Record open Company files. Capital. Supply chain. Verified 3 Aug 2026
DholeraDigital
India's semiconductor build-out, tracked from the ground
ConfirmedRs 91,000 cr
Next windowQ4 2026
Pages on record75

Front Page › Briefings › The supply chain quietly forming around the fab

Briefing

The supply chain quietly forming around the fab

Announcements can be staged. Air separation units cannot. Follow the suppliers and the fab's reality argument makes itself.

4 layers

Four supplier layers are now visible around the Dholera fab: equipment (ASML MoU, announced 16 May 2026), gases (INOX's Rs 500 crore hub, cold box reported erected), materials (Fujifilm exploratory MoU, 30 June 2026) and power (300 MW operational solar, 700 MW in build).

Source: Dholera Digital company register. As of 27 July 2026.

The logic of supplier evidence

A fab announcement costs a press release. A supplier ecosystem costs balance sheets: gas companies erect capital equipment, materials makers run siting studies, toolmakers allocate scarce engagement time. Suppliers commit these only against anchors they have privately diligenced. That is why this record weighs supplier behaviour above promotional volume: it is the market grading the fab's credibility with its own money.

The four layers on record

  • Equipment: ASML's lithography MoU REPORTED, the layer a fab can least improvise.
  • Gases: INOX Air Products' reported Rs 500 crore hub with a 200 TPD ASU, cold box reported standing REPORTED, the most physical of the supplier signals.
  • Materials: Fujifilm's exploratory MoU with the Gujarat State Electronics Mission REPORTED, diligence in progress, honestly labelled.
  • Power: ~300 MW operational with 700 MW building toward the 1,000 MW Phase I target DURABLE.

What would complete the picture

Chemicals and CMP consumables, test and packaging capacity, and precision logistics providers. When names in those categories appear with Dholera commitments, the register will file them the same way: tier first, hype never. The May 2026 Cabinet clearance of two further semiconductor units for Gujarat suggests the cluster logic is already compounding DURABLE on the clearance itself.

The four layers, and why each behaves like a different business

It helps to stop thinking of a fab's supply chain as one chain and start thinking of it as four layers with almost nothing in common. The first is process gases and bulk chemicals, which are governed by physics: nitrogen is cheap to make and expensive to move, so it gets produced next to the consumer or piped a short distance, and the supplier's plant is effectively part of the fab. The second is consumable materials, meaning silicon wafers, photoresists, developers, chemical mechanical planarisation slurries and pads, sputtering targets and photomasks. These are dense, high value and shippable by air, so they localise last and localise reluctantly. The third is capital equipment, spare parts and field service, which is a global oligopoly of a handful of firms and does not localise at all in any meaningful sense, though its service tail can. The fourth is utilities and the facility itself: power, ultrapure water, wastewater treatment, cleanroom envelope and vibration control. That layer is irreducibly local because it cannot be imported.

The consequence is that a semiconductor cluster forms in a predictable order. Gases and utilities arrive first because they must physically exist before a tool can be switched on. Equipment arrives as capital, not as industry. Materials follow years later, if at all, and only when volume justifies a dedicated plant. Anyone assessing whether a cluster is real rather than announced should therefore weight the layers differently rather than counting memoranda.

Why supplier capital expenditure is the cleanest diligence signal available

A memorandum of understanding costs a company a press release and a photograph. A gas plant costs it a balance sheet entry that cannot be reversed. That asymmetry is the whole basis of reading a project through its suppliers rather than through its own announcements, and it rests on a well understood idea in industrial economics: asset specificity. An air separation unit built at the boundary of a fab is worth a great deal to that fab and comparatively little to anyone else, because the value sits in the pipe rack running over the fence, not in the machinery. A supplier will only sink that money after it has seen the customer's construction programme, satisfied itself on the ramp schedule, and signed a long term supply agreement, typically structured on a take or pay basis so that the volume risk sits with the buyer. In other words, the supplier has run the diligence that outsiders cannot run, and its concrete is the visible residue of that private judgement.

On this reading, INOX Air Products beginning construction in October 2025 on a Rs 500 crore electronic specialty gas hub at Dholera DURABLE is a more informative data point than most of the partnership announcements attached to the fab itself. It is money in the ground from a party with no incentive to be early. Linde is separately reported to be advancing plans for a Dholera production plant REPORTED, though that report is single sourced and should be held loosely until a construction start or contract award is confirmed.

The signal has limits and they should be stated. Industrial gas suppliers also serve merchant markets, and Dholera's wider industrial base could absorb part of an air separation unit's output, so a gas plant is necessary evidence of a fab progressing but not sufficient proof of its schedule. It tells you the site will have gases. It does not tell you when wafers start.

Layer one: what a 200 tonne per day air separation unit actually implies

An air separation unit does something conceptually simple and industrially demanding: it compresses air, removes water and carbon dioxide, cools it until it liquefies and distils it into nitrogen, oxygen and argon. The air separation unit within the INOX facility at Dholera is specified at 200 tonnes per day DURABLE, and the hub is described as covering nitrogen, oxygen and argon alongside electronic specialty gases DURABLE.

Nitrogen is the volume story in any fab. It is used to purge process chambers, to inert load locks and transfer chambers, to blanket chemical delivery lines, to dry wafers, and increasingly to fill the internal atmosphere of wafer carriers so that a lot sitting between steps does not grow native oxide. Consumption is continuous and enormous, which is why a large fab takes nitrogen by pipeline from a neighbouring plant rather than by liquid tanker: above a certain steady demand the delivered cost of trailered liquid becomes indefensible. Purity requirements are severe, commonly six nines or better with moisture and oxygen specified in parts per billion, because the gas touches the wafer surface at every step. Oxygen feeds thermal oxidation and plasma ashing. Argon is the workhorse of physical vapour deposition and plasma processes.

The specialty side is a different discipline again. Silane, ammonia, tungsten hexafluoride, nitrogen trifluoride for chamber cleaning, hydrogen chloride and a range of dopant gases arrive in cylinders and ISO containers, and the local requirement is less about manufacture than about packaging, blending, analysis, cylinder management, recertification and safe handling of pyrophoric and toxic species with the abatement systems that go with them. That is what the word hub is doing in the INOX description, and it is the part with the longest tail of local employment.

Layer two: materials, where the gap is widest

Prime polished 300mm silicon wafers are made by a very small group of global producers, and none of them currently manufactures them in India. At the stated ceiling of 50,000 wafer starts per month TARGET the fab would consume in the order of 600,000 substrates a year at full utilisation TARGET, which is a serious volume by any standard and yet still small relative to the minimum efficient scale of a crystal pulling and polishing plant. Photoresists and their ancillary developers, edge bead removers and anti-reflective coatings are similarly concentrated, historically among Japanese chemical firms, and are qualified process by process rather than bought off a shelf, which makes substitution slow even when a second source exists.

Bulk wet chemicals are the most misunderstood item. India has a very large commodity chemicals industry, and it is tempting to assume that sulphuric acid, hydrogen peroxide, ammonium hydroxide and hydrofluoric acid are therefore solved. They are not. Semiconductor grade specifications for these chemicals are separated from industrial grade by several orders of magnitude in metallic and particulate contamination, and meeting them requires dedicated purification trains, dedicated packaging and dedicated logistics that never touch the industrial stream. Building that is a capital project in its own right.

Policy has now noticed. The second phase of the India Semiconductor Mission announced in the Union Budget for 2026-27 is directed specifically at equipment and materials manufacturing along with domestic semiconductor intellectual property REPORTED, and Fujifilm signed a memorandum with the Gujarat State Electronics Mission on 30 June 2026 to explore a semiconductor materials base REPORTED. Exploration is the correct word to use about it for now.

Layer three: equipment, and why the node dispute changes the tool list

The majority of a fab's capital cost is equipment, not building, and that single fact makes the unresolved node question consequential rather than cosmetic. A process centred on 90nm and 110nm is served by krypton fluoride and dry argon fluoride lithography with i-line steppers on non-critical layers, comparatively relaxed overlay budgets, fewer mask layers and a metrology load to match. A genuine 28nm capability requires immersion argon fluoride scanners, materially more masks per wafer and a step change in overlay and defect metrology. These are not the same factory with a different label on the door.

That is the context for the reporting by Bloomberg on 17 July 2026 that the fab will open mostly at 90nm REPORTED, corroborated by TrendForce on 20 July 2026 REPORTED, against the Tata Sons chairman's statement in the annual report for the year ended March 2025 that the company had chosen to start its chip journey at 28nm REPORTED. A Tata spokesperson has framed the plan as always having been to begin at 55nm and 90nm before introducing 28nm REPORTED, and PSMC's Eric Tang has described technology transfers as typically introduced gradually from more mature nodes REPORTED. The ASML memorandum announced on 16 May 2026 by both parties names the full range of 28, 40, 55, 90 and 110nm DURABLE, which is consistent with a phased tool installation rather than a single monolithic order.

Mature node equipment also has a deep secondary market. Refurbished and relocated tools are normal at 90nm and are bought through channels that do not generate headline purchase order announcements, which is one plausible and undramatic explanation for the absence of publicly valued equipment orders. The unglamorous corollary is the service layer: field engineers, spares depots, chamber and quartz part refurbishment, orbital welding of electropolished tubing, calibration laboratories. That layer is where local supply chain value genuinely accrues, and nothing about it at Dholera is publicly documented.

Layer four: power, water and the facility as a process tool

A fab is a continuous process. A momentary voltage dip lasting a fraction of a second can abort tools across the line and scrap the wafers in them, which is why the industry maintains a specific standard for equipment immunity to voltage sags and why a fab cares more about grid quality and dedicated substation design than about tariff. The 300 MW of the 1,000 MW Dholera solar park awarded to Tata Power Solar is commissioned DURABLE, with the remaining 700 MW under development against a March 2027 target TARGET. Solar is a cost and emissions instrument rather than a reliability instrument, because a fab needs firm power at three in the morning, so the meaningful questions concern firm capacity, redundant feeders and on site uninterruptible supply rather than installed megawatts.

Water is the other hard constraint. Fabs do not consume municipal water, they consume ultrapure water made from it, and the treatment train loses a substantial fraction of the feed, so the raw intake always exceeds the process demand even before reclaim systems recover part of the return. Dholera currently has a desalination plant of roughly 20 MLD DURABLE, with a tender issued on 9 July 2026 to appoint consultants for a 200 MLD seawater desalination plant REPORTED and bulk pipeline proposals including 100 MLD from Bhaskarpura lake in Surendranagar in the Gujarat Budget for 2026-27 REPORTED. Consultant appointment is the first step of a long procurement, not the plant.

The building itself is a process tool. Lithography and metrology impose vibration criteria that dictate slab mass, damping and foundation design, and the report of 28 November 2025 that Fugro, Cengrs and Geo Dynamics were engaged to redesign the main fab foundations after soil testing found the ground unsuitable REPORTED should be read in that light rather than as a mishap. The published record does not describe the ground conditions or the redesigned solution in engineering detail. Separately, the Malaysia based IAQ Group has been onboarded to design and install the cleanroom REPORTED.

What a complete cluster would additionally contain

Even with all four layers present, a fab alone is not a cluster. A complete ecosystem would add back end assembly, test and packaging, which in Gujarat exists at Sanand through Micron rather than at Dholera DURABLE, so front end wafers would travel for packaging. It would add chemical distribution and blending yards, cylinder logistics and recertification, industrial wastewater treatment capable of fluoride precipitation, ammonia stripping and solvent handling, precision machining and parts cleaning for chambers and quartzware, accredited calibration and analytical laboratories, and bonded warehousing with customs handling fast enough that a stranded tool part does not idle a line. The special economic zone notified for Tata Semiconductor Manufacturing at Dholera in April 2026 DURABLE is relevant here principally because zone status governs duty treatment on imported capital goods and consumables, which is a logistics instrument as much as a fiscal one.

The scarcest input is people. Tata has sent a couple of hundred staff to PSMC in Taiwan in reported batches of around 75 across equipment, yield, process technology and quality engineering REPORTED, and PSMC expects more than NT$20bn from consulting, technology transfer and training across the lifecycle REPORTED. Those numbers describe a knowledge transfer of real substance, and also the scale of what is being imported: process discipline, not just machines.

The honest negative findings

Several things that a reader would expect to find are not in the public record, and the absence should be stated plainly rather than papered over. No main engineering, procurement and construction contractor or principal civil contractor for the fab has been publicly confirmed, which is unusual for a project of this size and this stage. No valued equipment purchase orders have been made public. Widely circulated claims that the project was 50 per cent complete in April 2026, that foundations are complete, or that cleanroom installation is underway appear only on real estate and search optimised websites with no primary source behind them, and should be treated as unsourced until a filing, a contractor announcement or a government report carries them.

On schedule, commercial production guidance stands at mid-2028 TARGET. The older claim of first chip by December 2026 traces to ministerial statements and is not supported by the current record, and no first silicon has been reported. Among the partnerships, the Intel memorandum of December 2025 is explicitly exploratory and non-binding REPORTED, and the scope and Dholera linkage of the Qualcomm and ROHM announcements remain unverified REPORTED. Taken together, the supply chain evidence is thinner than the headline capital figure implies, with one clear exception: in the gas layer, someone has actually poured concrete.

Cite this: "Gases, materials, lithography, power: the supplier ecosystem assembling around the Rs 91,000 crore Dholera fab, and why supplier behaviour is the strongest evidence the anchor is real." Dholera Digital, 2026-08-03. https://dholera.digital