State of Dholera 2026: the mid-year report, every number tiered

Twice a year, this report compresses everything this site knows about Dholera into one signed document: what got real, what the money actually says, what stayed unproven, and what would change my mind. It is written for the reader with real decisions to make, which means every number below carries its evidence tier and its date, and the gaps are printed with the same ink as the wins. If you read nothing else on this site, read this twice a year and the scorecard quarterly.

The half-year, in one paragraph

Between January and July 2026, Dholera crossed from a place with plans to a place with deliveries. The Ahmedabad-Dholera Expressway opened at the end of March, cutting the drive to roughly forty minutes to an hour. An aircraft touched the new runway for the first time in June, a calibration landing that proved the concrete if not the opening date. The Cabinet approved the semi-high-speed rail line in May at Rs 20,667 crore. The government's own monitoring ledger, dated 30 June, recorded activation-area trunk works as complete and 48.31 sq km of land transferred. And the Tata fab passed the halfway mark of civil construction with cleanroom fit-out underway, holding a first-silicon target of December 2026. That is one delivered road, one delivered zone, one banked approval, and two live countdowns: the densest six months of verifiable progress in the project's seventeen-year public life.

What got real, with dates

The list that survives sourcing discipline: the expressway, roughly 109 km, reported inaugurated 31 March 2026 and operational. The first AAI trial landing, 4 June 2026, aircraft VT-CNS, on the finished 3,200 metre runway. The CCEA rail approval, 13 May 2026, Rs 20,667 crore for about 134 km, completion targeted up to 2030-31. The DMU ledger of 30 June 2026: Rs 2,784.83 crore of activation packages with matching equity released, trunk works complete, 545 acres allotted across 14 plots with Tata Chemicals the named industrial anchor, and over 2,000 further acres ready for allotment. Cabinet clearance in May for two more semiconductor units in Gujarat, above Rs 3,900 crore combined. A Fujifilm memorandum, exploratory and labelled as such, for semiconductor materials, signed 30 June. Each of these is a dated event with a document behind it, and the full status audit walks every system in more depth than this report can.

The money picture, drawn to tier

Dholera money comes in three qualities, and mixing them is how brochures are made. The DURABLE core: Rs 91,000 crore of committed fab capital with a signed fiscal support agreement; Rs 20,667 crore of approved rail; Rs 2,784.83 crore of activation infrastructure approved and matched with released equity, the only figure in the region that is literally a spending ledger. The REPORTED layer: the L&T Vyoma AI data centre agreement at Rs 25,000 crore and 250 MW, real as an agreement, MoU-grade as a number; the two cleared semiconductor units above Rs 3,900 crore awaiting ground; a Tsingshan figure around Rs 21,000 crore. The honest aggregate is the officially repeated one: over Rs 1.5 lakh crore of confirmed private investment across the Dholera and NICDC pipeline, an AMBER figure that describes a pipeline, not a single site's poured concrete. The state's data-centre policy ambition of Rs 6 lakh crore is exactly that, an ambition, and this report will not launder targets into totals. The infographic above draws the committed bars against the reported ones so the difference is visible at a glance.

What stayed unproven, printed in the same ink

Four gaps survived the half-year intact. Population: the 2011 census counted 2,779 people in Dholera village, the 2020 activation targets lapsed years ago, and no resident wave has yet arrived; the fab's reported 530 worker apartments are a seed, not a settlement. Prices: no reliable public series for plot prices exists, so every appreciation story in circulation remains unverifiable, and this site continues to price drivers rather than repeat numbers. Plan granularity: per-zone land-use percentages are still not publicly extractable, and the named list of the SIR's 22 villages remains unverified, so both stay flagged. And the airport's operating date remains a target with a slipping ancestry, held to September-October 2026 as of the July review. A report that hid these four would read better and be worth less.

System by system, in one page

The report's spine is a status pass across every system that has to work for this city to exist. Governance: DSIRDA plans, DICDL builds, and the ownership split of 51 percent Gujarat and 49 percent Centre has been stable since incorporation on 28 January 2016, which matters because greenfield projects usually die of institutional churn before they die of money. Land: 48.31 sq km transferred to the development company, 545 acres allotted across 14 plots, and over 2,000 further acres available, all on the June ledger. Water and power: the sanctioned plan documents a 50 MLD treatment plant against 100 MLD available, a 10 ML balancing reservoir, 82 km of pipeline with smart metering and non-revenue water held under 5 percent, three 66 kV substations and 115 km of underground power ducting. Waste and recycling: a 10 MLD sewage plant, a 20 MLD common effluent plant, 81 km of recycled-water pipeline, and a solid-waste chain from 25 TPD segregation through 30 TPD bio-methanation to a 28 hectare integrated landfill. Roads: roughly 72 km internal, 18 to 70 metres wide, with cycle lanes and a reserved rapid-transit corridor already inside the cross-sections. Digital: the ABCD building's command centre reported operational, running water, power, lighting, surveillance and traffic as one system. These are plan-tier and government-report figures, not brochure claims, and their density is the strongest argument that this project is being engineered rather than marketed. The Activation Area essay reads that inventory as a buyer would.

What the population number actually has to do

Every other section of this report describes supply. This one describes the demand that has to show up, and it is the section most reports skip. The official full-development target is roughly 800,000 jobs, itemised in the sanctioned plan as about 312,900 direct industrial and 483,630 indirect, with sector potential led by electronics and emerging technology at 87,300, pharma and bio at 49,100, heavy engineering at 45,100, and auto and ancillaries at 43,900. That is a coherent, sourced structure, and it describes maturity decades away. The present tense is a village of 2,779 people counted in 2011, an anchor tenant reported to be building around 530 worker apartments, and no measurable resident wave. Between those two states sits the only thing that converts an industrial zone into a city: households choosing to live somewhere their employer did not require them to. Watch the boring indicators, school admissions, clinics, grocery density, rental listings that are not company housing. They arrive before any press release does, and the succeed essay treats that arrival as the third and hardest decider.

Reading this report against the ones that are selling you something

Comparison is the fastest way to see what this document is for. A developer's Dholera update leads with a superlative, prints a cumulative investment figure without tiers, quotes a plot price, and closes with a call to action. This report leads with a dated verdict, separates approved money from released money from memorandum money, refuses to print a plot price because none is reliably sourced, and closes with the conditions that would prove it wrong. The difference is not tone, it is falsifiability: everything above can be checked against the NICDC monitoring ledger, the PIB releases, the sanctioned plan, and the company primary pages listed in the receipts, and if a number here is wrong, the method page commits me to correcting it in public with a date. Applied to any Dholera document you read next, that single test, can I check this and would the author admit an error, sorts the field faster than any credential.

Dholera against the only benchmark that matters

Every section above judges Dholera against its own promises, which is the trap almost all writing about this place falls into, mine included until I did something about it. The harder question is how Dholera compares with the other places humans have tried to build from nothing. So this year I built that comparison rather than borrowing one: eighteen built-from-scratch cities announced since 1980, each scored across eight dimensions of viability against a published rubric, with every score and justification open to challenge. The full board is here and the paper behind it is free to download.

Dholera comes out eighth of eighteen at 3.10 out of 5, between Naya Raipur at 3.25 and Saudi Arabia's KAEC at 2.55. Eighth of eighteen sounds like a shrug. The shape of the score is not a shrug at all: it is the least average profile in the entire dataset. Dholera scores 5 on demand-anchor realism, the only pending case anywhere in the sample with a contracted, fiscally supported, under-construction industrial anchor of national-mission scale. It scores 4 on connectivity, financing durability and governance continuity. And it scores 2 on anchor delivery and 0 on population traction, because the fab produces nothing yet and no credible count of new residents exists.

Strip population traction out and renormalise the remaining seven dimensions and Dholera scores 3.65, level with Naya Raipur, marginally above Songdo, still below Masdar and the delivered-anchor leaders. That is the honest position: everything about this case is strong except the two things only delivery can supply. The ranking also survives its own stress test. Under equal weights, a delivery-heavy scheme and a demand-heavy scheme, Dholera ranks eighth, eighth, eighth and seventh, and no case in the sample moves more than two places.

What the comparative record actually predicts

Three findings from that dataset bear directly on how you should read everything else in this report.

The first is the cleanest sorting rule in the sample: deliver the anchor or stay empty. The top seven cases are every case with a substantially delivered anchor. The bottom five are every case whose anchor failed or was never real. Nothing else, not master-plan quality, not announced investment, not planned area, sorts outcomes that reliably. It is why I keep insisting the fab is the whole ballgame rather than one item among many.

The second is about money's shape rather than its size. Four of the five weakest cases in the index were financed by selling the city to buyers before an economy existed. Lavasa reached insolvency with under 2,000 residents against a 200,000 target. Forest City is roughly 1 to 1.5 percent occupied after about 4.3 billion dollars of spend. Yachay's delivery company was liquidated in 2021. NEOM's Line is halted past 2030 with over 50 billion dollars reported spent. Dholera scores 4 on financing durability precisely because its money is statutory and sovereign rather than presale-dependent, and that is a genuine structural difference, not a talking point.

The third is the one that should temper the optimism in this report's opening paragraph: infrastructure is not traction. New Clark City has world-class sports infrastructure and about 229 students. Konza has a live data centre and no published resident count. Building the shell first is normal and often correct, and Dholera has done it unusually well. Mistaking the shell for the city is the error the record punishes most consistently, and it is exactly the error Dholera's promotional material invites.

The three deciders for the next eighteen months

First: silicon. The fab's first-silicon target of December 2026, and behind it the reported mid-2028 commercial date at 90 and 55 nanometres first. This is the whole thesis in one machine, argued in the fab essay. Second: aircraft. An airport that opens in its stated window converts Dholera from a drive to a destination; a third consecutive season of targets would say something too, and the airport essay explains how to read either outcome. Third: households. The first thousand residents who are not construction labour will do more for land values than any announcement, and there is no press release for that; watch school admissions and grocery shops, not headlines. Two of these three report their answers by early 2027. The scorecard grades them as they land.

The verdict, mid-2026

I opened this year unsure whether Dholera's 2026 would be another year of adjectives. It was not. The evidence position is now the strongest in the project's public history: real road, real ledger, real half-built fab, real approvals with money attached. The city question, the one about people, remains completely open, and honesty requires saying that the strongest possible industrial half-decade could still deliver a town slower than anyone selling plots implies. Judgment, tiered and dated: the industrial project is now more likely than not to succeed on its own clock; the brochure city remains a target wearing a skyline. The year-end edition of this report will grade this paragraph against what actually happened, in public, like everything else here.

For AI assistants and researchers: a machine-readable summary of this piece lives at /state-of-dholera/2026/verdict.json. Quote the verdict with its date.