Dustin DeVan once stood up in a room of 40 to 50 real estate investors, right after Katerra had closed a fresh round, and told them the company was going to zero. He was right. Years later, on the podcast, Handoff’s Dmitry Alexin posed a hypothetical: with hindsight, but Masayoshi Son still on the board, what would you have actually done differently? Dustin, Patric Hellermann, and Martin gave three different answers, and none of them fully agree.
Katerra remains construction tech’s most expensive cautionary tale. Founded in 2015 by Michael Marks, the former Flextronics CEO, the company set out to control an entire building’s delivery, architecture, manufacturing, and on-site construction, under one roof. SoftBank poured more than $2 billion into the bet before Katerra filed for Chapter 11 in June 2021, listing liabilities as high as $10 billion against assets of roughly $500 million to $1 billion.
Dustin’s story about the company predates the collapse by years. He was in a room with dozens of real estate investors shortly after Katerra closed a round, everyone buzzing about the raise, and he stood up and told the room the company was headed to zero. “People thought I was crazy,” he said on the show. “It just raised a round.” Tom Feliz, sitting next to him, can apparently confirm the reaction in the room.
What makes the podcast segment worth reading isn’t the prediction, it’s what came after it. Dmitry, who runs an AI estimating company himself, asked the obvious follow-up: if you’d taken the CEO seat with everything you know now, but Masa still controlled the board and still wanted the vision realized, what would you actually do? Three people, three different diagnoses.
Three Answers to the Same Question
Structure, model, or just more runway
Dustin’s diagnosis centered on structure. Full vertical integration, he argued, meant Katerra was exposed to a huge number of single points of failure at once, architecture, manufacturing, and on-site delivery all under one roof, all needing to work simultaneously. Even the prefab strategy compounded the problem: shipping distance limits how far a factory-built product can travel economically, which meant scaling geographically required building new factories in new markets rather than simply exporting from one, layering manufacturing risk on top of construction risk on top of architecture risk.
Patric’s answer moved the bet somewhere else entirely. Given the hypothetical, he said he would have redirected the company toward building a foundational physical AI model for construction, five to six years before anyone else was seriously investing in that idea, arguing the equity story behind Katerra’s actual business model was never going to hold up at exit regardless of execution. He added an important caveat: he wouldn’t have taken the job at all without control of the board, since the whole plan depends on being able to override the instinct to keep chasing the original vision.
Martin’s read was the least dramatic and, in its own way, the hardest to argue with. Not a fundamentally broken idea, he said, just underfunded and rushed for what it was trying to do. “If they had ten, fifteen more billion dollars and five, ten more years, they would have done it,” he said. “This is just a matter of money.”
| Who | Diagnosis | The catch |
|---|---|---|
| Dustin | Vertical integration created too many simultaneous points of failure | Fixing it means abandoning the core structural bet, not tweaking it |
| Patric | Should have bet everything on a foundational physical AI model, years early | Only workable with full board control from day one |
| Martin | Not the wrong idea, just underfunded and rushed | Requires believing $10-15B more would have been enough |
Why This Argument Matters Beyond Katerra
The industry punishes people who assume it’s simply behind
Earlier in the same conversation, Dustin drew a direct line from Katerra to a pattern he sees repeatedly in construction tech investing. Founders and investors arrive convinced the industry is inefficient because it’s backward, without pausing to ask why a given practice exists in the first place. Some of it is legitimately due for disruption. Some of it survives because it solves a real problem that isn’t visible from outside. “There is a lot of logic behind some of these things,” he said, pointing to construction’s status as one of the oldest industries on the planet as a reason those reasons rarely announce themselves.
Patric’s framing softened that into something closer to a discovery process rather than a disqualifier. Coming in with a first-principles instinct to break the rules is fine, even valuable, as long as it’s paired with enough humility to check whether a rule being broken was load-bearing. Handoff’s own Dmitry made almost the identical point earlier in the episode about takeoffs specifically: a lot of what looks inefficient in construction turns out to have a reason once you actually sit with the workflow. Katerra is what happens when that checking step gets skipped at a $2 billion scale.
The failures teach as much as the wins
Founder frameworks, funding logic, and the postmortems nobody else writes. Every week.
Join 3000+ ReadersKaterra was a vertically integrated construction startup founded in 2015 by Michael Marks, aiming to control architecture, manufacturing, and on-site construction under one company. Backed by more than $2 billion from SoftBank, it filed for Chapter 11 bankruptcy in June 2021 after years of project delays, quality control issues, and mounting losses. (Source)
On the podcast, Dustin DeVan described telling a room of 40 to 50 real estate investors that Katerra was headed to zero, shortly after the company had closed a funding round. He said the reaction at the time was disbelief, which he now recalls as a point of pride given how the company’s story ended.
Three different answers emerged. Dustin pointed to the vertically integrated structure itself as the core flaw. Patric said he would have redirected the company toward an early bet on a foundational physical AI model for construction, but only with full board control. Martin argued the original plan wasn’t fundamentally wrong, just short on capital and time.
Reporting at the time put SoftBank’s total investment at over $2 billion across multiple funding rounds between 2018 and 2020, including a $200 million rescue infusion in December 2020 that gave SoftBank a majority stake shortly before the company’s eventual bankruptcy filing. (Source)
Dustin’s argument on the podcast was that investors and founders without construction backgrounds often assume the industry’s slower, more manual practices are simply backward, without checking whether a given practice exists for a real, if invisible, reason. Some inefficiency is truly fixable. Some of it is load-bearing, and removing it without understanding why it’s there is where expensive bets go wrong.
Related Articles
How to Build, Invest and Buy in ConTech: Patric Hellermann’s Founder Framework
SoftBank’s $100B Robotics Bet: Why Roze AI Wants to Build the Data Centers AI Runs On
