Eleven modular projects, sorted by what they actually taught.

Case studies in this industry are usually chosen to flatter. We chose these eleven because they are well documented — by auditors, courts, bankruptcy filings, government releases and trade press — and because together they show a pattern that no single project does. Where a figure comes from a party with an interest, we say so.

The pattern

Sort the eleven by outcome and a split appears. The reliable successes are public or institutional programmes buying a narrow, repeatable product at scale: British Columbia's temporary modular housing, Hong Kong's Modular Integrated Construction, Singapore's PPVC towers, Croydon's build-to-rent. The failures are of two kinds: heavily capitalised startups that tried to build factory, software and supply chain at once, and "world's tallest" showcase projects whose ambition defeated the repetition that makes volumetric economical. The technology was rarely the problem. Financing, standardisation and procurement discipline almost always were.

Capital: Katerra, Veev, Factory_OS, Boxabl

Katerra — the cautionary tale

Katerra raised more than $2 billion, most of it from SoftBank's Vision Fund including an $835 million round in 2018 and a rescue in December 2020, grew to about 7,500 employees and some 700 projects, and filed for Chapter 11 in June 2021 without paying severance.¹ The proximate cause was the collapse of its lender, Greensill Capital, compounded by the pandemic. The structural cause, in the assessment of Volumetric Building Companies' chief executive, was that Katerra tried to vertically integrate an entire global supply chain simultaneously — factories, software, materials, design, general contracting — before proving unit economics in any one of them.¹ VBC subsequently bought Katerra's Tracy, California factory for $21.25 million, added a Pennsylvania plant, and has grown by keeping each business unit independently profitable.² The contrast is the lesson.

Veev — capital-cycle exposure

Veev raised roughly $600 million, reached a valuation above $1 billion, and shut down in November 2023, liquidating its US assets after new funding failed to materialise.³ Unlike Katerra it did not obviously mismanage; the capital markets for construction technology closed. The lesson is uncomfortable: a vertically integrated modular builder carries a factory's fixed costs on a startup's balance sheet, and survives only as long as someone will fund the gap between capacity and orders.

Factory_OS / Harbinger Homes — volume is not enough

Factory_OS, founded in 2017 in Vallejo, California, with Google as first customer and backing from Autodesk, Citi, JPMorgan and Saint-Gobain, built around 4,000 units, 90 percent of them affordable housing. It was acquired by private equity in 2024, renamed Harbinger Homes, and was reported in December 2025 to be ramping production; by February 2026 it had announced a potential closure and 280 layoffs, citing lost funding and cheaper competition from out-of-state and international factories.⁵ Real output, blue-chip customers and a social mission did not overcome a high-cost location in a thin-margin market.

Boxabl — throughput versus attention

Boxabl is the most publicised modular company in North America and one of the smallest by output: 2024 revenue of $3.4 million against a reported loss of $50.9 million, roughly 600 units shipped, and more than 160 employees.¹² The figures come from a critical crowdfunding analyst and should be read as such, but they are drawn from the company's own filings. Attention is not a production metric.

Showcase: 461 Dean, AC NoMad, Croydon

461 Dean Street — geometry defeats repetition

Forest City Ratner's B2 tower at Pacific Park, Brooklyn, was announced as the world's tallest modular building and opened in 2016 with 363 units at roughly twice the projected schedule.⁴ The wedge-shaped site meant that 419 bathroom pods required 179 different designs; façade panels were reported flapping in the wind; modules arrived water-damaged or incomplete. Forest City and its contractor Skanska sued each other for years and settled confidentially in April 2020.⁴ Volumetric construction's economics rest on repetition, and this building had almost none.

AC Hotel NoMad — the announcement that never became a building

Marriott announced the world's tallest modular hotel at 842 Sixth Avenue in 2018: 26 storeys, 168 steel modules, stacking a floor a day. Work stopped in 2020, the site changed hands, and in September 2024 the modular hotel was cancelled and replaced with a conventional condominium.¹⁰ Several trade outlets still describe the project in the present tense. Treat any "tallest" or "first" modular claim as unproven until the building is occupied.

101 George Street — what a showcase looks like when it works

Croydon's two towers — 44 and 38 storeys, 1,526 steel modules, 546 build-to-rent apartments — were delivered by Tide Construction with modules from its sister company Vision Modular Systems in roughly half the time of conventional construction, with power on by March 2020.¹¹ The difference from 461 Dean and NoMad is structural: developer, contractor and manufacturer were one integrated group with a pipeline of similar towers, building a repeatable rental product for institutional owners. Ambition was matched by repetition.

Public programmes: Toronto, BC, Singapore, Hong Kong

Toronto's Modular Housing Initiative — speed without controls

Toronto's Auditor General found that the city's programme completed 159 of 275 planned supportive units at three of five sites, exceeded its original $47.5 million budget by $33 million — 63 percent — and took five to thirteen months from construction start to occupancy. The causes were incomplete site due diligence before budgets were set, fragmented cost tracking and change orders approved without formal review.⁶ None of that happened in the factory. The modular method delivered what it promised; the procurement around it did not.

British Columbia — narrow scope, dedicated capital

Between 2017 and March 2019 the province and City of Vancouver delivered 606 supportive homes across ten sites — Nora Hendrix Place, for example, 52 units for $8.8 million — under a $291 million provincial programme targeting 2,000 homes.⁷ The scope was deliberately narrow: single-occupancy studios on city-owned land with a single provincial funder and a repeatable site type. That is the recipe, and it is the recipe Toronto's programme lacked.

The Clement Canopy — regulation creates specialists

Two 40-storey towers in Singapore, 505 apartments, with 65 percent of the superstructure delivered as prefabricated prefinished volumetric construction by Dragages.⁸ Singapore's Building and Construction Authority requires PPVC on designated government land sales sites, so contractors have built repeat experience that a one-off developer never accumulates. Regulation created the specialists; the specialists made high-rise volumetric routine.

Hong Kong MiC — the captive client

Hong Kong's government targeted Modular Integrated Construction for roughly 20,000 public housing flats, with the Housing Authority's first dedicated MiC project at mock-up stage by March 2023.⁹ At that scale the Housing Authority is exactly the long-horizon repeat buyer that de-risks factory investment — the structural advantage that no private North American developer can offer a manufacturer, and that Build Canada Homes is now attempting to replicate.

Six lessons

Standardise before you scale. Croydon, Singapore and Hong Kong repeated a product; 461 Dean did not. Prove one unit's economics before integrating the next. VBC survived by keeping units independently profitable; Katerra did the opposite. The factory is only half the schedule. Toronto's delays were on site, in due diligence and contracts. Capital cycles kill factories. Veev and Harbinger had output; they ran out of runway. Location is a cost of goods. A Vallejo factory competing with Idaho and overseas plants is fighting its own address. Announcements are not buildings. AC NoMad was "the world's tallest modular hotel" for six years and is now a condo.

What to do with this

If you are a developer, choose repeatable programmes over statement buildings, and audit the site before you sign with the factory. If you are a manufacturer, resist the customer who wants your factory to do something new; the next order from an existing client is worth more than the flagship. If you are an investor, price factory fixed costs against contracted backlog rather than pipeline, and ask what happens to the business in a two-year capital drought. And if you are a government, understand that the single most valuable thing you can offer this industry is not a subsidy but a decade of predictable orders for the same thing.