Reframe Systems raised $40 million in a Series A this week to make houses in sections in a factory, truck them to the site and crane them into place, with robots doing a growing share of the work.
The standard robotics answer to construction is one very large plant running a single product down an assembly line. Vikas Enti has no assembly line at all, buys ordinary industrial arms off the shelf, and says the problem is variety rather than volume.
"It is and we've also taken a stance where we have no concept of an assembly line."
Enti came out of Amazon Robotics, where he says the only lesson that mattered was to constrain the problem before applying automation. He gave the interview sitting inside one of his own units.
I listened to the full segment so you can skip it.
Here are the 7 takeaways that matter.
👤 Guest: Vikas Enti, co-founder and chief executive of Reframe Systems, which builds modular homes with robots and raised $40 million in a Series A this week; previously at Amazon Robotics
🎙️ Host: Akash Pasricha, who anchors TITV, The Information's live weekday show at 10 a.m. Pacific
📰 Published: 31 August 2026 on YouTube (The Information)
🔴 YouTube | ⏱️ 12 min
Key Takeaways
The product is not a robot, it is a factory that makes chunks of buildings
The sections travel on a flatbed truck and are craned into place on site
One accessory dwelling unit is a single module about 14 feet by 40 feet
A five-story apartment building later this year takes roughly 80-plus of them, some trapezoidal
Robots today fabricate the wall and ceiling panels, and the roadmap is the wiring and plumbing
He expects about 65% of the work content to end up robotically fabricated
He built the company with no assembly line, on purpose
He says earlier attempts misread this as mass production and built capital-intensive mega factories
Off-the-shelf arms plus his own vision software beat a purpose-built system at a fifth of the price
The company has no AI model of its own yet, and he says it is not the constraint
He expects robotics models to stay fragmented, because the robot is a tool and not the answer
His comparison is earth-moving equipment: a century-old, well-understood problem with a dozen manufacturers
1. Homes chunked into modules
Pasricha opened by asking how the robot actually works, and Enti reframed the question around the factory.
The company is building small factories across the country to produce high-performance homes, moving work that normally happens on a building site into a manufacturing setting
The physical process is the same one used for any prefabricated structure, done at higher precision. "We're effectively chunking up homes into smaller sections and components that we produce in a factory, ship them on a flatbed truck, crane them into place"
The experience he says he is trying to recreate is a home appearing overnight
What is new is the software and robotics underneath it, which he says make manufacturing significantly faster and significantly more cost-efficient than doing the same work on site
2. One module is 14 by 40 ft
The number of modules depends on what is being built. A single accessory dwelling unit — a second small home on an existing lot — arrives as one module
The dimensions are the size of a very large shoe box. "So, it's about right 14 feet wide by 40 ft long."
Asked whether that means a one-bedroom apartment: "Exactly. So, it's like it's a studio with a kitchen and a bath. In fact, I'm sitting in one of those units right now."
The shapes are not limited to boxes. Larger projects run to about six modules and include gable roofs, so the components are not all cuboidal
The largest project on the books is an apartment block. "We have a five-story apartment building coming up later this year that's going to have roughly 80 plus modules. Some will be trapezoidal."
"So we get to mix and match the component types as required by the project."
3. Shipping homes to California
Everything is manufactured in-house from ordinary materials. Reframe buys studs and sheet goods from a lumber yard or direct from the manufacturer, then makes the components itself
The software is what turns a building into machine instructions. He said the company models a building to the nearest fastener, pipe and wire, and translates that into instructions for its computer-controlled saws and routers and for the robot
There is one operating factory today, in Massachusetts, with his own unit next to it, and the company is now launching a production-scale plant called Fab One, also in Massachusetts
It has already shipped finished product across the country. "We have done work in California to help with the wildfire rebuild." Those were pilot projects
West Coast capacity is coming through developers rather than through Reframe alone. He said the company is in discussions with a couple of developers to open factories there, to be announced later this year
4. Robots frame the panels
The robots do component manufacturing, not assembly. "So today it's in component manufacturing. So our robots actually fabricating the wall panels and ceiling panels."
The roadmap is to push them into higher-value work — insulation, wiring components and plumbing inside the walls
He put a number on where that ends up. "So over time we'll see about 65 plus% of work content robotically fabricated and we'll still have a highly variable human component."
The point of the automation is to lower the skill required, not to remove people. He said the goal is builders of all skill levels, with instructions printed directly onto the wires and studs and shown on screen
"So we've had high school co-ops come in and do plumbing and wiring work."
"It feels more like assembling Lego bricks and assembling your IKEA furniture", with the output built to code
5. No assembly line at all
Pasricha said he was picturing rows of robotic arms of the kind Amazon runs in its warehouses, and asked whether that is what the facility looks like.
Enti confirmed the arms and then rejected the layout. "It is and we've also taken a stance where we have no concept of an assembly line."
His classification of the problem is what drives that. "We believe that this is a high mix manufacturing problem." Every project is different, so the plant has to make many things rather than many copies of one thing
He said earlier entrants got this wrong, treating homebuilding as mass production and trying to reach a very capital-intensive mega factory
The cost advantage comes from software over hardware. "What we've done instead is picked traditional off-the-shelf industrial arms, but built our own vision system, a lot of vision algorithms"
The result he claims is a $200,000 robot doing the work of a million-dollar robotic system, in a much smaller footprint
The plant is a collection of those arms, each able to build different components of the home over time
6. Classical vision, no model
Asked whether the vision system means Reframe trained its own AI model, Enti was direct about not having one.
He acknowledged the fashion and declined it. "I mean all the cool kids have models today."
The company is still training its vision systems using classical methods until it has enough data to justify training a full model
The absence is not holding anything up. "We're not limited by a model's capability to achieve our objectives in our near term."
The reason to keep going is the data, not the model. His stated end state is having built enough of a body of work and enough data sets to build the right system — one that keeps using off-the-shelf specialized hardware for tasks it was never designed to do, because Reframe has created the environment for it
7. Robot models stay split
Pasricha closed by asking how the model layer for robotics plays out: a few giants the way the AI labs dominate, or something more fragmented.
Enti said he does not have the perfect answer and gave his framework instead. A robot is a tool for solving a problem, so the discipline is to obsess over the problem — here, industrializing home building — and then build the right tools for it
His argument for fragmentation is that constraining the problem is where the leverage is. Reframe removes most world-model requirements by confining the work to a factory
"Like my whole experience and only experience from Amazon Robotics has been really like constrain the problem."
"So you have maximum leverage in applying automation in robotics."
The sequence he expects is problem, then architecture, then a model if one already exists. Define the problem properly, pick the solution architecture, then look for a ready-made model that saves engineering time on vision or manipulation — and if none exists, tune an open-source one or build from scratch
The analogy he used is construction equipment. "It's a well understood problem. It's been happening for 100 years but you still have a dozen different manufacturers." Even component models vary widely, because the use case is nuanced by project size and type of work
Asked to confirm the conclusion, he did. "Short answer yes. I think it'll be it'll stay fragmented for a while."
"It's a tool in the tool stack and depending on who's designing the overall solution you'll end up picking the right set of components to come together."
Bonus Insights
The funding is the news peg. The Information reported exclusively that day that Reframe Systems had raised $40 million in a Series A this week
Pasricha framed the segment around a wider point: the robotics companies with narrower ambitions are the ones more likely to reach commercial success in the near term, against the noise around humanoid robots
Enti was interviewed from inside one of the company's own studio units, next to the Massachusetts factory, and agreed to give a tour of it on a future appearance
This was the fifth of five interviews in one episode; the other four are written up separately
Enti's bottom line is that homebuilding is a high-variety manufacturing problem rather than a mass-production one, and that the way to automate it is to shrink the problem until cheap standard hardware can do the job — which is also why he thinks no single robotics model will end up running the industry.
Products, Companies & Tools Mentioned
Reframe Systems (His company: factory-built modular homes, one operating plant in Massachusetts plus a production-scale one called Fab One, $40 million raised this week)
Amazon Robotics (Where he worked before founding Reframe, and the source of the principle he keeps returning to: constrain the problem)
IKEA and Lego (His comparison for what assembling a Reframe home feels like for a worker following screen-guided instructions)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: