Company file

Tata Power Solar at Dholera

Published 3 August 2026. Facts verified to 27 July 2026 unless dated otherwise.

Tata Power runs about 300 MW of operational solar at Dholera on what it calls India's largest single-axis tracker.

Status tier: DURABLE

The quiet anchor

Semiconductor fabs are power-hungry and intolerant of outages. A city pitching itself to chipmakers and data centre operators needs generation it can point at, not promise. The operational 300 MW is exactly that, and it is why the solar story belongs in a business record and not just an energy one.

The verified record

• Operational: ~300 MW live at Dholera, Tata Power Solar. DURABLE
• Phase I target: 1,000 MW solar park. TARGET
• In development: 700 MW awarded to ReNew, SJVN, Vena and TEQ Green, with Gujarat Power Corporation Ltd as agency, target March 2027 (CEA quarterly report, September 2025). DURABLE on the award, TARGET on the date.

The renewables build-out also feeds the industrial pitch: green power is part of how Dholera courts AI data centre capital, from the L&T Vyoma MoU to the larger reported hyperscale plans on the ledger.

Why a fab buys firmness, not sunshine

Industrial buyers of electricity fall into two camps. One wants the lowest average tariff and can tolerate interruption. The other wants continuity above all else, because an interruption does not pause production, it destroys work in progress. A fab sits squarely in the second camp, and so, for different reasons, does a data centre.

The mechanism explains most of the siting decisions that follow. A wafer spends weeks moving through several hundred sequential process steps, many of them irreversible. A lithography scanner mid-exposure, an etch chamber mid-cycle or a furnace mid-ramp does not tolerate an unplanned stop, and the industry does not measure the threat in minutes. The SEMI F47 standard, which defines how process equipment should ride through a voltage sag, is written in cycles and fractions of a second, because a dip of a few hundred milliseconds propagating through a fab substation can abort tools across multiple bays at once. Recovery is measured in shifts: chambers requalified, lots scrapped, yield data quarantined. A single excursion of that kind can cost more than a year of tariff savings, which is why fabs tend to pay for reliability rather than shop for cheapness.

Data centres reach a similar conclusion by another route. Their loads are genuinely continuous, their contracts are written around uptime, and their carbon commitments are drifting from annual accounting, where a year of certificates offsets a year of consumption, towards hourly matching, where the operator wants clean generation on the grid in the same hour the servers draw it. That shift turns nearby renewable capacity from a marketing line into a procurement requirement. Wood Mackenzie's reading of the Indian market, that reliable and cost-competitive power has overtaken land and capital as the defining site-selection factor, is the same observation from the investor's side.

Gujarat's incentives follow the same logic. The Gujarat Semiconductor Policy 2022-27 carries a power subsidy of Rs 2 per unit for ten years alongside an electricity duty exemption DURABLE, and the Viksit Gujarat Data Centre Policy 2026-29, announced on 9 July 2026, offers Rs 1 per unit for twenty years plus full reimbursement of electricity duty DURABLE. Both are written for buyers whose meter never stops turning.

Tracking, and what it actually buys

No mounting or tracking configuration has been published for the 300 MW commissioned at Dholera DURABLE, so any description of that plant as fixed-tilt, tracked, or record-setting in either direction should be treated as unsourced. The choice is worth understanding regardless, because tracking is one of the few decisions in a solar plant that changes the shape of the output rather than only its size. A fixed-tilt array points its modules at a compromise angle and accepts that the sun spends most of the day somewhere else. A horizontal single-axis tracker mounts rows on a north-south torque tube and rotates them east to west, keeping the array closer to normal incidence from mid-morning to mid-afternoon. Dual-axis systems add a second degree of freedom to chase seasonal elevation, but add mechanical complexity for a comparatively modest gain, which is why most utility-scale plants stop at one axis.

What tracking buys is differently shaped energy, not simply more of it. Fixed arrays produce a narrow peak around solar noon. Trackers flatten and widen that curve, lifting the morning and afternoon shoulders, which matters commercially because those shoulders sit closer to the hours when demand is climbing. The trade is a set of new failure modes: slew drives, bearings and controllers a fixed array does not have, row-to-row shading managed by backtracking at low sun angles, a looser ground coverage ratio meaning more land per megawatt, and a wind stow strategy whose failure is a recognised cause of large-scale module damage in storms.

Two site conditions generally matter more on a flat coastal site than inland, though neither has been documented for this park specifically. The first is the ground. Coastal terrain of this type can bring a high water table and saline soils, which complicate pile-driven foundations and are hostile to galvanised steel over a twenty-five year design life. Something in that family surfaced on the fab side, where Fugro, Cengrs and Geo Dynamics were engaged to redesign the main fab foundations after soil testing found the ground unsuitable DURABLE. The published account does not set out what the testing found, nothing comparable has been published about the solar park's foundations, and the two problems should not be assumed identical. The second condition is soiling. Coastal dust and salt deposition reduce output between cleans, and cleaning consumes water in a district where water is the binding constraint, served today by a desalination plant of roughly 20 MLD with a tender issued on 9 July 2026 to appoint consultants for a 200 MLD plant REPORTED.

A solar park is a procurement structure, not a power station

The phrase 1,000 MW Dholera Solar Park suggests a single asset with a single owner. Indian solar parks are almost never that. The model separates two businesses: the park agency assembles land, builds internal roads and, above all, the pooling substation and transmission connection that lets generation reach the grid, then allots plots inside that envelope to competitively selected generators who build, own and contract their arrays separately.

At Dholera the agency is Gujarat Power Corporation Ltd DURABLE. That structure is why the 700 MW still under development carries four names rather than one: ReNew, SJVN, Vena and TEQ Green DURABLE, each holding a slice of the same envelope against a March 2027 date recorded by the Central Electricity Authority in September 2025 TARGET. Tata Power Solar's 300 MW is the first slice across the line, recorded as commissioned in the NICDC and DPIIT delivery monitoring report of 31 May 2026 DURABLE.

Three consequences are easy to miss. The headline 1,000 MW is a capacity envelope defined by land and evacuation capability, not an order that has been placed. The agency's economics and the generators' are independent, so shared infrastructure and individual plots advance at different speeds. And no developer is obliged to sell to a neighbour merely because the neighbour is close. We have found no published power purchase agreement linking any Dholera generator to Tata Electronics or to any other in-region industrial load, so the default assumption is grid sale on ordinary terms.

Commissioned, awarded, under development: three different facts

These are three unrelated states of the world, and renewable-energy reporting collapses them constantly. An award means a competitive process has concluded and a letter of award has been issued to a named developer. It establishes intent and a counterparty, not financing: financial close typically follows months later once a power purchase agreement is signed and lenders are satisfied, and in a market where module prices move quickly, an aggressive tariff can become uneconomic before a pile is driven. Under development is the widest of the three terms and the least informative, covering both a project just awarded and one energising its last string. Without a percentage-complete figure or an evacuation readiness date attached, it says only that the project has not formally stopped.

Commissioned is the term with real procedural content. In Indian practice a plant is synchronised to the grid, run through trial operation, inspected and then certified, sometimes in tranches, so a 300 MW plant may reach full commissioning through several part-commissioning certificates over months. That certificate makes the status auditable: it carries a date, an issuing authority and a quantum in megawatts. What it does not promise is performance. Commissioned capacity is nameplate capacity at the point of connection, and says nothing about yield, availability, or whether the network beyond the pooling substation can absorb full output at noon in March.

Reading the CEA as an evidence source

Most Dholera energy claims in circulation trace back either to a press release or to a website that copied one. The Central Electricity Authority sits in a different position in the information chain. It is the statutory technical body of the power sector, it collects returns from state agencies, transmission licensees and developers as routine, and it publishes on a fixed cadence: capacity and generation monthly, projects under construction and development quarterly. Nobody at the CEA is selling a project.

That makes a CEA line good evidence for the existence and identity of an award, which is why this page treats the 700 MW allocation as durable and March 2027 as a target TARGET. It matters equally what the source cannot do. The CEA logs what it is told, so a developer's optimistic commissioning date enters the table as a date rather than as a forecast, and slippage usually appears as a quietly revised entry in a later period rather than as a correction. Entries lag reality in both directions: a plant can be commissioned weeks before it appears, and a stalled project can sit in the tables for a year looking alive. The reports also cover none of the commercial layer, with no tariffs, no power purchase agreements and no financial close.

The method, then, is to read successive editions rather than a single one. A project whose target date moves twice while its capacity stays fixed has an execution problem, and one that reappears with a new capacity figure has usually been resized. Both patterns are visible from outside at no cost, and either beats an announcement as a basis for judgement. Our running version of that comparison sits on the promise versus delivery log.

Variable generation against a load that never blinks

Here is the arithmetic any serious discussion of green power for Dholera has to confront. Solar capacity is quoted at nameplate, meaning output under standard test conditions, while industrial load is quoted as a continuous draw. The two are not comparable. A well-built utility-scale plant in Gujarat converts nameplate into annual energy at a capacity factor that, in general Indian experience, sits in the high teens to mid twenties per cent, with tracking towards the upper end. On that basis, 300 MW of nameplate averages somewhere between a fifth and a quarter of its rating over a year, nights included. That is a general engineering approximation. No performance figure for this plant has been published.

The consequence is that a fab drawing a flat load cannot be run off a solar park of comparable nameplate. It can be supplied by the grid and matched against that park in accounting terms, which is what Indian corporate renewable arrangements do through open access, banking and captive structures whose rules the state regulator sets and revises. Physically, the electrons consumed at three in the morning come from whatever is running on the Gujarat grid at three in the morning.

Gujarat is unusually well placed for this and unusually exposed to it. The state has roughly 69 GW of installed capacity of which about 47 GW is renewable DURABLE, a remarkable ratio and also a balancing problem, since a grid carrying that much variable generation faces a steep evening ramp as solar falls away and demand rises. Serving large, inflexible industrial load on such a system requires firm capacity in reserve, storage, or contracts that let the load flex, and a fab cannot flex. No co-located battery storage has been announced for this park.

Everything above concerns kilowatt-hours, but what protects a fab from the excursions described earlier is local engineering: a dedicated high-voltage connection, redundant transformers, ride-through and uninterruptible supply inside the fence. No detail of the fab's connection design, contracted demand or backup arrangement has been published.

What 300 MW does and does not cover

The honest answer to what the operational capacity covers starts with an admission: Tata Electronics has not published a contracted demand figure for the Dholera fab, so any specific megawatt number attached to it in circulation is unsourced. General industry experience says a 300 mm fab running up to 50,000 wafer starts per month DURABLE is a continuous load in the tens to low hundreds of megawatts once fully ramped, much of it drawn not by the process tools but by the plant around them: cleanroom air handling, ultrapure water, process cooling, exhaust abatement and compressors that never switch off.

The fab is not the only claimant. INOX Air Products began construction in October 2025 on a Rs 500 crore electronic specialty gas hub including a 200 tonne-per-day air separation unit DURABLE, and cryogenic air separation is among the most electricity-intensive continuous industrial processes, run flat out because turning it down is expensive. Any data centre qualifying under the state policy needs a minimum IT load of 150 MW DURABLE, and the L&T Vyoma agreement for a 250 MW facility REPORTED is quoted as a single headline figure, with no published breakdown of whether it refers to IT load or to the larger draw the grid would see once cooling and losses are added. Stacked together, the 1,000 MW envelope DURABLE looks less like generosity than the first tranche of an energy build-out that must continue as long as the industrial one does.

Because so much of this rests on the gap between an award and an asset, it is worth naming what would justify a change of tier. Part-commissioning entries for any of the four 700 MW developers in a later CEA report would move part of that block from target to durable. A published power purchase agreement, or a regulator's open access approval, linking Dholera generation to a named in-region buyer would close the largest documentary gap on this file. A storage tender attached to the park would signal that firmness rather than volume has become the design objective.