NotesNo. 03AI
The house, off the rack
Clothing went from tailor-made to ready-to-wear. Physical AI could do the same for homes, and then go one step further.

Two hundred years ago, almost every shirt was made by hand. One tailor, one customer, one shirt at a time. Then came machines, standard sizes and factories, and clothing turned from a craft into a product. Today you can walk into a store and leave with a jacket that fits in ten minutes.
Houses never made that jump.
We still build houses the way people made shirts before factories: one at a time, by hand, outdoors, for one customer.
That is starting to change, and the reason has a name: physical AI. It means models that do not just write text but see, plan and move things in the real world. Put that inside a factory and the oldest industry on earth starts to look like a product business.
Why now
Three pressures are lining up at the same time.
- Not enough homes. Estimates of the US shortage range from about 1 million to 10 million homes, depending on who is counting. Even the low end is a lot of houses.
- Not enough hands. US construction needs 349,000 more workers in 2026 and 456,000 in 2027, just to keep up. Data centers are pulling electricians away from everything else.
- Robots that can handle variety. Old factory robots repeated one motion forever. The new ones can read a design and adjust. That matters, because no two houses are quite the same.
Off the rack, three ways
Startups are attacking this from different angles. Think of them as three ways to buy clothes.

The tailor’s workshop: make the parts in a factory. Promise Robotics takes a builder’s existing house plans, turns them into robot instructions and makes the wall panels in a factory. It calls the model Factory-as-a-Service. D.R. Horton, the largest US homebuilder by volume, signed a multi-year deal covering 1,250+ homes. Today the company announced a $40 million CAD Series B. The house still gets assembled on site. The hard, repetitive part moves indoors.
Ready-to-wear: ship the whole thing. Boxabl builds homes that fold flat, ship on a truck and unfold on site. It has built more than 700 of its small Casita units, listed on Nasdaq in July 2026, and signed a deal for up to 1,500 homes over three years with a multifamily developer. The open question, as with every modular company before it, is volume.
The jacket you take with you: a home that moves. This is where AC Future comes in, and it is the most interesting idea of the three. The Irvine startup, which raised a $15.6 million seed round, designs its homes with Pininfarina, the studio behind some of the most famous Ferraris. Its first product, the AI Transformer Home Trailer, expands at the push of a button to about 360 square feet, makes its own power and starts at $158,000. A drivable version unfolds into a 400-square-foot home and makes its own water.
AC Future’s bet is different. It is not trying to build houses faster. It is trying to make the home behave like a product you own, upgrade and take with you. First deliveries were planned for the second half of 2026. As of this week I could not find a public announcement that they have started. That is the milestone to watch.

We have seen this movie
Here is the part that keeps me honest. Factory housing has a long history of beautiful demos and painful endings. The best-known modular startup of the last decade, Katerra, raised more than $2 billion from SoftBank and filed for bankruptcy in 2021.
The lesson was not that factories do not work. It was that a factory is a fixed cost waiting for volume. If orders come in slowly, every empty hour on the line eats the savings. Clothing solved this with huge, steady demand and standard sizes. Housing has neither yet.
That is why the business model matters more than the robot. Promise Robotics plugs into big builders who already have the orders. Boxabl and AC Future have to create their own demand, one buyer at a time. Both can work. They just carry very different risks.
Follow it a few steps
If physical AI really makes homes into products, the effects go well beyond construction.

- Price like a product. You would see a sticker price, a spec sheet and model years, like a car or a phone.
- Finance like a product. A mortgage assumes a house appreciates. A product usually depreciates. Lenders and insurers will need new models, and whoever builds them wins a big market.
- Land becomes the scarce thing. If the box gets cheap and fast, the slow part is the plot and the permit. This is the same pattern I wrote about in the last note. Solve one bottleneck and the next one shows up.
- The home plugs into the grid differently. A home with its own solar and batteries, like AC Future’s, is a small power plant. Millions of them start to matter to utilities.
What I would ask as an investor
- Utilization. How full is the factory, really? That number decides the margin.
- Orders, not reservations. A signed contract with a builder is worth more than ten thousand waitlist sign-ups.
- The fitting room. Who handles land, permits and hookups? That is where projects die.
- Financing. Can a normal buyer get a loan for it today, and at what rate?
A shirt did not get cheap because someone built a better needle. It got cheap because of standard sizes, factories that ran all day and stores that sold to millions. Homes will follow the same path, just slower. The companies that win will not be the ones with the coolest robot. They will be the ones that keep the factory busy.
If you are building in this space, I would like to hear from you. I read every note myself.
Peter