preferential. Gujarat's semiconductor policy and similar state instruments offer subsidised industrial water tariffs. That subsidy works only as long as the basin remains hydrologically viable.
Public attention is high. Domestic media and downstream-village stakeholders are paying close attention. The first major drought-year incident with visible village impact will reshape the political envelope for the entire sector.

3. Three disclosure pressures stacking on every fab

TNFD · De-facto global

Nature-related financial disclosure

TNFD's LEAP approach explicitly requires "Locate" — geospatial identification of dependency and impact points. For a fab, that means watershed-level mapping of the intake catchment and the downstream return-water receiving body. Without that, a fab parent company's consolidated TNFD report is materially incomplete.

BRSR · India regulatory

Business Responsibility & Sustainability Reporting (Principle 6)

SEBI's BRSR Principle 6 covers environmental responsibility, including water consumption, intensity, recycling, and withdrawal source. For listed Indian holding companies (Tata Sons-linked entities, for example), fab water data flows directly into the BRSR narrative. SEBI has shown willingness to tighten BRSR thresholds further.

CDP Water Security · Investor

CDP Water Security disclosure

CDP Water Security responses now expect basin-level risk identification, intake-water source disclosure, and forward-looking water-stress scenarios. A 'A' rating requires evidence — not just intent. Lenders and equity investors increasingly reference CDP scores in financing decisions.

4. The water-balance math, plainly

Fab water balance, simplified to what an ESG team needs to defend:

Term	Order of magnitude	What it tells you
Raw intake	30,000–80,000 m³/day for a large fab	Total demand on the basin
UPW produced	10,000–30,000 m³/day	The "useful" water that touches wafers
Recycle rate	70–90% (industry target)	Determines net basin draw vs gross intake
Net consumption	≈ raw intake × (1 − recycle)	The water that does not return
Discharge water quality	Treated to permit standards	Determines downstream receiving-body load

Numbers in any individual case vary by node size, capacity utilisation, and the specifics of the water-recycling design. What stays constant is that net basin draw is a watershed-level question, and the upstream-forest portion of that watershed is what makes the basin's behaviour resilient or fragile across monsoon variability.

5. What watershed-level evidence looks like, for a fab

The same satellite stack we run for Japan answers the questions a fab needs:

Question	What we deliver
"Which basin does our intake actually draw from?"	HydroBASINS Lvl 10 catchment matched to the intake coordinates, with the upstream chain enumerated
"How does the upstream-forest water-yield contribution behave year to year?"	5-year Sentinel-2 NDVI history for the upstream forest mask + simplified water-balance estimate
"What did 2023's dry monsoon look like for our basin?"	IMD gridded rainfall + matching satellite vigour drop, by month
"How does our basin compare to alternative fab-location basins?"	Comparative water-yield score across HydroBASINS Lvl 10 polygons nationwide
"What is the downstream receiving-body load profile?"	Downstream basin chain with population-weighted village footprint
"Where is the cryptographic audit trail?"	TPM 2.0 attestation + Merkle hash + RFC 3161 timestamps on every report. Independently verifiable by lenders, CDP reviewers, and auditors.

6. A worked example: a hypothetical Gujarat fab cluster

Consider a hypothetical fab in the Dholera–Sanand corridor with a permitted raw water intake on the order of 50,000 m³/day, served by canal infrastructure that ultimately ties back to a Narmada-system distributary. The fab's sustainability team needs to answer: "What is the watershed exposure profile that backs this intake?"

With the morimieru stack:

Locate. Intake coordinates → HydroBASINS Lvl 10 catchment → trace upstream through the canal-fed sub-basins.
Evaluate forest contribution. Sentinel-2 NDVI mask of the upstream forest area + water-balance estimate using the simplified evaluation method (or its India-substituted variant). The number is a stable approximation, not the basin's true regulated flow — but it is the basin's natural-capital reservoir component.
Stress-test against history. Replay 2018, 2023, and 2025 (years with abnormal monsoon behaviour) using the same NDVI history. The pattern is the pattern; the fab can model what it would have looked like under each.
Disclose with evidence. Output: a TNFD-aligned location section, a BRSR Principle 6 quantitative table, and a CDP Water Security supporting attachment, all generated from the same dataset.
Re-run quarterly. Same pipeline, same code, same data sources — drift in upstream forest cover or rainfall regime shows up as a trend, not a surprise.

The fab does not need to share intake-volume numbers it considers commercially sensitive to use this. The watershed-side analysis is independent of the fab's internal water-balance model. The two come together only when the fab chooses to publish.

7. Stack: what stays, what swaps

Layer	Global source	India-specific source
Optical satellite	Sentinel-2 L2A (Copernicus, 10 m)	—
Climate (precip / T)	NASA POWER	IMD gridded products for finer monsoon detail
DEM	SRTM / Copernicus DEM	ISRO CartoSat where available
Watershed boundaries	HydroBASINS Lvl 10 (WWF, CC-BY)	India-WRIS finer sub-basins; CWC basin atlases
Land-use baseline	—	ISRO Bhuvan LULC + FSI ISFR
Industrial water tariff context	—	State semiconductor / industrial policy schedules
Water stress index	WRI Aqueduct global	WRI Aqueduct India + CGWB groundwater assessment
Cryptographic chain	TPM 2.0 + Merkle + RFC 3161	—

8. A 90-day disclosure-readiness pilot scope

Phase	Days	Output
1. Intake-point mapping	1–10	All intake coordinates loaded; HydroBASINS catchments identified
2. Upstream chain	11–20	Upstream basin enumeration; forest masks built
3. NDVI & climate history	21–40	2020 → 2026 NDVI per upstream basin; IMD rainfall overlay
4. Water-yield baseline	41–55	Simplified water-balance estimates per upstream forest block
5. Disclosure templates	56–75	TNFD Locate / Evaluate sections; BRSR Principle 6 tables; CDP supporting attachments
6. Cryptographic chain & handoff	76–90	Signed report bundle, exportable for parent-company consolidation

We have not yet run this in India for a semiconductor client. We have run every component step for our Japan forest pilots and have documented the India-layer substitutions in our India pilot article. The plan above is what we would propose to a fab, a state IDC, or a sustainability-led sovereign / lender stakeholder.

9. Honest constraints

This is the upstream-natural-capital side of the picture. It does not replace the fab's internal water-balance model. The two together produce a defensible disclosure; either alone is partial.
Canal-fed intakes blur the watershed. When water arrives through multi-basin canal infrastructure, the "natural" upstream basin is one input among several. We model the dominant contributor; the rest needs hydrologic engineering.
Industrial discharge analysis is out of scope. NDVI says nothing about discharge water chemistry. Receiving-body load needs in-situ monitoring.
Forecasts are not predictions. Historical replays show what the basin has done. Forward-looking stress tests need climate-model inputs (CMIP6 etc.), which add their own uncertainty.

10. Who this is for

Fab sustainability teams needing TNFD / BRSR / CDP evidence that is defensible against external audit.
Parent-company ESG and IR functions consolidating fab data into group-level disclosure.
Lenders and equity investors wanting their own independent water- exposure view before committing capital.
State IDCs and ISM-affiliated bodies designing the next round of fab-site selection and water-allocation policies.
Civil society watchdogs and downstream-village stakeholders who want transparent monitoring of fab water impact on shared basins.

11. Frequently asked questions

Q1. Aren't fabs already disclosing their water numbers?: Fab parent companies do disclose aggregate water KPIs in sustainability reports. What is typically missing is basin-level provenance and behaviour — i.e., where the water came from and how the upstream basin has behaved over recent years. TNFD's Locate step explicitly asks for that, and most disclosures stop short of it.
Q2. Does morimieru need access to confidential fab water data?: No. The watershed-side analysis runs entirely on public data (satellite + climate + watershed boundaries). The fab can keep its internal water-balance numbers confidential and still publish the watershed context alongside.
Q3. Doesn't every fab already have a hydrology consultant?: Yes, typically for permitting. That work is one-shot and proprietary. morimieru adds a public, repeatable, time-series layer that runs continuously and is auditable by anyone — which is what TNFD and CDP increasingly want.
Q4. How does this connect to CAMPA?: Where a fab site involved forest-land diversion under the FCA, the compensatory afforestation plot is a CAMPA case. Our CAMPA article covers that monitoring layer; the same pipeline ties to fab disclosure for the natural-capital chain.
Q5. Does this only help large fabs?: No. The pipeline cost scales with the number of intake basins, not the size of the facility. An ATMP / OSAT facility with a single intake gets the same evidence layer as a 300 mm logic fab.

12. References & sources

India Semiconductor Mission (ISM) — programme documents, Union Budget FY2026-27 announcements
Tata Electronics + PSMC + ASML — public statements on Dholera fab targets and timeline
Micron Technology — public statements on Sanand ATMP inauguration (Feb 2026)
Government of Gujarat — Semiconductor Policy and industrial water tariff schedules
TNFD — Recommendations of the Taskforce on Nature-related Financial Disclosures (Sept 2023)
SEBI — Business Responsibility & Sustainability Reporting (BRSR) framework, Principle 6
CDP — Water Security questionnaire (current version)
WRI Aqueduct — global and India water-stress indicators
India-WRIS / CWC — basin and groundwater datasets
Copernicus Sentinel-2 — open satellite data via AWS Open Data
WWF HydroSHEDS / HydroBASINS — global watershed boundaries (CC-BY)

Last updated 2026-05-26. Site-specific water demand numbers vary by node size, capacity utilisation, and recycle architecture; the ranges in this article are sector-typical and should not be read as company-specific.