The answer desk

Is the Dholera fab delayed?

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

Against the only attributable baseline, no. No company or government document has published a first-silicon date for the Dholera fab, so the December 2026 first-chip talk cannot be scored as a slip. Commercial production is guided to mid-2028, stated by Ashwini Vaishnaw and reported on 17 July 2026, and that is the first specific target this record will score. Trade coverage did describe the mid-2028 guidance as a delay against earlier expectations, and civil work was reported at about 50 percent in mid-2026 with cleanroom fit-out underway.

Delayed against what? The baseline problem

A delay is a measurement, and a measurement needs a baseline. Before anyone can say the Dholera fab is behind schedule, there has to be a schedule: a date, published by a party in a position to commit to it, against which reality can be compared. That is where this question falls apart, and the reason it falls apart is more interesting than the yes or no most coverage reaches for.

The public record contains ministerial expectations, a chairman's letter, press reporting and a large volume of repetition on property and search-optimised sites. What it does not contain, as far as this desk has been able to verify, is a company-issued construction and production schedule for the plant. So the honest answer is that no scoreable slip exists yet, and this page sets out the scoreboard that will make future slips scoreable, along with the movements in expectation that have genuinely occurred and should not be waved away.

What was actually stated, and by whom

The dated statements that exist, in order.

• 29 February 2024. Union Cabinet approval of the plant under the India Semiconductor Mission, at Rs 91,000 crore, up to 50,000 wafer starts a month, 300mm, stated node range 28 to 110nm DURABLE.
• March 2024. Ground-breaking at the Dholera site.
• 5 March 2025. Fiscal Support Agreement signed, the document that cites Rs 91,526 crore.
• Year ended March 2025. The Tata Sons chairman's annual report letter describing a decision to start the chip journey at the 28nm node REPORTED.
• About April 2026. The Tata Semiconductor special economic zone notified over 66.16 hectares, reported by Business Standard and the Times of India.
• 16 May 2026. The lithography memorandum of understanding with ASML, announced by both the ASML and Tata Electronics newsrooms, naming nodes of 28, 40, 55, 90 and 110nm REPORTED.
• 17 July 2026. Bloomberg reports the fab will open mostly at 90nm, with commercial production expected in mid-2028; the minister's guidance of commercial production by mid-2028 is reported the same day; TrendForce carries the report on 20 July 2026 TARGET.

Read that list carefully and one thing is missing throughout: a first-silicon date issued by the company. Approval documents commit money and capacity. The chairman's letter commits to a node philosophy. The equipment agreement commits to a relationship. None of them commits to a date on which wafers come out of the line.

Where December 2026 came from, and why it cannot be scored

The December 2026 first-chip expectation is the most repeated number in this subject. It traces to ministerial statements rather than to any company schedule. Our research pass of 24 July 2026 found that line still carried on the Gujarat government's official Dholera news page REPORTED, and property sites have amplified it into an apparent fact with a precision the underlying statement never had.

This record cannot score a slip against it, for a reason that is a discipline rather than a technicality. A ministerial expectation is a statement of intent by a party that is not building the plant. Scoring it as though it were a corporate commitment would mean holding a company to a date it never published, which is as unfair as letting it escape a date it did publish. The rule this desk applies is symmetrical: targets are scored against the party that stated them, and stated at the tier appropriate to who stated them.

What can be said is stronger than nothing. The public expectation clearly moved. Trade coverage said so directly: TrendForce's report of 20 July 2026 framed commercial production as delayed to mid-2028, and the Indian trade outlet DQIndia carried the same framing REPORTED. Anyone who followed ministerial statements in 2024 and 2025 and then read the July 2026 reporting experienced a real change in the expected timeline, and it would be evasive to pretend otherwise. The distinction this page insists on is between the expectation moving, which happened, and a company missing its own published date, which cannot be demonstrated because no such date exists.

The two milestones people are conflating

Much of the confusion collapses once first silicon and commercial production are separated. First silicon is the first wafers to run the complete process flow and come out electrically functional. It is compatible with low yield, no customer qualification and no revenue. Commercial production is output that a paying customer accepts, at yields that make the arithmetic work, after qualification.

Between those two events sits process qualification, yield learning and customer qualification, and in the industry generally that gap is measured in quarters rather than weeks. A December 2026 first-silicon expectation and a mid-2028 commercial production guidance are therefore not automatically contradictory: they could describe the same plan with roughly eighteen months between the two events, which is arithmetic on the two dates rather than a claim about the plan. Whether they do is exactly what no public document establishes. The mechanics of the gap are set out on our why fabs slip page.

The construction record, dated

What is verifiable about the build is narrower than the confident percentages circulating online, and it is worth setting down precisely.

• The foundation redesign. The Economic Times reported on 28 November 2025 that geotechnical specialists Fugro, Cengrs and Geo Dynamics were engaged to redesign the main fab foundations following soil testing REPORTED. The site sits on the coastal Bhal, and redesigning foundations after site investigation is an ordinary engineering event rather than a scandal. What has not been published is its schedule effect, and this record does not estimate one. The investigation is on our foundation redesign page.
• Civil progress. About 50 percent as of mid-2026, with cleanroom fit-out underway REPORTED. This record uses that characterisation rather than the confident month-specific percentages that circulate, because those trace to sites that sell plots rather than to the company or the government.
• Cleanroom contractor. The Economic Times reported in June 2026 that Malaysian cleanroom specialist IAQ Group had been onboarded to design and install the cleanroom REPORTED. Onboarding a contractor and installation being underway are different events, and the distance between them is where a great deal of progress reporting quietly lives.
• Equipment orders. No confirmed, valued semiconductor equipment purchase orders for this fab are public as of August 2026 REPORTED. Tool orders are frequently unannounced, so this is not alarming on its own, and it does mean anyone citing tool deliveries as evidence is citing something not on the record.
• The land and infrastructure layer. The NICDC Delivery Monitoring Unit report of 30 June 2026 records 14 plots allotted across about 545 acres, of which about 476 acres is industrial, with Tata Chemicals as the named anchor allottee.

The catalogue of construction claims this desk went looking for and could not confirm, including the widely quoted completion percentages and named engineering contractors, is published as a finding on our page of claims we could not verify. The full dated build ledger sits on the construction timeline.

The scoreboard

This is the table this record will score, and it is published before the outcomes so that the criteria cannot be adjusted afterwards.

Four milestones with dates have been met. One target is open and will be scored. Two have no published date and therefore cannot be scored, and this record will say so every time rather than borrowing a number from a seller's blog to fill the row.

Three ways this question gets asked badly

Against a seller's percentage. The most common form is a comparison between a completion percentage quoted on a property site in one month and another quoted somewhere else in a later month, from which a reader infers acceleration or stall. Those percentages have no published methodology, no stated scope, and in most cases no traceable origin. Two numbers with no method behind them cannot be subtracted from each other to produce a finding.

Against a ministerial date. The second form holds the company to a date a minister stated. Ministers speak about projects they support with a legitimate interest in momentum, and their statements are worth recording as expectations. Treating them as corporate schedules produces both false alarms and false comfort, and it lets the party actually building the plant avoid ever publishing a commitment.

Against the original node framing. The third form treats the 90nm reporting as evidence of delay. It is not a schedule question at all, it is a specification question, and it deserves to be scored on its own terms rather than folded into the timeline. A fab that opens at 90nm instead of 28nm has changed what it will sell, not necessarily when it will sell it. The official range remains 28 to 110nm, Bloomberg reported a 90nm opening, the company says the plan was always to begin at 55nm and 90nm, and the chairman's letter for the year ended March 2025 said the journey would start at 28nm. Those statements are not reconcilable from public sources, and this record prints the contradiction rather than resolving it. Whether opening at 90nm is a weakness is a separate argument, taken up separately by this desk, and the short version is that non-leading-edge is not the same as obsolete.

Underneath all three of those errors sits a method, and this record states it deliberately narrowly. A target enters the scoreboard only when it is specific, dated and attributable to a named party, and it is scored against that party rather than against the project in general. When the window closes, the target is marked met, missed or superseded, with the evidence and its date attached, and a missed target stays on the page rather than being quietly replaced by a newer one. Where a target moves before its window closes, the movement itself is recorded, because a series of revisions is often the most informative thing about a schedule.

The reason for publishing the method before the outcome is that scoring rules invented after the fact are worthless. This subject has an unusually large number of interested parties on both sides, sellers who need the timeline to be short and critics who need it to be broken, and a record that decides its criteria in advance is the only kind that can be checked by either.

Which slip risks are actually live

Saying no slip is scoreable is not the same as saying the schedule is safe. The risks that historically move fab dates are specific, and several of them are live here.

• Tool lead times and hookup. Equipment delivery slots for a global build-out wave are contended, and installation and hookup regularly take longer than the plan. With no public order record, this risk is invisible from outside.
• Process qualification. Bringing every module inside its window across chambers and shifts is the classic source of quiet quarters.
• Talent. Engineers who have taken a plant through a full ramp are scarce everywhere, and scarcer in a country building its first large commercial logic fab.
• Utilities and supply. Ultrapure water, bulk and specialty gases and chemical logistics all have to be operating at fab standard before qualification, not after.
• Customer qualification. Even a plant producing on time can miss a commercial-production date if the parts have not cleared a buyer's qualification programme.

What would change the answer

Three kinds of evidence would convert this page from no-scoreable-slip to something more definite. A company-issued schedule with dates, which would give the record a real baseline. A dated statement moving the mid-2028 guidance, which would be a scoreable slip against a target this record has already logged. Or evidence of production, meaning a first-silicon announcement, a customer qualification, or tool deliveries, which would score the target in the other direction.

Until one of those arrives, the position stands as stated at the top. The fab is not demonstrably late against any baseline it published, because it published none. The public expectation did move from an end-2026 first chip to a mid-2028 commercial production guidance, trade press called that a delay, and this record logs both the movement and the reason it cannot be scored as a broken corporate promise. The narrower date question is handled on when the Dholera fab opens, and the node contradiction that runs alongside it on the node question.