The answer desk
Which companies are investing in Dholera?
Published 3 August 2026. Facts verified to 27 July 2026 unless dated otherwise.
Confirmed and building as of 27 July 2026: Tata Electronics with PSMC (Rs 91,000 crore fab) and Tata Power (about 300 MW solar, operational), with ReNew, SJVN, Vena and TEQ Green developing 700 MW of solar park capacity. Announced or MoU-stage: ASML, L&T Vyoma, Adani, Embraer, Reliance, Tillman, Tsingshan, Jabil, INOX Air Products and Fujifilm. A solo Vedanta fab remains an unconfirmed claim.
The distinction that answers the question properly
Investing means different things at different tiers. The confirmed tier has government paper and physical construction. The reported tier has announcements and MoUs, which are real corporate behaviour but not yet delivered capital. Every name above is filed with its tier, date and source in the company register, and every rupee claim sits on the capital ledger.
One correction worth repeating, because search results blur it constantly: Micron is at Sanand, not Dholera, and the collapsed Vedanta-Foxconn JV should never be presented as a live Dholera project. More here.
What counts as an investment, and why the distinction matters
The word "investment" does a great deal of work in coverage of industrial corridors, and it usually conceals four very different commitments. The weakest is a memorandum of understanding, which in Indian practice is frequently a non-binding statement of intent signed at a summit, carrying no capital call, no land, and no penalty for walking away. Above that sits a land allotment, which requires a payment and a development obligation but can still be surrendered. Above that sits a binding fiscal or offtake agreement with the state, which creates enforceable milestones. At the top sits money actually spent: foundations poured, equipment ordered, payroll running. Coverage tends to treat all four as equivalent because the headline number is the same in each case, and the headline number is usually the full lifetime capital expenditure envelope, not the amount committed this year.
Dholera makes this problem acute because the announcement flow has been unusually heavy relative to the built stock. A useful discipline when reading any list of companies at Dholera is to ask three questions of each entry. Has a specific site been identified? Has an instrument been signed, and of what type? Has any third party independent of the company or the state confirmed physical work? A name that fails all three is a press release, not an investment, and it belongs in a different column of the ledger.
The confirmed tier, and the instruments behind it
Only a small number of entities can be placed in the top tier at Dholera, and each of them sits there because of a specific, dated instrument rather than an announcement.
- Tata Electronics with Powerchip Semiconductor Manufacturing Corporation of Taiwan, at roughly Rs 91,000 crore, approved by the Union Cabinet on 29 February 2024 DURABLE. The instrument that upgrades this from approval to commitment is the Fiscal Support Agreement signed on 5 March 2025, under which the Government of India committed to fifty per cent fiscal support DURABLE. A technology transfer agreement with PSMC was completed on 26 September 2024 DURABLE.
- Tata Power Solar, whose 300 MW block within the 1,000 MW Dholera solar park is recorded as commissioned DURABLE. This is the least discussed entry on the list and, in engineering terms, one of the most solid, because a commissioned generating asset is a physical, inspectable thing rather than an intention.
- INOX Air Products, which began construction in October 2025 on a Rs 500 crore electronic specialty gas hub including a 200 tonnes per day air separation unit DURABLE.
- Tata Chemicals, named as anchor investor in the government's own delivery monitoring report DURABLE.
Note what the confirmed tier has in common. Every entry is either bound by a signed fiscal instrument with the state, physically under construction, or already built. Nothing qualifies on the strength of a stated ambition.
The allotment count, and what twelve plots actually tells you
The delivery monitoring report produced by NICDC and the Department for Promotion of Industry and Internal Trade, dated 31 May 2026, records twelve plots allotted at Dholera, of which 436 acres are industrial DURABLE. It also records that industrial and commercial or residential land remained readily available for allotment as of that date DURABLE. For context, a separate official summary, the PIB release of 26 February 2026 on the Economic Survey 2025-26, shows Phase I corridor cities collectively allotted 350 industrial plots DURABLE.
Twelve is a small number against a Phase I activation area of 22.5 sq km with trunk infrastructure works recorded complete DURABLE, and against a declared region of roughly 920 sq km DURABLE. It is not, however, evidence of failure, and reading it that way misunderstands how greenfield industrial cities fill. Anchor-led corridors follow a step function rather than a curve. The anchor commits, then a long quiet period passes while the anchor builds, during which its suppliers watch but do not sign, because a supplier's capital case depends on the anchor's ramp being real rather than announced. Gas, chemicals and materials suppliers move first because their product is undifferentiated, their contracts are typically take-or-pay, and their plants are usually sited adjacent to the customer they serve. Everything else waits. The presence of INOX under construction DURABLE and Fujifilm exploring a materials base under an MoU with the Gujarat State Electronics Mission dated 30 June 2026 REPORTED is exactly the pattern you would expect at this stage: the gas cluster arrives before the component cluster.
The honest reading of twelve allotments is therefore that the ecosystem has not yet formed, and that the allotment count is the correct metric to watch, because it is a payment-backed number published by the state rather than a promotional one. Whether it inflects sharply depends almost entirely on whether the fab reaches sustained production.
The announced tier: partnerships, MoUs and what each one binds
Around the fab sits a set of relationships that are real, dated and sourced, but whose commercial content varies enormously. ASML and Tata Electronics announced a memorandum of understanding on 16 May 2026, confirmed by both the ASML press release and the Tata Electronics newsroom, naming the 28, 40, 55, 90 and 110 nanometre nodes REPORTED. Confirmation from both parties is meaningfully stronger evidence than a single-party claim, and an equipment relationship with a lithography supplier is a prerequisite for any fab, so this one carries weight even as an MoU.
Intel's MoU of 8 December 2025 is explicitly non-binding and framed as exploring manufacturing and packaging of Intel products REPORTED. Qualcomm announced a partnership on 20 February 2026 for automotive modules in India, though the scope and the Dholera link are unverified REPORTED. ROHM announced a strategic partnership on 22 December 2025 with scope unverified REPORTED. Linde is reported to be advancing plans for a Dholera plant producing oxygen, nitrogen and argon, though this rests on a single trade publication and should be held loosely REPORTED.
On the display side, the Union Cabinet on 5 May 2026 approved Crystal Matrix Limited for an integrated compound semiconductor facility at Dholera, covering 72,000 square metres per year of Mini and Micro-LED display panels plus 24,000 sets of RGB Mini and Micro-LED gallium nitride epitaxy wafers DURABLE. Cabinet approval is a stronger instrument than an MoU, because it is a government decision under the central semiconductor incentive framework rather than a statement of intent, but approval is not construction, and no construction status is publicly established DURABLE.
On the data centre side, L&T's Vyoma vehicle has an agreement with the Gujarat government for a 250 MW green AI-ready facility at Dholera valued at Rs 25,000 crore, reported by Prasar Bharati's News on AIR on 20 February 2026 REPORTED. Adani's AI-ready hub of up to 7.5 GW via AdaniConneX and the Embraer-Adani MoU for civil aircraft final assembly at Dholera airport around 2028 are both at announcement stage REPORTED TARGET. Tillman Global's roughly Rs 83,000 crore hyperscale plan is single-sourced and should be flagged hard REPORTED; the same group separately signed a Rs 15,000 crore MoU with Andhra Pradesh for 300 MW at Visakhapatnam, which is a distinct project and a common source of conflation REPORTED. Tsingshan at roughly Rs 21,000 crore for steel and EV batteries, and Jabil at roughly Rs 1,000 crore for silicon photonics, are also announcement-stage REPORTED.
Why data centre announcements cluster here, and what actually decides the outcome
The concentration of data centre announcements around Dholera is not accidental, and understanding the mechanism helps separate the ones likely to convert from the ones likely to lapse. Site selection for large compute has shifted decisively. Wood Mackenzie's assessment is that reliable, cost-competitive power has overtaken land and capital as the defining factor REPORTED, and the logic behind that is straightforward: across a twenty-year asset life, the cumulative electricity bill of a hyperscale facility is very large relative to both the land cost and, at sufficient scale, the construction cost. Land is the smallest of the three line items in most such comparisons.
Gujarat's pitch is built precisely on that shift. The state has 69 GW of installed capacity, of which 47 GW is renewable REPORTED, and the Viksit Gujarat Data Centre Policy 2026-29, announced on 9 July 2026, offers a Rs 1 per unit power tariff subsidy for twenty years, full electricity duty reimbursement for twenty years, up to 4 per cent interest subsidy, and full stamp duty and registration exemption on land, with an additional 2.5 per cent capital subsidy specific to projects inside Dholera SIR DURABLE. The qualifying threshold is a minimum 150 MW IT load DURABLE, which is a deliberate filter: it excludes colocation minnows and targets hyperscale and AI training campuses.
Two constraints sit against that. The first is water. Two rules of thumb circulate in the industry, and they are worth quoting separately rather than deriving one from the other, because they do not reconcile: one megawatt of IT load is often put at something like 25 to 26 million litres of water a year in an evaporatively cooled design, while a 100 MW facility is commonly cited at roughly two million litres a day REPORTED. Real draw depends heavily on the cooling architecture chosen, and the gap between those two figures is itself a signal that any site-level water number should be interrogated rather than repeated. Dholera's current desalination capacity is around 20 MLD DURABLE, with a tender issued on 9 July 2026 merely to appoint consultants for a 200 MLD seawater plant TARGET. Consultant appointment is the beginning of a procurement cycle, not the end of one. The engineering answer is closed-loop or air-cooled design, which trades water for electricity and capital, and any operator serious about Dholera at scale will be making that trade explicitly.
The second is competitive gravity. CBRE places Mumbai at more than half of India's operational inventory, with Mumbai, Chennai, Delhi-NCR and Bengaluru together at roughly ninety per cent of tier-I capacity REPORTED. That concentration exists because latency to users and proximity to subsea cable landings matter for inference and for anything user-facing. The counter-argument for an inland greenfield site is that AI training workloads are far more latency-tolerant than inference, which is genuinely true, and Gujarat plans two cable landing stations TARGET. Whether Dholera captures training capacity is therefore a question about power price and grid firmness, not about incentives, and the incentives are already generous.
The two errors that contaminate almost every list
Two mistakes recur so persistently in coverage of Dholera investment that any list containing either should be treated as unverified in full.
The first is Micron. Micron's Indian facility is at Sanand, not Dholera DURABLE. The two are separate locations, and the plants are not comparable in kind: Sanand is a back-end assembly, test and packaging operation, which takes finished wafers, singulates the dies, bonds and encapsulates them, and tests the packaged parts. Dholera's Tata-PSMC plant is a front-end wafer fabrication facility, where silicon wafers are patterned through hundreds of deposition, lithography, etch and implant steps DURABLE. The capital intensity, the cleanroom class, the water and gas loads, the equipment vendors and the skill profile all differ substantially. Front-end fabs are the harder, slower, more expensive proposition, so conflating the two makes India's front-end fab progress look considerably further along than the public record supports.
The second is Vedanta. The Vedanta-Foxconn joint venture collapsed in 2023 when Foxconn exited DURABLE. Claims of a solo Vedanta fab at Dholera are unconfirmed and should not be carried forward REPORTED. The persistence of this entry is instructive about how these lists propagate: one aggregator copies another, the original caveat is dropped for brevity, and a dead joint venture keeps appearing in 2026 investment tables.
Reading the fab's own signals, because everything downstream depends on them
Since almost every other name on the list is, directly or indirectly, a bet on the fab, the fab's own status is the single most informative input. Here the public record is uncomfortable and worth stating plainly. Bloomberg reported on 17 July 2026 that the plant will open mostly at 90 nanometres rather than 28, with commercial production in mid-2028, and TrendForce corroborated on 20 July 2026 REPORTED TARGET. A Tata spokesperson said the plan "has always been to start with 55nm and 90nm before introducing 28nm", and PSMC's Eric Tang noted that technology transfers are typically introduced gradually, starting with more mature nodes REPORTED. That is standard practice and technically sound: yield learning on a new site is easier at relaxed geometries, and a fab that starts mature and tightens later is following the industry's normal path. It nonetheless contradicts Tata Sons chairman N Chandrasekaran's statement in the annual report for the year ended March 2025 that "We have chosen to start our chip journey at the 28nm node" DURABLE. Both statements are on the record and they do not reconcile.
Two further disclosures matter for anyone assessing schedule risk. Economic Times reported on 28 November 2025 that Fugro, Cengrs and Geo Dynamics were engaged to redesign the main fab foundations after soil testing showed the ground unsuitable DURABLE. That detail deserves weight, because a wafer fab is among the most vibration-sensitive structures built anywhere: lithography tools require the slab to hold sub-micron stability, which generally means piling to a competent load-bearing stratum and a heavily isolated mass foundation. That is solvable engineering, but it is expensive and it consumes calendar time before any tool arrives. Separately, Malaysia-based IAQ Group was onboarded to design and install the cleanroom, reported by ETtech in June 2026 REPORTED.
What is not public is as telling as what is. There is no publicly confirmed main EPC or civil contractor for the fab, and no publicly valued equipment purchase orders DURABLE. Claims that the site was fifty per cent complete in April 2026, that foundations are complete, or that cleanroom installation is underway appear only on real-estate and search-optimised sites with no primary source behind them, and should be discarded. In a fab build, the equipment order book is among the most reliable schedule signals available, because process tools carry long lead times and vendors tend to disclose large bookings. Until valued orders appear in the public record, the mid-2028 commercial production guidance given by Ashwini Vaishnaw on 17 July 2026 TARGET is the most credible date available, and the older "first chip by December 2026" target, which traces to ministerial statements, is not supported by evidence. No first silicon had been reported as of early August 2026 DURABLE.