Gropyus's insolvency filings confirm the bet we made on AUAR: in housebuilding, small factories on the builder's site will beat gigafactories.
Gropyus is the flagship of German serial timber construction, with around €400m of equity behind it. On 3 November, it filed for insolvency for several of its subsidiaries.
Headlines will lump every factory-built housing company together. But the filings say more about how Gropyus was financed (it paid for its own plant and carried its own projects). And how capital and risk are split is the heart of our AUAR thesis.
Three arguments support our view:
- Gigafactories put fixed cost and project risk on one balance sheet. AUAR spreads the fixed cost across small, mobile factories and leaves project risk with the builder.
- The market can absorb hundreds of small factories, and AUAR built V3 on 2 years of field data to serve it.
- Competitors are converging on smaller factories, and AUAR runs the smallest capital unit among them.
Why decentralised factories will beat gigafactories
The past decade of industrialised housebuilding is full of examples:
- Katerra raised around $2bn, mostly from SoftBank, and filed for Chapter 11 in June 2021. Its Spokane CLT plant cost about $150m and sold out of bankruptcy for $50m (a third of what it cost to build).
- Veev raised more than $600m at a $1bn valuation in 2022 and shut down at the end of 2023.
- ilke Homes entered administration in June 2023 owing more than £319m, including £68m to Homes England, which expected to recover a fraction.
- L&G Modular stopped production in Leeds in 2023 after losing about £235m since 2016.
- TopHat lost money every year it traded and stopped in November 2024, 8 months after scrapping its plan to take Corby from 800 to 4,800 homes a year.
- Harbinger (formerly Factory_OS) warned in February 2026 that it may close its Vallejo plant.
Gropyus is the latest addition. The trigger was an investor payment that was contractually due and never arrived.
But the numbers go further back. In the first half of 2026, Gropyus lost €51.4m on €29.6m of revenue (about €1.70 lost for every euro billed). On 30 June it held €11.2m of cash against €101.3m of short-term debt, according to the half-year report of Vonovia, which owns a quarter of the company.
Everyone knows the fixed-cost and transport arguments. What gets less attention is how these companies kept their lines busy: they turned into developers. They took on land, planning and sales risk to feed the factory. L&G's leadership blamed UK planning delays. Katerra tried to own design, materials, manufacturing and contracting at once. Gropyus runs its plant in Richen and also delivers the buildings, around 1,150 apartments across 11 projects. It even cites Berlin's debate on socialising large housing portfolios as a strain on a market that matters to its business. That's a political risk no factory should be carrying. So in each case, one expensive plant carried both the fixed cost and the demand risk.
Gropyus is the closest German comparison to AUAR: timber elements, made by robots, assembled on site. AUAR splits the risk the other way round. Its factory sits on the builder's site and moves to the next one when the site finishes or stalls. The builder keeps land, planning, sales and the timber, and pays AUAR per square metre of panel produced. The builder no longer has to invest in a factory. AUAR's job is to keep its factories busy, and that's a risk worth holding.
One MicroFactory costs a fraction of Gropyus's plant and can generate around £1m of revenue a year without running at full capacity. So it pays for itself well within its first year (its build cost is a small slice of that revenue). When a builder slows down, the factory moves to another one within weeks.
TopHat and ilke needed years at full volume to break even.
The market can absorb hundreds of small factories
The US alone makes the case. Builders started 943,000 single-family homes there in 2025, and 94% have a timber frame (the exact product a MicroFactory makes).
One factory is designed to frame about 200 homes a year. Framing 1 US home in 10 would take about 470 factories, or close to £500m of revenue a year. That's one country and one product. And because AUAR charges per square meter, that figure tracks how much the US builds. Factory speed only changes how many factories it takes to get there.
V3 is built on 2 years of field data
This part we share only with you. AUAR spent most of 2025 turning its V1 and V2 deployments in Europe and the US into a third-generation factory. Three lessons from the field shaped it.
- Make it work on any site, with any timber. V3 rolls out of its container on a chassis and needs no foundation. Its vision and handling systems cope with American timber (less straight and consistent than European stock). In V2, the arm had to push every panel through the saw, so the arm was always the bottleneck. In V3, the saw runs on its own actuators, and design capacity goes from about 120 to about 200 homes a year. It makes floors and roofs too, where V2 only made walls.
- Ship the people with the machine. Builders get the most out of a factory with experts beside it. So a field operations team, modelled on Palantir's forward-deployed engineers, now goes out with each factory.
- Fit the price to how builders budget. Builders allocate spending project by project. The upfront payment now buys production credits they draw down per project, so the cost lands in the right budget.
Industrial robotics usually takes 3 to 5 years to get from V1 to V3. AUAR did it in 24 months, with a rebuild in the middle. The team changes the product when the field data tells them to, and they came out of 2025 focused on one customer: housebuilders building 150 or more homes a year.
Competitors are converging on smaller factories
Weinmann, Hundegger, Randek and House of Design will keep selling lines to component plants. Their customers still carry a fixed plant, so we see them as suppliers or acquirers before rivals.
The real competition comes from newer players:
- BotBuilt (North Carolina) frames wall panels with second-hand automotive arms. It raised $12.4m in late 2023.
- Promise Robotics (Alberta) had raised about C$34m by its 2023 Series A.
- Reframe Systems (Massachusetts) raised $40m in August 2026. Its first microfactory runs on under $5m of equipment for up to 250 single-family homes a year, and Reframe delivers the homes itself.
- Cuby is closest to AUAR in concept: containerised factories near sites, sold to local operators, at a launch cost below $10m for about 200 homes a year on roughly 3 acres.
Reframe and Cuby produce more of the finished house, so the numbers don't compare one to one. AUAR still runs the smallest capital unit in the field by an order of magnitude, and it leaves homebuilding to homebuilders.
Its edge sits in what a competitor can't buy off the shelf:
- In-house tooling. AUAR designs its own end-of-arm tools for nailing, cutting and assembling natural timber, the hardest part of making industrial robots work with wood.
- The cassette standard. AUAR holds a patent on its cassettes and connectors. Once a builder's design library and structural approvals reference them, switching means re-engineering and re-approving every house type.
- The software lead. MasterBuilder already handles the inputs builders actually send, down to paper plans. And every deployment feeds production data into the next release.
What comes next
The next milestone is 10 V3 factories on builders' sites, running at target output and paid for by the metre.
We've seen this play out with ecoworks: once a new construction method proves its margins and delivery, the industry follows. For AUAR, 10 running factories are that proof.
They give lenders the operating data they need to finance a fleet. And they turn US builder interest into signed volume (AUAR's US team is led by a former production homebuilder COO), which puts distance between AUAR and well-funded US rivals.
That's why AUAR is raising its Series A now: to take the fleet from 10 factories to 100.