The Methodology Bet Survived Katerra. It Lives In Robotics Now

Silver bullets rarely describe themselves as such and yet during its peak being the silver bullets was essentially the business model for Katerra. Then Chapter 11 happened in June of 2021 and it was seemingly proof that vertically integrating an industry from the design layer down to the panel factory does not, on its own, fix the production system for construction. The industry absorbed the lesson as a verdict. Construction can’t be rethought from the methodology layer. Manufacturing logic doesn’t translate. Move on.

That conclusion is half right.

When Stuart Maggs, then Director of Commercial Product at Field AI and now Commercial Director at Algorized, joined the Construction Conversations podcast, the conversation landed on exactly that distinction. My argument on the call was that Katerra's bankruptcy had been misread as a verdict on the underlying thesis, when it was really a verdict on the execution. The diagnosis was correct. Construction's production system is overdue for an end-to-end rethink. The way Katerra tried to deliver that rethink (developer, manufacturer, GC, and tech company at the same time, financed by capex) is what folded. Stuart agreed without hedging. "They were onto the right idea though, because they were basically saying everything you're saying. But they just didn't do it." This is the argument I have been making for years, now with a robotics company that just raised $405 million willing to make it out loud.

Key Takeaways

• Katerra was a verdict on one operating model, not on the methodology thesis. Vertically integrating design, manufacturing, GC services, and tech inside one company at $2 billion-plus in capex was the failure mode. The underlying argument, that the production system can be rethought, is still defensible.

• The new methodology bet is software-first, hardware-light, and built to license, not own. Field AI raised $405 million without a factory plan. The thesis is to own the autonomy layer, license it to robotics companies, and let everyone else carry the capex of the physical layer.

• Hardware has been good enough for years. Intelligence has been the bottleneck. Stuart’s Boston Dynamics Spot example: Spot fails the moment the environment changes because it follows a static map. Field AI’s bet is that the brain that handles job-site uncertainty is where the defensible product lives, and that brain is software.

• The tool is still rarely the bottleneck. Incentive structures are. Stuart’s drywall illustration: trades get paid by quantity installed, not by what is finished. A robot that ignores the unit of pay will not get adopted no matter how good the chassis. The procurement model still has to change before the methodology bet pays out at the trade scope.

• Receipts in construction robotics are still concentrated in reality capture. Stuart conceded that 360 capture and laser scanning are “pretty mature, we’re delivering this now.” Everything else requires a per-vertical value case. Diligence on construction robotics in 2026 starts with “what scope are you billing for today.”

• By 2030, a data center build will have more than ten robots on site. Stuart's forecast was specific. Five Field AI units doing 360 capture from first day to last, a drone overhead, two Dusty robots laying out from the model, two more doing material transport, and a couple doing some flavor of trade work. None of them require a factory to exist. All of them require an autonomy layer to work.

Katerra wasn’t the end of construction’s methodology revolution.

What broke Katerra was the choice to attempt the methodology bet as a developer, a manufacturer, a general contractor, and a technology company all at once. Not the thesis underneath. The thesis is still defensible. Construction wastes a staggering amount of value on a production system that has been left alone for too long. Stuart’s view, which I share, is that a willingness to rethink the process end-to-end could deliver something faster, cheaper, more reliable, with a healthier supply chain. That part of Katerra was right. The receipts for being right cost $2 billion.

The procurement side is adapting too. EllisDon's 2026 ConTech Accelerator, applications open through June 30, gives construction tech startups direct collaboration with the GC's teams and a defined path to pilot solutions on active projects. That is the buyer-side counterpart to what is happening on the venture side. Smaller bets, faster proof points, structured on-ramps that didn't exist in the Katerra cycle.

Now look at Field AI. The company raised $405 million in 2026. The capital is not going into factories. It is going into engineering on an autonomy stack that runs on someone else’s robots. Stuart’s read on the underlying technology is that hardware has been good enough for years, and the bottleneck has been intelligence. He used Boston Dynamics’ Spot to make the point. Spot is a beautifully engineered platform that fails the moment the environment changes, because it follows a static map. Job sites change every day. Field AI’s brain doesn’t sit down when the hallway closes. It semantically understands the environment and finds another way to accomplish the mission. That brain is a software product, not a manufacturing one.

The vision extends one layer further. I asked Stuart if Field AI was essentially trying to be a Google Play Store for robotics. He agreed. Every robotics startup that launches in 2027 should, in his read, use ROS for the foundation and Field AI’s stack for navigation, because there is no point doing the work yourself. That is a structurally different bet from Katerra, which tried to own the application AND the platform AND the supply chain AND the project. The new bet is to own one defensible layer and to license it broadly. The same pattern repeats across the post-Katerra category. ICON has scoped its concrete 3D printing into specific geographies rather than promising universal coverage. Aro Homes, founded in part by people who lived through Katerra from the inside, designs the product to a specific site type rather than asking one factory to produce variety. Prefab consultant Amy Marks summarized the shift cleanly: prefab is in a better place today than during the Katerra cycle, but ambitious startups are not leading the way anymore. The category survived. The execution model didn’t.

Here is where my standing skepticism kicks back in. The tool has rarely been the bottleneck in this industry. The incentive structures around the tool usually are. Stuart said it himself when we talked about drywall. UK installers get paid by quantity of backboard installed, not by what gets finished. A robot that ignores how the trade is priced does not displace the trade. His three-rule adoption test made the same point from the other side. If a robotics solution does not help you win the next job, make day one cheaper, or make day two cheaper, “why would you?” The new methodology bet has not solved that. And in the receipts category, Stuart was honest that the only mature construction robotics use case today is reality capture and 360 capture. “We’re delivering this now.” Everything else still requires a per-vertical value case. The new bet has technological credibility. It does not yet have the procurement plumbing or the dollar-attributed case studies to back it across most of the trade scope.

The most uncomfortable counter is the one Katerra would have made before it folded. Software does not pour concrete. Code does not bolt a connection. An autonomy stack still has to land on hardware that someone has to manufacture, ship, maintain, and warranty. Field AI’s bet depends on a hardware partner ecosystem that has not fully solved its own unit economics. The methodology is being rethought from one direction. The job still has to get built from the other. That is the same friction Katerra ran into, attacked from a different layer.

What changes my read on this round is the discipline. The previous attempt collapsed under the weight of trying to solve every layer at once. The new attempt is willing to do less and focus its growth on closing the gap with the field. The startups that win this round will be the ones that pick a thin software wedge, combine it with outcome focused delivery and use capex to fully integrate a workflow. The GCs and owners who win will evaluate these vendors alongside a keen understanding of their own IP and the data that feeds it.

The expensive lesson Katerra paid for the industry was that you can be right about the diagnosis and still go bankrupt on the execution. The current wave has read the receipt. Whether that is enough is the question I am watching.

About Stuart Maggs. Stuart is Commercial Director at Algorized, an advanced people-sensing company. He previously served as Director of Commercial Product at Field AI and remains an advisor to the company. His background spans architecture, robotics research, and early-stage 3D-printing work, with continued focus on construction-adjacent technology bets.