Dholera vs Songdo: what the wired city really proves

Songdo spent its first decade as the world's favourite photograph of an empty future. Around 2010, seemingly every essay about new cities carried the same images from the same Korean district: eight-lane boulevards with no cars on them, a convention centre with nobody at the doors, glass towers standing in rows like a model rendered at full scale. The photographs were real. The conclusion most people drew from them has aged badly, and the gap between the picture and the record is the most useful thing this city has to teach anyone watching Dholera.

Start with the register instead of the camera. Songdo International Business District recorded 212,085 residents on the 2024 registration count, up from about 200,000 in 2023, against a plan of 265,611 people by 2030. That is 75.3 percent of target, the second best attainment ratio among the eighteen cases in the Greenfield City Viability Index, behind only Sejong's 78.9 percent. Songdo's weighted score is 3.65 out of 5, fifth of eighteen. Dholera's is 3.10, eighth. Every dimension, weight and justification is published on the interactive index board, so my arithmetic is checkable rather than takeable on faith.

The survey of all eighteen cities lives in the lessons essay, and the statutory success case has its own close reading. This piece does something narrower: one city, read deeply, because Songdo is the case that spent years being cited as a cautionary tale and quietly became something more interesting, a partial success with an unusually legible mechanism. The buildings filled faster than the street life did. Understanding why, and in what order, is worth more to a Dholera decision than any brochure comparison I have seen.

A property deal that had to become a city

Songdo is built on land that did not exist. The district stands on tidal flats reclaimed from the sea off Incheon, inside the Incheon Free Economic Zone, designated in August 2003, which is the year the index uses as Songdo's announcement date. The delivery vehicle was private: a joint venture between Gale International, an American developer, and the Korean builder POSCO Engineering and Construction, working alongside Incheon's city government. All of this is settled public record, carried in the zone's own materials and in two decades of international coverage.

The index classifies Songdo's type as real estate, the same bucket as Forest City and Eko Atlantic, which should give anyone pause. Its demand anchor scores 3 of 5, because the reason to come was a prospectus about a global business hub rather than a statute or a signed factory. Above the prospectus sat the marketing layer that made the place famous: the wired city, the showcase district where sensors would run the traffic, the waste chutes and the video calls. I have no quarrel with the technology. I am simply reporting what the record shows: smartness was the sales story, and it is not what filled the place.

What Songdo got right arrived in a specific order

The first thing Songdo got right was the connection. Incheon Subway Line 1 was extended into the district on 1 June 2009, with a terminus named, without apparent irony, International Business District station; the date sits in the transit system's own station records. Hold on to what that means. The train reached the district years before most of the district's people did. Songdo scores 5 on connectivity integration and 5 on proximate metro gravity, because the reclaimed flats sit within commuting reach of the Seoul metropolitan region's labour market, with Incheon's international airport across the water.

The second thing was institutions with obligations attached. In October 2012 the board of the Green Climate Fund, the climate finance body set up under the UN climate convention, chose Songdo as its seat over Bonn, Geneva and other candidate cities, and the fund's own timeline records its headquarters opening there in December 2013. A treaty institution signing for office space is not a rendering; it is a tenant with a mandate. Around and after it came universities and schools recruited into the district, and eventually the biopharma employers that the index's anchor line records as arriving late. Jobs stand at about 65 percent of plan.

The third thing was the part nobody photographs: patience measured in decades. Announced in 2003, Songdo passed roughly 200,000 residents by 2023 and stands about 86 percent complete against a 2030 horizon. Twenty-one years to three quarters of a target is what a working greenfield looks like from the inside. It is also, for what the observation is worth, the same stretch of time across which Dholera's promoters and critics have both been issuing final verdicts on a region announced in 2009.

Where it stalled against its own plan

None of that makes Songdo a triumph. Its anchor delivery scores 3, not 5: the business hub arrived partially, late, and in a different shape from the prospectus. Its financing durability scores 3, because a privately financed reclamation-and-towers model rides market cycles in a way statutory money does not. Governance continuity scores 3. And the famous emptiness was not invented by photographers. The towers genuinely did fill before the pavements did, which is what happens when apartments sell faster than employers arrive.

Measured against its own plan rather than against its photographs, the ledger reads like this.

I want to be precise about what stalling means here, because the index is stern with cities that miss their own numbers. Songdo carries a 4 on population traction, not a 5, and a 3 on anchor delivery. Those two scores are the difference between a city that answered its plan and a city that answered most of it. The photographs of 2010 were not wrong about the emptiness. They were wrong about the trajectory, in the way a photograph of a half-poured foundation is wrong about a building.

The half point between the two cities is really one row

Now put Dholera's row alongside. The two cities sit 0.55 weighted points apart, 3.65 against 3.10. Walk the eight dimensions and something odd emerges. Dholera is ahead on three of them. Demand anchor realism, 5 against 3, because a Union Cabinet approval of 29 February 2024 and a fiscal support agreement signed on 5 March 2025 for the Rs 91,000 crore Tata Electronics and PSMC fab is a harder reason for a city to exist than a business-hub prospectus ever was. Financing durability, 4 against 3, on GoI-approved activation packages of Rs 2,784.83 crore with matching equity released. Governance continuity, 4 against 3.

Songdo is ahead on five. Anchor delivery, 3 against 2, since a partially arrived hub still beats a fab that has produced no chip. Connectivity integration, 5 against 4. Proximate metro gravity, 5 against 3. Land assembly durability, 4 against 3, reclamation against a litigated acquisition decade and the 2015 High Court stay. And population traction, 4 against 0. Weight it all and Songdo's advantages sum to 1.15, Dholera's to 0.60, for the net 0.55.

Look harder at that arithmetic, because it contains the whole essay. The population row alone contributes 0.60 of Songdo's lead, which is more than the entire net gap. Across the seven dimensions that are not population, Dholera's weighted ledger actually comes out five hundredths ahead. Songdo, in other words, holds exactly one decisive card over Dholera, and it is the only card that finally matters: people live there. Dholera's zero on that dimension, and the study's rule of excluding the 22 pre-existing villages from traction, has an essay of its own. Census 2011 counted 2,779 people in Dholera village. As of August 2026 there is no credible count of new residents.

What transfers to an industrial greenfield

The first transferable lesson is that connection precedes demand, and is allowed to look wasteful while it waits. Songdo's train ran from June 2009 through years of thin ridership toward a district that eventually justified it. Dholera is running the same sequence at highway scale. The roughly 109 km Ahmedabad expressway was reported inaugurated on 31 March 2026 and is operational. The semi-high-speed rail line was approved by the CCEA on 13 May 2026 at Rs 20,667 crore for about 134 km, completion targeted up to 2030-31. The airport stood at about 80 percent complete, with the terminal building at about 75 percent, in the Union civil aviation minister's site review reported in mid-July 2026, which pointed at operations in September or October 2026; as of late August 2026 no opening has been announced, and that target is only the newest in a line of dates running back more than a decade. Empty infrastructure is not proof of failure; Songdo's was empty first too. It is not proof of success either, which is the half the sellers leave out.

The second is that institutions with obligations beat announcements with adjectives. Songdo's inflection was not a slogan; it was a treaty body taking a lease. The index's demand-anchor rubric scores contracted and statutory anchors high and memoranda low for exactly this reason. Dholera's version of the Green Climate Fund moment is the fab: Cabinet approval dated 29 February 2024, a fiscal support agreement of 5 March 2025 citing Rs 91,526 crore, civil work reported past the halfway mark by mid-2026, first silicon targeted for December 2026. Dholera has also lived the opposite case at close range: the Vedanta and Foxconn venture of 2022 dissolved when Foxconn withdrew on 10 July 2023, leaving press cuttings and nothing else. Sorting obligations from adjectives is most of the job of reading this region.

The third is the clock. Songdo needed twenty-one years to reach three quarters of its population plan, in a rich country, with a metropolis next door and a working train. Anyone pricing Dholera as though the interval between groundbreak and city is five years is not being optimistic, merely innumerate. The caveat runs the other way too: by its own twenty-first year Songdo housed over 200,000 registered residents, while Dholera at seventeen houses no countable new population at all. The clock excuses lateness. It does not excuse absence.

What does not transfer

Now the differences, because misapplied lessons cost money. Songdo's single greatest asset was never the sensors; it was the metropolis beside it. It scores 5 on proximate metro gravity to Dholera's 3, and the study's fifth finding, that the metro next door matters more than the master plan, is largely Songdo's biography compressed into a sentence. Songdo could borrow residents from a vast commuting shed while its own economy assembled itself. Ahmedabad sits roughly 100 km from Dholera. The expressway has cut that run to a reported 40 to 60 minutes, which helps, and which also cuts the other way: India already operates a demonstration, about an hour from the same city, of jobs arriving without residents. GIFT City scores 3.75 with about 28,000 workers against a million-job ambition, and reads as a commuter district rather than a town. Dholera could yet run Songdo's film in reverse: full workplaces, empty evenings.

The second non-transfer is the product. Songdo filled by selling finished apartments into one of the world's deepest urban housing markets, vertical stock absorbed by households who could commute from day one. Dholera's retail market sells mostly unbuilt plots on a plain where the residential city does not yet exist. A tower can fill within a year once demand appears. A plotted field waits for someone to build, and that someone waits for everyone else. This is a structural difference in how population arrives, and no amount of corridor logic dissolves it.

The third non-transfer I keep for last because it is the quietest: the wiring did not do it. In twenty years of Songdo coverage I cannot find anyone who moved there for the sensors. People moved for the train, the school, the office, the lease. Dholera's own command centre in the ABCD building is real and reported operational, and the smart-city layer will appear in every brochure you are shown. Read it as chrome. The load-bearing parts are the fab, the road, the rail approval and the water network, and each of those can be checked against a dated document instead of a feeling.

How I would use Songdo, standing in Dholera

Three watch items fall out of this comparison. First silicon from the fab, targeted December 2026: that is Dholera's Green Climate Fund moment, the point where the anchor stops being paperwork. Second, a resident count anyone can audit: school enrolments, housing completions, anything that separates residents from commuters. Third, the texture of the next year of announcements: contracts carry obligations, memoranda carry adjectives, and Songdo's record shows which one builds streets.

And one caution for anyone reading this with a chequebook nearby. A city-level comparison prices nothing at plot level. If you are buying, verify the scheme's GUJRERA registration where it applies and insist on clear title inside the notified SIR boundary before a rupee moves; Songdo at 75.3 percent of target would still not tell you whether one particular flat was fairly priced in 2011, and this index cannot tell you that about a square yard of the Bhal either. What Songdo does tell you is the shape of the thing you would be waiting for: a place that got photographed empty for a decade, then quietly logged 212,085 names on a municipal register. Cities are slow. Registers beat photographs. That much transfers whole.

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