The answer desk

When will the Dholera fab open?

Published 25 August 2026. Facts verified to 25 August 2026 unless dated otherwise.

There is no supported first silicon date for the Dholera fab, and this record does not print one. What exists is guidance: Union minister Ashwini Vaishnaw said on 17 July 2026 that commercial production is expected by mid-2028, and Bloomberg reported the same date the same day, corroborated by TrendForce on 20 July 2026. Civil work was about 50 percent complete in mid-2026 with cleanroom fit-out reported underway. The December 2026 first chip talk traces to ministerial statements rather than any company milestone, and it is not a fact.

Open means three different things, and the confusion is the whole problem

Almost every disagreement about the Dholera fab's date is really a disagreement about the word open. A wafer fabrication plant does not open the way a shop opens. It passes through at least three distinct events, separated by many months, and coverage routinely reports one of them using the language of another.

• Mechanical and building completion. The structure is finished, the cleanroom envelope is sealed and certified to its particle class, utilities are live, and the plant can accept process equipment. A ribbon can be cut here, and often is.
• First silicon. The first wafers run through the installed tools, usually as test and characterisation lots rather than as product. First silicon proves the line functions. It proves nothing about yield, and the wafers are generally not sellable.
• Commercial production. The process is qualified, yields are commercially viable, customers have qualified their own devices on the line, and the fab ships parts against orders. This is the only one of the three that has revenue attached, and it is the meaning that matters to suppliers, customers and the public purse.

The gap between the second and the third event is where most fab schedules are actually won or lost, and it is measured in quarters rather than weeks. Any answer that gives a single date without saying which of the three it refers to should be discarded on that basis alone.

The only guidance that is actually sourced

On 17 July 2026 the Union Minister for Electronics and Information Technology, Ashwini Vaishnaw, stated that the plant is expected to begin commercial production by mid-2028. TARGET Bloomberg reported the same date on the same day alongside its finding that the fab will open mostly at 90nm rather than the 28nm presented in 2024, and TrendForce carried the report independently on 20 July 2026. REPORTED

That is the whole of the supported forward guidance. It is a target, it comes from the sponsoring minister rather than from a company filing, and this desk files it as a target and will score it on delivery rather than treat it as an outcome. What it is not is a first silicon date. No source publishes one, and the distinction was worth a correction on this site: an earlier version of these pages carried mid-2028 as a first silicon date, which conflated the third event above with the second. Mid-2028 is the commercial production guidance. Every other reading of it is our error or someone else's.

The December 2026 claim, and why it does not appear here as fact

A first chip by the end of 2026 has circulated widely for two years and still appears in property marketing, aggregator posts and a good deal of search optimised writing. Its provenance matters. It traces to ministerial and executive statements of intent made around and after the Union Cabinet approval of 29 February 2024, not to a company milestone announcement, a filing, a supplier confirmation or any dated engineering event. No source has published that a wafer has been fabricated at Dholera. As of 25 August 2026, none has been reported by any party. REPORTED

Two facts sit against the claim rather than merely failing to support it. The first is arithmetic on the construction record below: a plant with roughly half its civil work done in mid-2026, with cleanroom fit-out underway rather than complete, has not yet reached the point where process tools are installed and qualified, and tools cannot run wafers before they are installed. The second is the sponsoring minister's own guidance of 17 July 2026, which points at mid-2028 for commercial production. A December 2026 first silicon would not be impossible in principle for a differently sequenced project, but it is not what this project's own record describes. This desk treats the claim as unsupported, which is a stronger statement than unproven: the evidence points the other way.

The construction record, dated

The defensible answer to when the fab opens is built out of what has actually happened, each item dated, with a tier tag on the entries a skeptic would press on.

• 29 February 2024. Union Cabinet approval of the Tata Electronics and Powerchip fab under the India Semiconductor Mission, at a stated Rs 91,000 crore for up to 50,000 wafers a month on 300mm wafers across 28 to 110nm.
• March 2024. Ground breaking at the Dholera site.
• 26 September 2024. Technology transfer agreement with Powerchip completed.
• 5 March 2025. Fiscal Support Agreement signed, with the central government committing to half of the eligible project cost; the associated release cites Rs 91,526 crore. DURABLE
• 28 November 2025. The Economic Times reports Fugro, Cengrs and Geo Dynamics engaged to redesign the main fab foundations after soil testing found the ground unsuitable. REPORTED
• Around April 2026. The Tata Semiconductor special economic zone at Dholera, covering 66.16 hectares, is notified, with press reporting citing about 21,000 jobs. DURABLE on the notification, and the job figure remains an administrative projection rather than a headcount.
• 16 May 2026. ASML lithography memorandum announced by both newsrooms, naming 28, 40, 55, 90 and 110nm.
• 31 May 2026. The NICDC and DPIIT Delivery Monitoring Unit report records twelve plots allotted at Dholera, 436 acres of industrial land, Tata Chemicals as anchor investor, and trunk infrastructure works complete in the 22.5 sq km activation area. DURABLE The full line by line reading is on our progress report page.
• Mid-2026. Civil work reported at about 50 percent, with cleanroom fit-out underway. REPORTED The plant is not structurally complete, and claims that it is do not survive sourcing.
• June 2026. IAQ Group reported onboarded to design and install the cleanroom.
• 17 July 2026. Commercial production guided to mid-2028; the fab reported to open mostly at 90nm. TARGET

Read as a sequence rather than as a list, that record describes a project in the envelope and fit-out phase of a fab build, roughly two years after ground breaking, with equipment installation still ahead of it. The dated build ledger is maintained separately on our construction timeline.

The dependency chain that decides the date

Fab schedules are not estimates of effort. They are chains of dependencies in which no link can start early, which is why apparently unglamorous construction facts carry so much schedule weight.

Foundations and structure must be complete before the cleanroom envelope can be built, because the envelope hangs off the structure and inherits its vibration behaviour. The envelope must be sealed and certified before tools move in, because a tool installed in a dirty room is a tool that will be recleaned. Bulk gases, specialty gases, ultrapure water, chemical distribution, exhaust and abatement must be live before a tool can be hooked up, because a tool without utilities is a crate. Tools are installed in sets, then accepted against contractual specifications for overlay, focus and throughput, then qualified for the fab's own process layers, which means running wafers and measuring them. Only then does yield learning start, and only after yield learning does customer qualification begin, which is itself a multi month exercise for automotive and industrial parts.

Each of those steps has a duration set by physics and procedure rather than by ambition, and each is gated by the one before. That structure is why fabs slip in the way they do, which our page on schedule risk covers as an engineering subject rather than as an accusation. It is also why a foundation redesign reported in late 2025 is genuinely schedule relevant while a memorandum signed in mid-2026 mostly is not.

Is mid-2028 plausible against the remaining work?

A target can be tested against the work still in front of it without predicting anything. From mid-2026, with civil work reported at about half and cleanroom fit-out reported underway, the remaining sequence to commercial production contains a known set of stages: completing the structure and the envelope, certifying the cleanroom, commissioning bulk and specialty gases, ultrapure water, chemical distribution and abatement, installing and accepting tools in sets, qualifying the process layers, running yield learning, and then qualifying customer devices. Whether that remainder fits into the two years the guidance implies is exactly what the target is asserting, and nothing in the sequence makes it impossible, because the stages overlap rather than run end to end: utilities commissioning proceeds while fit-out finishes in other bays, and tools are installed in one module while another is still being sealed. What this desk will not do is convert that observation into a prediction, because the durations that would be needed to do so are not published for this project.

What makes the target harder here than the sequence suggests is not construction. It is that this is a first fab for the operating organisation. Process engineers become useful after they have seen a full production cycle, and equipment engineers after they have seen the failure modes their tools actually exhibit rather than the ones the manuals describe. Powerchip's involvement is designed to compress exactly that gap, through the technology transfer agreement completed on 26 September 2024. DURABLE Engineer secondments to Taiwan in batches have been reported alongside it, and reporting is all they are. Whether the compression is enough is the genuine open question on this schedule, and it will not be visible until tools are running wafers.

Two further items sit outside the company's control. The first is customer qualification, which is a matter of the customer's calendar rather than the fab's: automotive and industrial buyers run their own reliability programmes on their own timelines, and a fab can be technically ready months before a customer says yes. The second is the equipment delivery queue, which is set by the toolmakers' order books and is invisible from outside until orders are announced. Neither is a reason to doubt the target. Both are reasons to score it rather than to assume it.

How this desk will score the date

A target is only useful if someone writes it down and checks it later, which is the practice this record exists to keep. The scoreable commitment is the one stated on 17 July 2026: commercial production by mid-2028. The scoring rules are set in advance so that they cannot be adjusted afterwards.

• The baseline is mid-2028 for commercial production, attributed to the minister's statement of 17 July 2026, not to any earlier or later restatement.
• First silicon is not scored, because no party has published a first silicon date to score against. If one is published, it enters the ledger on the day it is published, with its source.
• A slip is recorded only against a dated public statement by a principal, meaning the company or the government, not against a target invented by coverage.
• Progress claims require a named source with a date. A construction percentage on a real estate site with no primary source behind it is not evidence, and this record has published that finding before.

By those rules, and this is the part readers usually find unsatisfying, there is no scoreable slip to report at present. The earlier 2026 talk was never a supported target, so it cannot slip; the mid-2028 guidance has not yet come due. The separate answer on whether the fab is delayed works through that argument in full, and the promise against delivery record for the wider project sits on the timeline.

What would move the date, in either direction

Six observable events would change the answer on this page, and they are worth watching in preference to statements.

• Cleanroom certification or a tool move-in announcement, which converts building progress into equipment progress.
• Valued equipment purchase orders or delivery confirmations, which are the first hard evidence of an installation schedule.
• Hiring patterns for process and equipment engineers attached to the Dholera ramp, which precede tool qualification by months.
• A process design kit release, which signals that the process is stable enough to expose to customers.
• A first customer qualification announcement, which is the true gate on commercial production.
• A dated schedule statement from the company itself rather than from the ministry, which is the source class this record is thinnest on and the one that would carry the most weight.

Until one of those lands, the answer stands as written: no supported first silicon date, commercial production guided to mid-2028, a plant in envelope and fit-out at roughly half its civil work in mid-2026, and a record that will score the target rather than repeat it.