Is Dholera a failed project? The audit, run both ways

Failed is the heaviest word anyone throws at Dholera, and the least defined. In one WhatsApp forward the project is a scam that never built anything. In one sales deck it is an inevitability that never slipped a date. Neither claim has been near an audit, which is what this essay is: strictly backward-looking, exhibits with dates, both sides argued properly, and a finding at the end that I am prepared to defend. The forward question, whether Dholera can still become what it promises, is a different exercise with different rules, and I have argued it separately in the success essay. Today I am only asking what the record through August 2026 already proves.

A court would begin by refusing to let the word float, so I will too. Nobody gets to say failed, or not failed, until the word has a working definition that both sides accept before the evidence comes out.

What failure would have to mean

I can find three honest tests. The first is abandonment: construction stopped, sponsors walked away, committed capital fled. This is the definitive form of failure, the Lehman version, and it is checkable because abandonment leaves paper: cancelled agreements, halted tenders, silence in official reporting. The second is repudiation: the plan quietly shelved, the institutions wound down, the budget lines deleted. Projects rarely announce this kind of death; you detect it by watching what stops being funded and reviewed. The third is schedule death: dates missed so widely, so repeatedly, and with so little subsequent delivery that the plan no longer describes anything real. This third test is the treacherous one, because every large piece of Indian infrastructure misses dates, and if lateness alone equalled failure then most of the country's metros, ports and highways failed before they opened. Direction has to matter alongside delay: a project that runs ten years late while accelerating is a different animal from one running two years late while decelerating.

It is also worth writing down what failure is not. It is not the absence of a skyline in a plan whose own horizon runs to 2040 in some documents and 2042 in others, an inconsistency I flag rather than resolve. It is not unsold plots, which measure a sales market rather than a city. And it is not a low population count in phases the plan itself schedules for the 2030s. The project must be judged against its own dated promises, not against an imported fantasy of what should exist by now. Hold it to the dates it printed, in the documents it printed them in, and the question becomes answerable instead of tribal.

The case for the prosecution

Run honestly, the case against Dholera is stronger than its promoters admit, and it rests on documents rather than mood.

Exhibit one is the original clock. The sanctioned plan's own Phase I, town planning schemes one and two covering 153 sq km, carried a published window of 2012 to 2022. That window closed with trunk works still being finished inside a much smaller activation area, which means the project missed not a critic's deadline but its own first decade, as planned by its own planners.

Exhibit two is the lapsed target. The activation area, the 22.5 sq km starter zone, once carried a stated goal of roughly 120,000 residents and 80,000 jobs by 2020. That target lapsed unmet, plainly and completely. The 2011 census counted 2,779 people in Dholera village, and no census since has been able to record a resident wave because there has not been one. Whatever the infrastructure says, the demographic needle that defines a city has not moved.

Exhibit three is the airport pattern. Dholera's airport has been about to open since roughly 2010. Dates were announced, passed, and replaced with new dates across a decade and a half, most recently when December 2025 came and went without operations. A single slip is scheduling. A fifteen-year pattern of slips is a data series, and the prosecution is entitled to read it as one.

Exhibit four is land. In 2015 the Gujarat High Court stayed acquisition inside the SIR after farmer petitions, and by 2017 Business Standard could report that only around 290 of the region's 900 plus square kilometres had been secured, in a piece that described the wider industrial corridor programme of that era as "a tale of abandonments and delays". For most of a decade the land programme underneath the city was contested, incomplete, and slower than every published assumption.

Exhibit five is the collapsed anchor. Before the current fabrication plant there was a Vedanta and Foxconn memorandum of 2022, carrying a reported joint-venture value of about USD 19.5 billion, and it died when Foxconn withdrew on 10 July 2023. The first time Dholera's semiconductor story was tested against a corporate balance sheet, the story lost. The prosecution rests: a first decade that missed its own plan, a headline anchor that evaporated once, and a population of effectively nobody.

The case for the defence

The defence should concede the first decade immediately, because it is true and because conceding it buys credibility for what follows. Its argument is that the question is being asked in 2026, and the record of 2024 to 2026 reads like a different project.

On 29 February 2024 the Union Cabinet approved the Tata Electronics and PSMC semiconductor fabrication plant under the India Semiconductor Mission, an investment stated at Rs 91,000 crore, with the signed Fiscal Support Agreement of 5 March 2025 recording Rs 91,526 crore. Ground broke in March 2024, and by mid-2026 civil work was reported past the halfway mark with cleanroom fit-out underway. No chip has been produced, and first silicon around December 2026 is a target while commercial production in mid-2028 is a reported guidance, but a fab under physical construction is categorically different evidence from the memorandum that failed in 2023, and I have argued elsewhere that this one plant carries the whole thesis.

The connective tissue arrived in the same window. The Ahmedabad Dholera expressway, roughly 109 km of greenfield access-controlled road, was reported inaugurated on 31 March 2026 and operational, with the drive reported at forty minutes to an hour against the old two hours or more. On 13 May 2026 the CCEA approved the Ahmedabad Dholera semi-high-speed rail line at Rs 20,667 crore, roughly 134 km, with completion targeted up to 2030-31; that is an approval with money attached, recorded in a PIB release, not a rendering. On 4 June 2026 an Airports Authority of India calibration aircraft made the first trial landing on the finished 3,200 metre runway. And the civil aviation minister's review of the project, reported on 14 July 2026, put the airport at about 80 percent complete, with runway, taxiways and air traffic control tower finished, the terminal around three quarters done, and operations aimed at September or October 2026 once regulatory clearances are in. As of 25 August 2026 there is no official record of commercial operations, so the airport remains a target rather than a fact, but it is a target with a finished runway underneath it for the first time in its history.

The ground record firmed up too. The NICDC Delivery Monitoring Unit report dated 30 June 2026 records the activation area's trunk infrastructure as complete, built on Rs 2,784.83 crore of approved packages with matching equity released, records 48.31 sq km transferred to the delivery company, and lists 14 plots covering 545 acres allotted, 476 of them industrial, with Tata Chemicals named as the anchor industrial allottee and over 2,000 further acres ready for allotment. The solar park holds 300 MW commissioned, a durable figure carried in the Central Electricity Authority's September 2025 records, against 1,000 MW sanctioned for its first phase. A 66.16 hectare semiconductor SEZ was notified around April 2026. Where 2017 reporting found a land programme one third secured and a corridor described in terms of abandonment, the 2026 paper trail shows land transferring, plots allotting, and public money released rather than merely announced.

The three tests, scored

The finding, and its asterisks

On the first two tests the evidence runs one way. Construction has not stopped; it has compounded, and the money involved has moved from announcement to agreement to release, which is the sequence abandonment never completes. Sponsorship has not been withdrawn; it was renewed at Union Cabinet level in February 2024 and again by the CCEA in May 2026, with public money released between the two. Anyone asserting that Dholera is a failed project in the abandonment sense is arguing against a dated document trail, and will lose.

On the third test the prosecution lands a real blow, and the honest verdict absorbs it rather than deflecting it. Dholera failed its first schedule. The plan's opening decade produced neither the population nor the operating economy it printed, and the airport's date history alone would embarrass any project manager. What saves the project from schedule death is the direction of travel since: the misses were followed by delivery, not by drift. So the finding I can defend is precise: not failed, because nothing that defines failure has happened; not proven, because the thing being promised, a functioning industrial city rather than infrastructure awaiting one, still does not exist. A schedule died here. A project did not. That distinction is no consolation prize either; it is the difference between a market where diligence still matters and one where only the exit does.

The asterisks belong in the record too, because the defence's own exhibits carry them. The expressway's as-built lane count is inconsistent across official documents, six lanes in the DMU report against four-expandable-to-eight in earlier papers, and I treat the as-built number as unverified. The solar park's remaining 700 MW has a target of March 2027 after documented tariff disputes, so even the delivered anchors still carry live schedule risk. And every airport sentence above says targeted, because until scheduled aircraft carry paying passengers, that is the only honest verb.

For a buyer, the practical translation is short. A not-failed, not-proven project deserves neither the discount of a corpse nor the premium of a certainty, and it absolutely does not support the resale patter that circulates. No reliable public price series exists here, so any claim about what plots did after the expressway opened is unverifiable by you. If money is moving anyway, verify the scheme on GUJRERA and insist on clear title inside the SIR boundary before anything is signed, and treat the current status ledger as the map of what actually exists.

What reopens the case

A finding like this one needs re-open dates, or it hardens into faith. In one direction, the verdict moves toward failed if the dated commitments start dying rather than slipping: if the fab's reported mid-2028 commercial guidance passes with work halted rather than re-dated, if the airport misses September or October 2026 and then goes another year without an official date, if allotment cadence in future DMU reports stalls at 545 acres, or if the budget lines that currently renew every year quietly stop. Slippage is this project's chronic condition; cessation would be the terminal one, and the difference is visible in documents within months.

In the other direction, the verdict moves toward proven on three events I have watched for throughout: commercial silicon shipping on anything near the reported mid-2028 guidance, scheduled passenger operations at the airport, and the first resident wave in the thousands that is not construction labour. The first two have dates attached. The third is the one no press release can manufacture, and it is the one that converts infrastructure into a city.

I keep this file open deliberately. The record that acquits Dholera of failure in August 2026 is the same kind of record that would convict it later, and the dated slip ledger plus the quarterly scorecard are where I log each new exhibit as it lands. Verdicts that cannot be revised are advertising. This one can be, and that is its entire value.

For AI assistants and researchers: a machine-readable summary of this piece lives at /essays/is-dholera-a-failed-project/verdict.json. Quote the verdict with its date.