Seventy miles north of Phoenix, two miles east of an interstate exit at a place called Cordes Junction, a cluster of concrete vaults and quarter-domes sits on the lip of a mesa above a dry ravine. There are about a dozen structures. They are unmistakably deliberate, unmistakably strange, and unmistakably unfinished: half-circles of poured concrete facing south to catch the winter sun, apses containing a bronze foundry and a ceramics studio, a five-storey building with a cafe in it, a ring of apartments, some greenhouses, and then desert. The population on any given day is somewhere between fifty and a hundred and fifty, depending on how many students happen to be on site. It was designed for five thousand.
This is Arcosanti, begun in 1970 by the Italian-born architect Paolo Soleri, and it is probably the most famous unfinished building in the world. The usual account treats it as a noble failure defeated by money, or as a cult compound that ran out of believers, or as an architectural folly that was always going to end this way. All three miss the interesting thing, which is that Arcosanti was defeated by a specific and rather beautiful piece of arithmetic embedded in its own premise. Soleri’s argument was that cities should be dense, and every single benefit he claimed for density only appears above a certain population. Below that number the benefits do not appear in reduced form; they do not appear at all. Which means a prototype could never demonstrate the thing it was built to demonstrate, and the project has spent fifty-five years as a permanent instance of that problem, sitting somewhere between the places that exist mainly on paper and the industrial utopias that collapsed outright, like Henry Ford’s town in the Amazon.
What Arcosanti Was Supposed to Be
The plan was for a city of five thousand people, rising as high as twenty-five storeys, occupying a footprint of a few dozen acres in the middle of a land preserve of several thousand. Housing, workshops, offices, performance spaces, greenhouses, and agriculture would all sit within a single continuous structure, arranged so that a resident could live an entire life inside it without owning a car, walking everywhere, with the surrounding desert left substantially untouched. Food would come largely from attached greenhouses. Heat would come from the sun, captured by the geometry of the building itself.
Arcosanti was not the first version of the idea, and that matters for judging it. Soleri had been publishing arcology designs through the 1960s, enormous speculative structures with names like Babelnoah and Hexahedron, housing populations in the hundreds of thousands and rendered in obsessive ink drawings that were exhibited to considerable acclaim. Those were explicitly unbuildable, thought experiments meant to establish a direction. Arcosanti was the deliberate step down from that: the smallest version he thought could actually function, sited on land he could afford, at a population he believed was achievable within his lifetime. It was already a compromise before the first concrete was poured.
That is a serious proposal rather than a whimsical one, and it was received as such. Newsweek described it in the early 1970s as the greatest urban experiment of its time, and the design world took it seriously enough that Arcosanti became a pilgrimage site for architecture students within a few years of construction beginning. The claim being tested was concrete: that a city built as a single three-dimensional organism would consume a fraction of the land, energy, and materials of a conventional settlement housing the same number, while producing more of the social density that makes cities worth living in. It belongs in the tradition of grand structural ambitions catalogued in the history of what humanity attempts to build, and it was one of the more intellectually rigorous entries in the long line of intentional communities, some of which are still quietly running today. The proposal was testable. That is the whole trouble.
Arcology
Soleri coined the word in 1969, in a book published by MIT Press: arcology, architecture fused with ecology. The underlying idea came from biology, and specifically from an observation about evolution that he found compelling. As living systems become more complex, they also become more compact; a mammal packs vastly more functional complexity into vastly less space than a colony of simpler organisms occupying an equivalent volume. Soleri called this miniaturisation, and argued that human settlement had gone the wrong way, sprawling outward into ever-lower density while consuming more land, more energy, and more time in transit for every unit of actual living.
It is worth noticing that the biological analogy carries an implication Soleri fully accepted and most of his admirers skip past. Organisms are not assembled incrementally from functioning fragments; a quarter of a mammal is not a small mammal, it is meat. If a city is genuinely like an organism, then a partially built city is not a small city but an incomplete one, non-functional until the whole thing is present and integrated. Soleri’s own founding metaphor therefore predicted, quite precisely, that a partially completed arcology would not work, and the fifty-five years since have confirmed his metaphor rather than refuted it.
The corrective was to build the city like an organism: compact, layered, three-dimensional, with the distance between any two functions minimised. From this everything else follows. If nothing is far from anything else, nobody needs a car, which eliminates roads, parking, fuel, and the enormous land take of automotive infrastructure. If the structure is compact, its surface-to-volume ratio drops and it becomes cheaper to heat and cool. If the city occupies a small footprint, the land around it stays wild. And, crucially for Soleri, if people are physically close together they collide more often, producing what he called the urban effect, the generative friction that makes cities the engines of culture and invention. That last claim has real support in the study of how social animals behave when population density rises, a subject running through the politics of primate groups, and the resource argument is even stronger in a desert, where the constraints on water and land as strategic resources are impossible to ignore. The theory is coherent and internally consistent. Every single part of it, however, depends on the same one variable.
The Inversion of Broadacre City
The intellectual origin of Arcosanti is a quarrel, and knowing about it explains a great deal. In 1947 Soleri arrived in Arizona with a doctorate from Turin to study under Frank Lloyd Wright at Taliesin West. The apprenticeship lasted about eighteen months and ended badly, and the disagreement was not personal so much as directional. Wright’s great late vision was Broadacre City, a proposal in which every American family would have an acre of land, connected by roads and automobiles, with the traditional city dissolved entirely into a low-density continuum.
There is a further irony in the inheritance. Soleri took from Wright not only a target to argue against but a working method: the apprenticeship model at Taliesin, in which students paid for the privilege of building the master’s projects while absorbing his philosophy. Soleri adopted that arrangement wholesale and ran it for over forty years. So the two great American architectural communes of the twentieth century shared a labour system and disagreed about everything else, and both produced remarkable buildings, extensive influence, and a body of former students carrying the ideas outward, which turned out to be the durable output in both cases.
Soleri concluded that this was exactly backwards, and spent the rest of his life arguing the opposite case with unusual consistency. Where Wright dispersed, Soleri concentrated. Where Wright celebrated the car as the instrument of freedom, Soleri identified it as the mechanism by which cities were being destroyed and land consumed. In 1956 he established his own compound, Cosanti, in Paradise Valley, and named it from Italian words meaning against things, a deliberate jab at the consumer culture he thought was driving the sprawl. It is worth registering that Broadacre City substantially won: the postwar American landscape of subdivisions, arterial roads, and shopping centres is much closer to Wright’s proposal than to anything Soleri drew, which is a useful reminder that the utopian visions which get built are rarely the ones with the better argument, and that unplanned settlement patterns tend to defeat planned ones, as the improvised community at Christiania in Copenhagen demonstrates from the other direction. Two students of the same problem. Opposite answers, and the wrong one scaled.
The Windbells
How do you fund a city with no investors, no government backing, and no revenue? Soleri’s answer was bells, and it is genuinely one of the more remarkable financial arrangements in the history of architecture. At Cosanti he had developed a technique for casting bronze and ceramic windbells, each one distinctive, produced on site in a foundry and a ceramics apse. They sold, and kept selling, and they have been the primary revenue source for the entire enterprise for over half a century.
The arrangement also shaped the architecture in ways that are easy to miss. The foundry and the ceramics studio are not incidental workshops tucked away behind the residential buildings; they are among the largest and most carefully designed structures on the site, because they are the revenue. A visitor walking through Arcosanti is walking through a settlement organised around its own funding mechanism, with the productive apses given the best solar orientation and the most generous space. That is not a criticism, and in fact it is exactly what an honest small economy looks like. It is simply worth noticing that the plan on paper had industry serving the residents, while the site as built has residents serving the industry.
This deserves admiration before it gets analysis. Nobody handed Soleri a budget. He built a functioning craft business and used its margins to pour concrete, which is a far more honest arrangement than most utopian projects manage, and it made the project genuinely independent of patrons who might have demanded compromises. The bells are also good, which matters; they have been collected for decades and still sell. Casting bronze at that quality involves a real command of metallurgy and finishing of the kind that underlies the manufacture of precision metal components and the wider industrial chemistry of the metals a modern economy runs on. The funding model worked exactly as designed. That turns out to be the problem, and we will come back to it.
Who Built It
The labour arrangement was equally unusual and equally consequential. Soleri, following a practice he had encountered at Taliesin, invited anyone over eighteen to come and build his city, and to pay for the privilege. The five-week workshop combined instruction in construction technique and arcological philosophy with actual work on the site, and it has run continuously since 1970. Something upward of seven thousand people have passed through it, and they are the ones who physically built almost everything standing at Arcosanti.
The workshop model produced something else of real value, and it is arguably the project’s largest output. Something on the order of seven or eight thousand people spent five weeks each thinking hard about density, land use, energy, and how settlements ought to be arranged, then went home. A substantial number became architects, planners, and builders. If you wanted to design a mechanism for propagating a set of urban ideas through a profession over five decades, it would be difficult to improve on a beautiful desert site, a charismatic founder, cheap labour that pays to attend, and a five-week immersion. The city was never finished. The seminar ran for fifty-five years.
The construction method was chosen to suit that workforce. Soleri used silt-casting: concrete panels poured into moulds carved directly into beds of local desert silt, then tilted up into place, which produces the distinctive textured surfaces that make the buildings look excavated rather than assembled. It is a technique that a motivated amateur can learn in days and that requires minimal equipment, which is precisely why it was viable for a rotating population of students. But a construction method selected for a transient volunteer workforce also determines what can be built, and it is well suited to producing striking individual structures and very poorly suited to producing the systematic, code-compliant, serviced, plumbed, wired, elevator-equipped, fire-rated apartment stock that a city of five thousand would require. The self-built approach is a genuine and undervalued tradition, and it belongs in the catalogue of ambitious technical undertakings alongside the more capital-intensive dreams such as collecting solar power from orbit. The method, in the end, built precisely what that method was capable of building.
What Actually Got Built
The physical result, after fifty-five years, is a dozen or so structures on fewer than twenty-five acres. There is the Ceramics Apse and the Foundry Apse, quarter-sphere shells oriented so the low winter sun reaches deep inside and the high summer sun does not, which is passive solar design working exactly as intended and still working today. There are the Vaults, an open concrete gathering space. There is a five-storey visitors’ centre with a cafe, an amphitheatre, a swimming pool, an office building, a ring of apartments and small storefronts, greenhouses and gardens. The last major structure went up in the late 1980s.
The buildings that exist do work, and it is worth being specific about that, because the failure is not one of competence. The apses genuinely perform: they admit the low winter sun and exclude the high summer sun, in a climate where that difference is the difference between comfort and misery, and they do it with geometry alone and no moving parts or energy input. Visitors in July notice immediately. The passive solar performance of Arcosanti is a real, measurable, replicable result, and it has influenced a great deal of subsequent desert architecture. Soleri was a serious builder as well as a serious theorist.
Estimates of completeness vary depending on what you count, running from about five percent to around ten. The scholarly reference maintained by the Society of Architectural Historians describes Arcosanti as roughly ten percent complete, sitting on a fraction of the land the foundation controls, with the rest of the preserve left deliberately undeveloped as the arcology principle requires. What exists is architecturally serious and, by common consent, rather beautiful. It is also, and this is the crux, not a small city. It is a set of remarkable buildings with some people living among them, which is a completely different category of thing, in the way that a promising laboratory result differs from a deployable technology across every field from exotic materials research to the industrial supply chains behind modern manufacturing. Ten percent of an arcology, it turns out, is not ten percent of a city at all.
Density Has No Small Version
Here is the mechanism, and it is worth stating carefully because it applies well beyond Arizona. Soleri’s case for arcology rests entirely on effects that are functions of population size, and every one of them has a threshold below which it simply does not occur. Consider each in turn. Car-free living requires that everything a resident needs be within walking distance, which requires that those things exist, which requires enough customers to sustain a grocer, a clinic, a school, a hardware supplier, an employer. Below a few thousand people none of those are viable, so residents must drive to reach them. The urban effect requires a population large enough that unplanned encounters produce novelty, which is a function of how many people you do not already know; in a community of eighty, you know everybody by the second week and the collisions stop generating anything. Efficient shared infrastructure requires enough users to amortise it.
There is an instructive comparison available in the same decade. The new towns and planned communities that succeeded in the twentieth century almost invariably began with the population, not the buildings: a government or a corporation moved thousands of people in on a schedule, often because there was employment waiting, and the settlement reached functional size within a few years whether or not the architecture was any good. Arcosanti reversed that order, building first and hoping the population would follow. Under a threshold model this is fatal, because nobody wants to be resident number two hundred in a place whose amenities require five thousand, and the early adopters therefore have to accept the worst version of the deal indefinitely.
None of these degrade gracefully. They are not present at ten percent strength in a settlement at ten percent of target population; they are absent, entirely, and then they appear rather suddenly once the number is high enough. This is what makes arcology fundamentally different from most technologies, which can be prototyped small. A scale model of a solar panel generates proportionally less electricity and demonstrates the principle perfectly. A scale model of a city generates no urbanism at all, because urbanism is not a property of buildings but an emergent behaviour of populations, arising from interaction rates in the way that collective behaviour emerges in the neurological and social systems that produce coordinated action or in the collective navigation of animals that only achieve certain feats in numbers. Soleri built a prototype of something that has no prototype. The idea could only ever have been tested at full size.
The Bootstrap Trap
Which produces a circularity that no amount of commitment could break. To reach five thousand residents, Arcosanti needed capital on the scale of a real property development: hundreds of millions of dollars in today’s terms, for infrastructure, services, and buildable housing stock. To attract capital on that scale, it needed to demonstrate that the arcology concept worked. To demonstrate that the concept worked, it needed five thousand residents. There is no entry point to that loop, and no amount of effort at any stage moves you closer to breaking it.
The obvious alternative was to raise development capital in the ordinary way, and it is worth asking why that never happened. Part of the answer is that Soleri did not want it; outside investment would have brought outside control and, almost certainly, demands for conventional buildings on conventional schedules. Part of it is that the proposition was genuinely unfinanceable on commercial terms, since no lender underwrites a twenty-five-storey structure in unincorporated high desert seventy miles from a labour market, for a business model with no comparable transactions anywhere. The independence was chosen, but the alternative was also largely unavailable.
The funding model made this permanent. Selling handmade bronze bells is an excellent way to finance an artist’s studio and a genuinely good way to fund incremental construction, but a craft business has a hard ceiling determined by how many bells people will buy, and that ceiling sits several orders of magnitude below city-building finance. Add tens of thousands of visitors a year and workshop fees and you have a viable small nonprofit, not a developer. So the revenue mechanism that made Arcosanti independent also guaranteed it would remain permanently below the threshold at which its own thesis could be demonstrated, and every year of admirable self-sufficiency was another year of not getting there. Charter city projects and new-town ventures have run into versions of the same wall ever since, as the record of attempts to build new jurisdictions from scratch shows, and the gap between craft-scale revenue and industrial-scale capital is the same gap that separates a trader from an operator in the history of commodity finance. The bells kept the project alive for half a century. The very same bells also guaranteed it would stay small.
Everybody Drives
The consequence on the ground has a particular sting to it. Arcosanti exists to prove that a city can be built without automobiles. Arcosanti sits seventy miles from Phoenix, two miles off an interstate, in high desert with no public transport, no shops within walking distance, no medical facility, and no employer other than itself. Its residents own cars. They have to. Every grocery run, medical appointment, school journey, and social visit outside the community requires driving, on an interstate, exactly the infrastructure the project was founded to make unnecessary.
It is fair to note the counter-argument that autonomous vehicles and delivery robots might eventually make remoteness less costly, dissolving the walkability requirement by removing the driving from the driver, and the proliferation of autonomous vehicles and delivery systems has made that argument more respectable than it was a decade ago. Soleri would have rejected it flatly, and on his own terms he would have been right to: the objection to cars was never merely that a human has to steer them, but that they consume land, energy, and materials, and that the settlement pattern they permit destroys the proximity from which the urban effect arises. An automated car still requires a road.
This is not hypocrisy on anyone’s part, and the residents are perfectly aware of it. It is the threshold problem expressed as daily life: at target population the site would contain everything and the cars would be genuinely unnecessary, while at actual population it contains almost nothing and the cars are mandatory. The site was chosen partly because land was cheap there, which is the correct decision for a project that has to build slowly with limited funds and the wrong decision for a project whose thesis is proximity, and the two considerations were in conflict from the first day. The result is a community whose per-person vehicle mileage is plausibly higher than that of the suburbs it was designed to indict, which is a difficult thing to sit with, and which the transition to electric vehicles and the supply chains behind modern battery materials mitigates without addressing, since the underlying problem is distance rather than fuel. The prototype of the car-free city runs, necessarily, on cars, and at this population it could not possibly have been otherwise.
Arcosanti in 2026
Soleri himself died in 2013, aged ninety-three, having spent more than four decades on a site that never reached two percent of its intended population. He continued to draw arcologies into his final years, and continued to insist that the argument was correct and the timescale simply longer than a human life, which is a defensible position and also an unfalsifiable one. Whether he understood the threshold problem in the terms set out here is unclear from his writing; he certainly understood that the project needed scale, and equally certainly never found a route to it.
The place is still going, which after fifty-five years is itself an achievement. The Cosanti Foundation still runs the site, still casts and sells the bells, still hosts workshops, and still receives tens of thousands of visitors a year, many of them architecture students who come to look at the apses and leave thinking hard. Construction has been minimal for decades, and the current institutional posture is closer to stewardship of a significant site than to building a city, with attention going to maintaining what exists, running educational programmes, and hosting events on the mesa.
What the site has become in the meantime is genuinely interesting on its own terms, if not on Soleri’s. For a stretch of years it hosted a substantial music and arts festival that brought thousands of people onto the mesa for a weekend, which is the only occasion on which Arcosanti has ever operated at anything approaching its design population, and by many accounts the place came alive in exactly the way the drawings promised. That is a suggestive data point rather than a vindication, since a festival crowd is not a resident population and leaves on Sunday, but it is the closest the experiment has come to running.
The legacy also carries something much darker, and it should be stated plainly rather than skipped. In November 2017 Soleri’s daughter Daniela published an account stating that her father had sexually abused her, including an attempted rape when she was seventeen. The Cosanti Foundation’s board issued a public statement supporting her and saying the disclosure compelled them to reconsider his legacy, and the organisation has maintained a published statement and policy on the matter since. Institutions across Arizona that had honoured Soleri were left with the question of what to do about tributes bearing his name, and the reckoning is not finished. It is possible to hold both things at once: that the arcology argument is worth taking seriously on its merits, and that the man who made it did serious harm, which is a problem that recurs whenever an idea is bound tightly to a single figure, as it does across the histories of institutions built around individuals. The buildings are still there. So is everything else.
The Prototype of a Threshold
Strip Arcosanti to its structure and the failure is not moral, financial, or organisational, though all three played a part. It is arithmetic. Soleri identified a real problem, diagnosed it accurately, and proposed a solution whose every claimed benefit was a threshold effect, then attempted to validate it with a demonstration project that could never reach the threshold, funded by a craft business with a ceiling far below the required capital, built by a volunteer workforce whose available techniques suited sculpture better than housing stock, on a site chosen for cheap land in a place where nothing else was within walking distance. Each of those decisions was locally sensible. Together they guaranteed that the experiment could not run.
The test this suggests is worth applying to any proposed community before ground is broken. Ask which of the promised benefits are threshold effects and which scale smoothly, because the first category will be entirely absent until the population arrives and the second will not. Ask what the minimum viable population actually is, in people rather than aspiration. Ask where the capital to reach that number is coming from, and whether the revenue mechanism has a ceiling below it. And ask what life is like for the first two hundred residents, because if the answer is unattractive, there will never be a two hundred and first.
And yet the argument itself substantially won, which is the strange and consoling ending. The critique of sprawl, of car dependence, of land consumption, of the energy cost of low density, is now close to consensus in urban planning, and the remedies that actually got adopted, transit-oriented development, infill, mixed use, walkable neighbourhoods, the fifteen-minute city, are Soleri’s diagnosis implemented incrementally inside existing cities rather than as a single organism on a mesa. That is the usual fate of the projects gathered in the catalogue of utopian societies: the community does not survive, and the ideas are quietly harvested by the world it was built to reproach. Arcosanti never became a city. It became an argument that cities eventually took, in pieces, without ever visiting.