How to Build Homes 10x Cheaper

You're looking at half a million dollars to build a new house in Australia right now. But the physical stuff going into it, the raw iron ore, limestone, sand, and clay pulled out of the dirt, costs closer to three thousand. Everything we hate about the housing market sits right in the middle of those two numbers.

A home costing $500,000 to build today can actually be done for $69,000. It doesn't need some magic breakthrough in nanotechnology either. We're using machines that are being built today and chemistry that's been in textbooks for a hundred years, backed up by maths you can check yourself. The car industry figured this out a long time ago. If cars had developed like houses since the fifties, buying a new Corolla would take nine months and cost you $400,000. When it finally turned up it would probably leak, and we'd just call it character. In housing, we just accepted it and called it the property market.

Costs get smeared completely thin across about thirty different trades, not to mention transport, paperwork, and materials. That's why one clever trick has never solved the problem. Factory-built housing companies always attacked the labour side of things, and then died because they got the other math wrong. They'd get killed by the cost of transport, sit empty during slow years, die waiting for council approvals, or get crushed as material prices went up. Terraform hits all four at the exact same time. We make the materials in our own solar-industrial plants using nothing but rock and sunlight. We ship the homes flat-packed in containers, then the factory-finished components bolt together on site in eight days. Thousands of homes share one single certification. By halving the labour, halving the materials, dropping the overheads and dumping the parts nobody needs, the whole thing lands at a tenth of the price. The rest of this is how we do it.

I am going to pull apart a half million dollar invoice for a standard 200 square metre house, because the industry never does this where people can see it. We are only looking at the cost to actually build the thing. The land isn't in these numbers. Everything is rounded, and unless I flag it, the money is Australian.

You stare at the breakdown and realise quickly that there is no giant monster to slay here. The absolute biggest trade line takes up just 9% of the cost. The money is scattered in tiny piles across thirty different trades. This scattergun spread is why every genius invention aimed at housing, from the printed wall to the robotic bricklayer, ends up changing nothing about the final price. They kill off a single line item and go home. Crack open that $398,000 block of trade lines. You find roughly $180,000 of labour sitting next to $218,000 of materials. A straight split. Every factory housing startup in history attacked the labour side of that equation. In a hundred years, nobody seriously took on the materials side. Hold onto that thought. The materials side is where three grand worth of dirt and metal goes into the top of a machine and a $218,000 bill falls out the bottom. The rest of this invoice isn't even for construction. Around $102,000 pays for margin, finance, design, and navigating council approvals. It is pure margin piled on top of the cost of risk and waiting around. The builder's margin line sits at $70,000 simply because throwing a house together outdoors takes nine months. Somebody has to price in nine months of rain, screw-ups, variations, and the chance that a contractor goes bust. You are literally paying $70,000 for the risk of rain, and it gets its own line on the bill.

People with very deep pockets have spent a century trying to industrialise this mess, and the graveyard they left behind is incredible to look at. General Panel built a massive factory meant to spit out ten thousand homes a year and died after building about two hundred.¹ Lustron got 37 million dollars from the US government back in 1948, which was an insane amount of money at the time. They actually built 2,500 really good steel houses in two years before going bankrupt anyway.² Stirling Homex ended up pretending that unsold modules rotting in paddocks were revenue, and some of their executives went to prison.³ Toyota is arguably the most capable manufacturer on the planet. They waded into housing in 1975 with bottomless patience and a massive home market. Half a century later they hold a polite niche and have caused absolutely zero cost revolution.⁴ Katerra burned through over two billion US dollars in roughly six years.⁵ Veev set fire to six hundred million US dollars and collapsed twenty months after they raised US$400 million at a billion-dollar valuation.⁶ A whole British modular building sector popped up, burned bright, and completely died within ten years. Not one of them failed because they couldn't weld straight. They failed because they refused to respect a few brutal realities of physics and maths.

Shipping physics hits you first. A shipping container packed to the roof with smartphones is worth ten million dollars. A shipping container carrying a finished module of a house is mostly just air, worth maybe thirty thousand dollars. But the freight company charges you by volume, not value. Getting massive finished modules down a highway means paying $12,000 to $25,000 a house by the time you pay for oversize permits, escort utes, and crane hire (those little warning flags cost extra). All of that shipping expense immediately eats whatever 15% margin you managed to claw back inside the factory doors if you dare try to sell a house more than a few hundred kilometres away. So your catchment area physically shrinks. Your factory gets capped at a small size and you never actually manage to hit real scale. Trying to ship empty air on flatbed trucks while paying road train rates is exactly what trapped every modular builder that ever existed.

Then the factory itself starts bleeding you. A factory is nothing more than a giant fixed cost divided by whatever you manage to push out the door. If you look at Britain, they ran an incredibly clean, tragic test case for this. Legal & General threw down a massive factory that was supposed to churn out 3,500 modular homes every single year. They kept the doors open for about eight years but never managed to get anywhere close to filling the floor, and they finally shut the whole operation down in 2023 wearing £359 million in accumulated losses.⁷ Meanwhile Urban Splash's modular division and Caledonian Modular died right next door to them. When a factory runs at half speed, every house it manages to sell has to carry the cost of the empty space left behind on the factory floor. Housing demand swings wildly, crashing by 40 to 60 percent across market cycles. A factory without a guaranteed floor of demand is just a ticking clock waiting for a bad year to wipe it off the map.

The absolute most infuriating waste begins before anyone even picks up a hammer. That $20,000 line for design, engineering, and council approvals gets spent all over again on every single house. Because local councils cling to their own discretion, every single home gets assessed as if nobody has ever built anything like it before. They treat every house like a forever prototype. The fix for this is obvious, because the one massive exception over the last century proves it. American manufactured homes are built to one short national code, and they consistently cost about half as much per square metre as building on site.⁸ The factory was never the secret weapon there. The secret was spreading the fixed costs of design and approval over a massive number of homes.

But absolutely none of that even touches the real problem, which is materials. That $218,000 chunk for materials is bought at the end of an absurdly long, fat chain. A mine digs it, a mill processes it, a wholesaler stores it, a freight network moves it, a retailer sells it. Each one of those guys takes their margin. And every time the price of diesel jumps or a wage deal goes through anywhere in that supply chain, the extra cost just gets passed downstream until it hits the final buyer. The raw cost of materials in Australia jumped by a third over a five year period and it is absolutely never going to drop back down to where it was.¹⁰ People selling commodities do not hand out refunds. If you rely on finished materials, your input costs are totally set by strangers who only know how to jack the price up. The only way you can actually escape this trap is by looking at what makes the materials expensive in the first place. The most expensive ingredient in any building material is the massive amount of raw heat required to manufacture it. Today, we can finally generate that heat using solar power, and solar happens to be the one industrial input in human history that actually gets cheaper every single year.

To get the prize you actually have to fix all four things at once. If you only manage to fix three of the calculations, you don't succeed. You just fail a lot slower, which is an incredibly expensive way to go bankrupt. Saying you will cut the cost to one tenth sounds like empty talk until you lay out the multiplication involved. You need to take the labour half and halve it. You need to halve the materials half. You take 40% straight out of the transport, the overheads, and the council approvals. Then you delete roughly 30% of the subsystems that usually cause trouble, meaning things like plaster, paint, and ducts are just gone. When you do the maths on that, 0.5 multiplied by 0.5, then by 0.6, and finally by 0.7 gives you 0.105. Someone has managed to achieve every single one of those factors independently. However, they absolutely refuse to add up unless a single group owns the entire process. If you sell a very cheap panel to a conventional builder, it instantly turns into the builder's profit margin. Building a fast factory that has to wait for a slow approvals process just means you own a warehouse with a lot of ambition. You have to own the whole equation for it to multiply.

Labour is the best place to start. Terraform should not be thought of as a house design, nor is it a chassis. What it actually is, is a library containing prefabricated steel sections and components. We have wall sections. We have roof sections. We have floor cassettes, bathroom pods, and kitchen modules. We have wiring harnesses and a plumbing core. All of these are engineered so they come together quickly on the site, allowing you to build everything from a small studio to a massive family home. The principle we use to govern this is quite ruthless. If there is a clever idea, it has to live either at the design desk or somewhere on the factory line. None of it is allowed to live on the site. The factory handles the thinking, and the site just handles the bolting. Because components combine exactly the way words combine into sentences, the buyer gets an enormous amount of variety. The factory gets repetition because it is only ever manufacturing the words. We have two rigid constraints that set our grid dimensions. The bathroom pod alongside the plumbing core forms the first constraint because they are incredibly complicated objects to manufacture. What travels on the roads dictates the second constraint, meaning we are bound by the internal width of a shipping container. Composition above that grid is completely free. A family home wrapped around a courtyard uses the exact same parts as a studio behind an existing house, or a long thin home sitting on a battleaxe block. They are all different in plan but identical in parts. The factory never sees the house, it only sees the components, so offering variety costs us absolutely nothing.

This does completely change the invoice. A concrete slab is usually $45,000 worth of earthworks, formwork, pouring, and then a 28 day cure where you spend a month checking the weather app. That becomes $12,000 of screw piles and a steel subframe. One machine installs it in a single day, load-verifying each pile by its installation torque, and the entire frame is laser-levelled to two millimetres. The Earth has a lot of opinions about what constitutes 'level', and threaded pile heads absorb all of them in about an hour because the house never touches the ground. Because the rollformer cuts and punches and labels every piece of steel, the 2,000 hours humans normally spend measuring in the weather suddenly becomes 200. Tape measures vanish. Workers construct walls flat at bench height where the whole operation remains completely dry. Trying to squeeze three licensed trades into the footprint of a standard bathroom creates a three-week fight in a phone booth, so we bypassed that entirely by building ours as an independent pod on its own cell that passes a wet test long before seeing a delivery truck. That turf war is finished before delivery day arrives, allowing the whole unit to drop into place with a single crane lift and just two final connections. We mounted the heavy external services right onto the outside face of that pod to carry the meter box alongside the hot water unit, heat pump, and battery bay. Because those components leave the line fully fitted and tested, swinging the bathroom into position installs your entire services wall in the very same motion. The installation timeline kicks off with screw piles going into the dirt on Monday morning, allowing the main sections to be assembled across the next few days. We are usually inside running our final commissioning checks around the eight-day mark. There is no glue anywhere in the thing, which was deliberate, and the knock-on effect is a bit odd once you notice it: every join is a bolt or a clip or a coupling or a lap that reseals, so the whole build runs backwards just as happily as it runs forwards, and a crew that knows what they are doing can have a standing home broken back down into container-ready sections inside about a day. Movable dwelling stops being the regulatory loophole everyone quietly exploits and turns into something you can literally do, which means the house is an asset you pick up and take with you when you leave the land rather than something you walk away from.

Wiring and plumbing are already installed when every section leaves the factory. The digital model has the routes designed into it, and the same machines that form the panels cut those routes. They are fitted at bench height on the line. The electrician, plumber, and insulation installer usually take turns wrecking each other's work inside a normal wall cavity. Production stations that never conflict replace them entirely. The digital model resolves any crossing before the first panel is even created. Every circuit and pipe is tested before despatch, and each run ends at a keyed connector or push-fit manifold on the edge of the section. When you join the sections on site, you join the services. If you want to add a power point in ten years, you don't open a wall. You just unclip a strip, because the joint cover strips are shaped as raceways. The outside face arrives finished too, with sections fully clad or carrying a modular cladding system over a drained cavity. The weather line is made and tested indoors, so the facade is a catalogue selection rather than a messy construction process. Things like scaffolding and external painting and rendering just exit the industry. Every section has a factory-fitted weather membrane behind the cladding, carrying an overhanging taped flap that folds onto the next panel. Layering those overlapping flaps creates a continuously sealed envelope across the entire structure, turning a seam-seal job into a rapid process we confidently verify with a blower-door test. We finish the weather seal by clipping the exterior cladding directly over those section joints so there is zero reliance on running a bead of silicone to maintain the line, because trusting silicone for a lasting seal was never going to survive the distance.

The quiet part underneath all the hardware, and honestly the bit I care about most, is that the design isn't a drawing at all, it's factory data. A normal set of plans is a PDF that someone has to stand on site and interpret, and every time a human interprets it you have basically scheduled a defect for some point down the track. Ours isn't like that. The whole house lives as one model, and that model just compiles, the way code compiles, straight out into the cut files the rollformer needs, the bills of materials for each pod, the harness schedules, the approval dossier, all of it. The drawings and the parts are the same object, so they can't contradict each other, because there is nothing to disagree. We do the engineering once, properly, against the real load paths sitting in that file, then we log the torque on every bolt and proof-test the panels, and because we actually have that test data we get to throw away the structural over-sizing that a normal builder carries as a hedge against workmanship no inspector is ever going to look at. On a muddy site chucking in a bit of extra steel is just humility, fair enough. In a factory where you have measured the thing, it's weight the buyer paid for and didn't need.

When we add all those savings together, the roughly $500,000 cost of a typical build is dragged down to somewhere around $300,000, and this is exactly where nearly every single failed construction startup stalled and died. They usually hit a 40% saving right before watching road transport eat about half of it before the materials even reached the site, and then the next housing-market downturn inevitably wiped out whatever scraps were left.

What every one of them completely missed was the jump into second gear, meaning the actual difference between just building something indoors and genuinely manufacturing a product. Dragging construction into a shed, where you basically just have a bloke with a nail gun standing at a static bench knocking wall panels together, is really nothing more than site speed under a roof. It completely fails to change the underlying maths of how the whole thing actually assembles.

Real manufacturing means breaking panel assembly into totally isolated stations, where each station gets one single job and a jig that holds tolerance every single time, being fed parts that the rollformer has already cut to exact lengths and clearly labelled.

If we do that, the exact same labour hours produce two or three times the amount of wall. That labour cost we dragged indoors falls a second time, dropping from about $60,000 toward $25,000. We pressure-test absolutely everything before it even leaves the factory floor, which means every circuit and every pipe is checked, and every single bathroom pod is filled with water. Doing this removes the ~5% of a conventional build that is purely rework, like callbacks, supervision, or paying someone $80/hr to stand around watching a junior fix a mistake. On a normal site, that 5% hides deep within the trades and simply eats the builder's $70,000 margin, mostly because you will never see a subcontractor hand you an invoice with a line called mistakes. Our nesting software cuts panels from coil and sheet exactly the way a tailor cuts a suit. We are running at under 2% waste versus the 5-10% a normal site simply throws in a skip, meaning about $15,000 of materials is never bought in the first place. The running total now sits somewhere near $225,000.¹³

The biggest prize of all is about $115,000, and it sits buried in the materials themselves. You cannot claim it unless you decide to actually make the materials yourself, which led us straight to Terraform Materials. The plants we build, which we call Terraform Solar-Industrial Plants, operate nothing like the massive, dirty sites they replace, beginning with how they physically sit on the land. They scale up in simple one-megawatt chunks. A megawatt of solar panels, stretching over maybe two hectares of dirt, is wired directly to a megawatt of hot processing. That means kilns, melters, and smelting cells all lined up in a row. Dirt and sunlight go into one end, and Terraform Steel, Glass, Glass Wool, Aluminium, and Cement pour right out the other. We brand every single output, trace it right down to the batch, and drop it on a public price list.

These panels are not just some piece of infrastructure sitting in a paddock nearby. They are half the plant, in the exact same way wings are half an aircraft. If you need more cement capacity, you just add megawatts. When you need a steel plant, you line up the megawatts. Capacity grows linearly by simply adding blocks. There is never a giant central machine that ruins everything if you decide to switch it off at night.

These plants exist solely because of a truth about solar that most people only seem to half-notice, which is that panel prices have fallen for four decades and are still falling, so the electricity itself is heading toward being nearly free. Most engineers completely stop thinking right there. The catch is always waiting on the next line. If you bolt a factory to a solar farm, it only runs while the sun is up, which works out to about six good hours a day if we are being honest about Melbourne weather. A heavy line running for six hours carries four times the capital cost per tonne of one running around the clock. Pushing free electricity through an eye-wateringly expensive machine does not give you cheap steel. It just gives you a very expensive machine sitting completely idle in the dark eighteen hours a day. So the next thing that actually has to fall is the cost of the machinery the panels plug into. Conventional industry spent a century designing expensive plants fed expensive energy around the clock, and that was honestly right for that era. Terraform designs cheap plants fed free energy for six hours a day, and at today's panel prices the six hours simply win. Every piece of equipment, from the kilns to the presses, is built specifically to survive the sun's timetable. The methods that let a melt freeze off in the afternoon without wrecking the gear are essentially what our patents cover. These plants were built for intermittency, not for fighting it.

A few design rules run deep through every plant we build. We start by grabbing the panels' direct current and wiring it straight into our resistive heating elements, which operates on exactly the same principle as heating up water for your morning coffee. There are no inverters involved here. We use absolutely no grid connection, and we certainly do not rely on some football-field battery farm. Your toaster has run this exact architecture for a century without ever needing its own substation. Stripping the power electronics out entirely deletes $0.15-0.25 per watt of capital and simply avoids the 6-8% of energy those electronics usually eat, while also getting rid of the least reliable kit on site. The control system just ramps up with the morning light, coasts comfortably through a bit of cloud on the heat we banked in the furnace mass, and then ramps down at dusk. Everyone else seems to treat variable solar as a massive problem needing massive batteries, but these plants treat it as nothing more than a shift roster. Behind each plant's meter a daytime megawatt-hour costs us about $20. When we buy panels at seventy cents a watt and run them 1,600 hours a year for 25 years, it works out to $17.50, so we just call it $20. A fuel-fired plant pays roughly five times that much.

We also stick to mild conditions in very small increments, and steel is probably the best way to show exactly why this matters. Conventional steel requires a massive integrated mill loaded with a blast furnace, coke and sinter, billions to build, and thousands of staff. It runs continuously for decades because switching it off literally destroys the whole operation. The Terraform plant takes iron ore in and puts steel out through a protected, patent-pending process, which will not be described here, and that is exactly the point. The energy per tonne falls well below the conventional route, and the operating conditions are mild enough that the unit can be fabricated from ordinary steel by a local workshop down the road. Each megawatt, including the panels, costs about a million dollars instead of billions. It switches off nightly without a single complaint, and simply feeds the rollformers making the house frames. The exact same protected approach covers cement for the cladding and lining boards, as well as the civil works sitting under the communities, which I will describe even less. We then rely entirely on the absolute cheapest feedstocks on Earth, and glass shows this best. Glass is basically just melted sand. Glass wool is the exact same melt spun into a fibre, so one small electric melter running on plain sand and recycled cullet gives us both double-glazed units and insulation where the raw material cost rounds right down to zero. Pairing three tonnes of cheap glass wool with cheap double glazing actually lets the whole ducted heating system come entirely out of the plan. And here the national comedy has to be said out loud. Australia is the world's largest exporter of iron ore and digs up nearly a quarter of the world's bauxite.¹¹ We ship the mountains overseas to be cooked with someone else's coal, then buy them back as window frames at up to fifty times the price of the dirt, and we run the whole transaction under the best sun on the planet. That isn't a supply chain. It's a prank, and the plants end it. Aluminium uses the same logic just one notch up. Bauxite goes in and alumina comes out, and then direct current splits it into metal, mostly since aluminium has always been a DC process. A normal smelter gets absolutely wrecked by an interruption, which is exactly why nobody sane runs one on sunlight. Terraform solved that by designing hardware that can just switch off the moment the sun goes down, and cells that live comfortably on the sun's schedule are part of what the plant programme has cracked and part of what the patents protect.

The Terraform Materials catalogue just keeps sliding further down the house's bill of materials. The test for admission stays the same. The pre-finished internal linings and external cladding boards come straight out of the press using our own cement and sand. The kitchens and pods grab fired products from those exact same kilns. We're talking about wall tiles, sanitaryware, and sintered stone benchtop slabs. The timing of this is absolutely ridiculous, considering Australia just outlawed engineered stone outright.⁹ The country now imports basically everything in that category, dropping a market entry advantage in our laps that you couldn't buy for any money.

We don't buy bench surfaces. We cast them. We mould kitchen benchtops, shower trays, and bathroom vanities featuring integral basins from our own sand, cement, and chopped glass fibre. They cure in the identical chambers. When you cast a shower tray in a single piece and build the falls and upstands right into it, you kill off the failed shower membrane entirely. That membrane is the most common building defect ruining floors across the nation, and we just cross it off the worry list. Over on the framing lines, we take our coated steel and fold it into drawers and cabinet carcasses. By the time a kitchen leaves the factory floor, the only things in it we didn't build ourselves are the appliances, runners, and hinges. Glass that isn't turning into windows or insulation gets foamed down into lightweight aggregate to use as road base or drainage across our own developments. We even galvanise the steel and form the screws pinning it all together.

Every single material candidate has to survive a brutal gauntlet of three checks. I have to know if the current shelf price is mostly plant costs and stacked margins rather than the actual ingredients. We check if the feedstock is already sitting there as a near-free by-product. Finally, we figure out if the market price props up margin layers that a solar-industrial facility would naturally bypass anyway. If you answer yes to all three, you get a plant.

When you run on solar, the math looks totally different. The market price is whatever the incumbent product commands right now. Our target is the Terraform list price once we hit early production scale. These are rounded, representative figures, but staring at the gap between those columns is the whole point.

Look at those numbers closely. Something genuinely absurd happens. For every single material we produce, the physical inputs of feedstock and energy combined amount to between 2% and 15% of the market price. It sounds like a joke, but it's completely true. This points to the absolute subtlest argument in this entire essay. Cheap solar energy on its own basically saves you nothing. You might manage to knock $55 off a $1,400 tonne of steel. The energy bill was never the prize we were chasing. The actual prize is that nearly free daytime power lets you build daytime plants that are almost disposable. The remaining 85% to 98% of the price tag is simply the multi-billion-dollar capex of traditional plants, dragging five layers of margin behind them. Solar doesn't slash the cost of the material itself. It slashes the cost of the machines making the material, and the money was always hiding in the machines. That 85% to 98% is just old plant architecture designed back when energy was expensive. Then you wrap it in freight and stacked margins, throw a hi-vis vest on it, and call it the cost of materials.

Our target price isn't the physical floor. It's the bare physics combined with our machine amortisation and our labour at early scale. Every time we add a machine, that target drops lower. Megawatts are the whole capital strategy. We don't run a single monolithic plant anywhere in the company, because monolithic plants are exactly how people manage to lose a billion dollars. A solar-industrial setup is just megawatts placed side by side. You can add capacity a few thousand tonnes at a time. A single megawatt costs roughly a million dollars, solar panels included. A local workshop fabricates it. One crew runs it. We drop it wherever the market and feedstock sit, totally ignoring where the grid happens to be. One megawatt of cement plant spits out several thousand tonnes annually, amortising at tens of dollars a tonne over ten years. We proved the first plant's process at bench scale for a few grand. Sunlight and dirt are the primary inputs.

Apply this to the invoice. Bringing the energy-intensive materials in-house and only buying true commodities takes a savage axe to the bill. At early scale, the materials budget drops from $218,000 to roughly $100,000. Reaching steady state pushes that number into the floor at around $40,000. An 80% reduction happens in three specific waves. We strip out materials the structural design doesn't require at all. We manufacture everything else at bare machine cost. Finally, since the factory buys the remaining generic parts in bulk right at the gate, we completely bypass the massive trade and retail margins. The running total hits about $140,000.

Cheap materials make the big deletions possible. This is the cash that normally leaves a site and is never seen again. Plastering and painting suck $40,000 out of the original bill, existing entirely to force separate sheets of wall to pretend they are one continuous surface. Car makers abandoned that fantasy a hundred years ago. They gasket their panel gaps, design them in, and respect them. I have never once heard someone complain about the shutline on a Porsche. So our linings arrive from the factory already coloured. Every single join forms a crisp shadow line using a clip-on strip engineered to automotive tolerances. The total cost of this theatre is about $4,000. Fixing damage just means you unclip a strip. Ordinary houses leak heat like a sieve, which is why ducted heating and cooling costs $15,000. We pack three tonnes of glass wool—made for almost nothing by the melter—behind the walls, add double glazing, and give the thermal load to refrigerant lines. These run to a single outdoor heat pump that also handles the hot water, paired with a fresh-air heat exchanger. Total cost is $4,500. Drainage isn't some crazy scavenger hunt spanning the entire floor plan anymore. We just grab the bathroom, the kitchen, and the laundry and stack them back to back on a single plumbing wall. The entire waste tree ships from the factory as a welded, pressure-tested block. You plug it into the street connection once. The absolute cheapest pipe you can buy is the one you deleted entirely. The running total drops to roughly $110,000.

The final chunk of the original bill is the $102,000 of risk, paperwork, and time. This gets handled by structure instead of machinery, which is great because structure is far cheaper. The car industry solved certification ages ago. When you buy a new Corolla, nobody hauls that specific car off to a crash testing facility. They approve the core model geometry once. After that, they just register the individual units that roll off the line. We certify the Terraform system the exact same way. We push it through once for structure, fire, acoustics, weather, and energy. It sits there as a portfolio of pre-approved performance solutions, with the test evidence literally stapled on. Running a two to three million dollar certification programme across ten thousand homes comes to $300 a home. That absolutely demolishes the $20,000 of bespoke paperwork in a standard bill. It's a 98% discount that you purchase just by doing the documents in the right order.

Site approvals shrink down to what is genuinely local: the block, wind, connections, and soil. In Australia, our smallest product breezes straight through the existing secondary dwelling and movable dwelling rules. We start pulling in revenue inside the current law while kicking the door open for everything else. As for inspections, we answer those by handing over something building sites have never managed to produce. Total evidence. Every single panel carries a serial number. Every station takes photos of its output. We log the torque of every bolt, and we keep every pod's water test on file. We treat the factory to an annual audit like an aerospace supplier. A surveyor can open a single file packed with more verified proof than a hundred physical site visits, discharging the mandatory inspection stages immediately. Compliance gets significantly cheaper while getting drastically stronger. That's the only type of regulatory reform that ever actually happens. Licensed trades stay licensed, but the boundary shifts. Nobody asks a plumber to fabricate a dishwasher, because a dishwasher is a certified appliance. The licensed work is simply plugging the thing in. We certify the pod as a plumbing product and the harness as an electrical assembly. Site work for licensed trades collapses down to almost nothing. The electrician, the plumber, and the refrigeration tech each spend about an hour on location. The factory floor swallows roughly ninety percent of the trade hours that used to happen on site.

The housing boom and bust cycle historically gutted British factories. We answer that by just owning the demand. Terraform builds two distinct types of internal development. Our smallest product lets us build movable dwelling communities for holiday parks, tourism, and resorts under existing rules. We also develop residential communities for families. These serve as the fundamental base load. We plan them years in advance. When the wider housing market slows to a crawl, these projects absorb our factory capacity at bare cost. If paying customers flood in, we just step our internal projects aside. The tourism sites keep generating cash long after they are built, steadying the business exactly like the pipeline steadies the factory. External sales to farm stays, other developers, and families swing the way markets always swing. Because a stacked section is just generic until we actually configure it, a quiet month's production can happily wait in the warehouse without going out of fashion. The factory never actually needs the housing market to be in a good mood. We just need there to be a next project, and we own that project. Writing that sentence feels remarkably relaxing.

Margin drops to match the risk it genuinely covers. An eight-day site schedule backed by factory quality control and joints tested twice doesn't carry nine months of weather risk. The $70,000 margin line falls heavily toward a manufacturing margin on a much smaller base. The $12,000 finance line vanishes alongside the timeline it was funding. The terrifying gap between the build starting and the family actually moving in shrinks from nearly a year to about a week. Transport gets priced properly. A home ships as roughly twelve flat-packed sections, packing sixteen to a high-cube container. That's three-quarters of a container per home, costing $2,000 to $3,000 to reach almost any port. That beats paying $12,000 to $25,000 dragging a finished module a few hundred kilometres behind escort utes. Cement never travels at all, because the machines make it from the local dirt right next to the market. Dense components travel. Heavy and bulky materials stay home. This model lets a handful of large, fully loaded factories serve an entire continent. Continental throughput pays for the whole operation.

One structural choice stops all our construction savings silently evaporating upstream before the buyer gets a look. We put our prices in public. Every home model has a standing national price, and every single component has a list price. Every Terraform Material, the steel by the tonne, glass wool by the bale, cement by the bag, sits right there for anyone to see. That includes other builders. You won't find quotes or variations here. When you certify a system and throw it together in eight days, nothing is left to vary. The dirt itself is the lone exception, where we price piles off the soil test, mostly because the Earth refuses to publish its own price list. A published price stops intermediaries from quietly absorbing it as their own margin, and open supply means our customers can always bypass anyone trying to take a cut just by paying list. This whole thing runs on the commercial logic of volume. Drop the cost to a tenth and suddenly the pool of people who can actually buy a home expands massively. Selling into a pool that big on a razor-thin published margin earns a company far more than trying to harvest a scarce product for maximum profit. It's rare, but the lowest price and the most profitable strategy are the exact same thing this time.

I need to own up to something about the overheads line, because factory ventures love hiding their blind optimism right there. A budget of five thousand dollars per home assumes you've got one massive factory churning out around five thousand homes a year, highly utilised. That money covers the building, writing down the equipment, indirect staff, maintenance, and insurance. The second your utilisation drops by half, that line item doubles. That simple bit of maths is the entire financial justification for our community pipeline existing at all.

Here is the destination, line by line.

Look at the numbers: $69,000 against $500,000 is 0.138. The multiplication I did earlier in this post predicted 0.105. Having two calculations arrive from different directions and land a few hundredths apart tells you something. Either they are both wrong in the exact same direction, or you'd better start building the company. Cumulative volume closes the remaining gap. It is the most dependable force in manufacturing: costs drop as cumulative volume doubles. That exact curve dropped solar panel prices by 99% without anyone pulling off a miracle.¹² People just manufactured a huge amount of them while paying close attention to what they were doing.

To be upfront about how this sequences out, that table is our steady state. Right now, the product we ship comes in 30-40% below conventional cost, made from factory sections using bought materials. Over the first two years, we'll see the first solar-industrial plants show up megawatt by megawatt, each one packing its own panels. Those will run alongside some incredibly boring but lucrative electrification wins in rollforming, glazing assembly, and glass wool. We're talking low single digit millions, paid back entirely by the materials delta on each home we ship out. Replication comes after that. More megawatts, the iron ore furnace, the aluminium line, a glass melter supplying windows and insulation from a single melt. We commit capital megawatt by megawatt, with each carrying its own payback, so we never have to bet the house on one giant plant. Then we bring automation to the panel and pod lines. Software will take any floor plan and spit out cutting files and approval documents, and then the first fully complete Terraform communities will start going up, acting as the factory's permanent order book.

How do you build housing ten times cheaper? You have to admit there is no single answer, mainly because the cost doesn't have a single cause. You go after the whole mess at once inside one system. Shove every decision into the design file and the factory, leaving the site as nothing but a bolting exercise, and your labour costs halve. Quit buying materials through five layers of margin built on expensive heat. Make them from soil, ore, sand, and sunlight in megawatt-scaling solar-industrial plants, and materials halve too. Get the system certified once instead of doing it for every single house forever. Ship dense components instead of wasting highway space on bubbles of air. Let an eight day build wipe out the enormous risk that margin and finance were invented to price in, and your overheads plummet. The absolute cheapest part of any building is the one your design doesn't even need anymore, so we delete the plaster, the paint, the ducts, and the vast majority of the pipes. Stacking those four reductions gets you down to one tenth. It matches the other calculation perfectly. The learning curve does what it has done for every other manufactured product and handles the final stretch.

I need to explain exactly why this company takes the shape it does. None of this post appeared from nowhere. Everything sits on years of research and development across my companies, covering diagnostics, materials, and manufacturing. I'm talking about the unglamorous kind of R&D where the majority of your experiments fail but the invoices show up regardless. Australia has the R&D Tax Incentive, a program that's supposed to back exactly this kind of work, and on paper it looks world class. In reality, it's a refund you receive, spend on your next round of research, and then have to defend for years when the assessment arrives long after the cash is gone. We have been contesting $5.7 million in research and development refunds with the tax office since 2022. I will not litigate the dispute here; that's what the tribunal is for. I'm calling out the pattern because any founder in this country instantly knows what I'm talking about. We live in a place that practically throws a parade for a bloke digging a rock out of the ground, then audits the engineer who actually attempts to make something useful out of it. The department sets the timeline however they like, they drop the burden of proof onto the innovator, and fighting them costs more cash than a startup usually has left. I took design away from that lesson, not bitterness. I built the company described in this post so it needs absolutely nothing that anyone can retrospectively withdraw. No grants. No subsidies. No incentives. No favourable rulings. You can't claw back the sunlight. We don't need a private binding ruling to use the rock. We built every system above so that applying the rulebook exactly as written is the only thing we need, and the only thing we ask of government.

The whole point of this arithmetic is what it actually buys. Family formation is collapsing across the developed world, and the cost of housing is the biggest lever anyone has to fix it. A society where a young couple can snag a home for a single year of their salary is a place that actually has kids, takes risks, and gets things built. The dirt under our boots already contains the base ingredients for cement, steel, aluminium, and glass. Every single day, the sun blasts 10,000 times our civilisation's energy use right onto that exact same dirt entirely for free. Our first bench process already proved it works.¹⁴ Getting the cheap, replicable megawatts is all it takes to build the solar-industrial plants that bring them together.

We are building them now.

References

1. Gilbert Herbert, The Dream of the Factory-Made House: Walter Gropius and Konrad Wachsmann, MIT Press, 1984 (online edition); Pacific Coast Architecture Database, General Panel System House. ↩
2. Brian Potter, "The Lustron Home," Construction Physics, 2023; Ohio History Connection, Lustron Corporation records. ↩
3. United States v. Stirling, 571 F.2d 708 (2d Cir. 1978); In re Stirling Homex Corp., 579 F.2d 206 (2d Cir. 1978); Harvard Business School Case 173-193, "Stirling Homex (A)". ↩
4. Toyota Motor Corporation, 75 Years of Toyota, "Initiatives in the Housing Business". ↩
5. TechCrunch, "SoftBank-backed construction giant Katerra said to be shutting down after raising billions," 1 June 2021; Architect Magazine, "Katerra's $2 Billion Legacy," June 2021. ↩
6. TechCrunch, "Prefab home builder Veev reportedly shutting down after reaching unicorn status last year," 27 November 2023; Calcalist, 26 November 2023. ↩
7. Construction News, "L&G to stop making modular homes," 5 May 2023; Building, "L&G's failed modular housing business records £61.1m loss," 30 September 2024. ↩
8. Harvard Joint Center for Housing Studies, "Comparison of the Costs of Manufactured and Site-Built Housing"; NAHB, Eye on Housing, "Manufactured Homes: An Alternative Means of Housing Supply," 2025; US Census Bureau, Manufactured Housing Survey. ↩
9. Safe Work Australia, "Silica: Engineered stone ban". ↩
10. Australian Bureau of Statistics, Producer Price Indexes, Australia, Input to the House construction industry, March quarter 2025. ↩
11. US Geological Survey, Mineral Commodity Summaries; BMI Global Bauxite Mining Outlook via Australian Mining, 2025. ↩
12. Our World in Data, "Solar (photovoltaic) panel prices"; IRENA, Renewable Power Generation Costs. ↩
13. Get It Right Initiative (UK), research reports on error and rework in construction. ↩
14. Terraform internal specifications and modelling: TF-SR-001; TF-CS-001. ↩