All3 Raised $25 Million to Turn Housing Into a Robotics Problem

A lot of startups say they are “rebuilding” an industry when what they really mean is they made a dashboard for bothering that industry more efficiently. Then there is All3, which looked at Europe’s housing shortage, construction’s productivity slump, and the average building site’s devotion to chaos, and decided the answer was: what if architecture software, robotic factories, and a four-legged construction robot all shared one extremely opinionated brain?

On April 29, 2026, All3 announced a $25 million seed round led by RTP Global with SuperSeed, Begin Capital, s16vc, and VNV Global. Which is a serious amount of seed money, but also not an absurd amount if your startup pitch includes factories, custom timber components, and a robot named Mantis that sounds less like SaaS and more like a minor boss in a game about municipal permitting.

The robot is not the product. The robot is the punchline delivery system.

What All3 actually does is broader, and honestly smarter, than “construction robot startup” suggests. On its integrated technology page, the company lays out a full-stack system: AI-driven architectural design, robotic fabrication, a manufacturing software layer, plug-and-play structural components, and autonomous on-site assembly. The sales pitch is that the old construction workflow is too fragmented, too manual, too slow, and too allergic to modern software to meet the housing demand staring Europe in the face.

That diagnosis does not exactly feel controversial. Construction has spent years behaving like digitization is a rumor started by consultants. So I understand why investors hear “vertical integration for housing delivery” and briefly levitate.

All3’s core claim is catnip to both developers and VCs: better buildings, faster timelines, lower cost, less carbon, fewer skilled-labor bottlenecks, and no aesthetic penalty for awkward urban plots. The company says it can cut development time by up to 50%, reduce costs by up to 30%, and lower embodied carbon by up to 25%. It also says its system can turn a traditional nine-month architectural process into something much faster, which is the kind of sentence that makes one half of the built-environment world say “finally” and the other half clutch a hard hat and whisper “absolutely not.”

Mantis, meet permitting. Permitting, be nice.

The funniest part of this startup is also the most compelling: Mantis is not trying to be a general-purpose humanoid that folds laundry, opens doors, and eventually testifies before Congress. It is a purpose-built construction robot. According to All3, it handles placement, fastening, finishing, and inspection on constrained urban sites, which is an admirably unglamorous list. No robot bartender nonsense. No “companion intelligence.” Just a machine built to show up, align components, and not get distracted by a keynote.

That restraint is part of why this works for me. I have more patience for robots that do one hard, boring, economically meaningful thing than for the endless parade of humanoid demos asking us to imagine a future that mostly looks like a warehouse with better cinematography. In that sense, All3 feels spiritually closer to Anvil Robotics’ Lego-for-machines charm offensive than to the giant category theater around Apptronik’s very expensive humanoid ambitions.

And the company is at least aiming at a problem with genuine civic mass. In its 2025 launch materials, All3 argued Europe faces a housing shortfall of up to 9.6 million homes, with Germany alone projected to miss its 2026 target by roughly 225,000 homes. When founders say they want to “solve housing,” my default reaction is usually to hide the zoning maps and back away slowly. But when the proposed solution is to attack design, manufacturing, and on-site assembly as one system, that begins to sound less like TED Talk garnish and more like an actual operating thesis.

Why investors keep falling for “physical AI,” and why this time I get it

SuperSeed’s public writeup is refreshingly direct about the bet. The firm says the real opportunity is bringing AI into the physical world, and that construction is a perfect example of a huge sector where productivity has barely budged. It also praises founders Rodion Shishkov and Slava Bocharov for building Samokat into a $1.5 billion exit and for applying the same automation discipline to construction. In another SuperSeed post, the fund calls All3 exactly the kind of company it built its new fund for: “Physical AI” aimed at the real economy, not just more software nibbling around the edges.

“Physical AI” is a phrase with a very high conference-to-substance ratio, so I do not surrender to it lightly. But if you want to sell me on embodied intelligence, “build apartments faster on weird urban plots” is dramatically better than “my laptop now has ten autonomous interns and a mood board.” We have already spent enough time asking whether AI agents actually produce money or just attractive screenshots. All3 is making a more grounded wager: if software can make design and fabrication tighter, and robotics can remove on-site bottlenecks, maybe the unit economics of building real homes improve enough to matter.

That is at least a grown-up fantasy.

The part where I ruin the pitch deck a little

Now for the loving skepticism. Construction is the graveyard of elegant diagrams. Every bold construction-tech startup eventually has to negotiate with weather, regulation, financing cycles, supply chains, neighborhood politics, union realities, site access, inspectors, and the ancient universal law that the thing delivered to a job site is never quite the thing everyone thought they approved. A startup can have brilliant robotics and still get tackled by one missing document and a miserable Tuesday.

There is also a lot packed into this model. All3 is not just building software. It is not just making robots. It is not just selling prefab systems. It wants to coordinate design generation, compliance logic, component fabrication, manufacturing capacity, and on-site assembly into one coherent machine. That is ambition bordering on imperial. It is also the kind of ambition that can become genuinely defensible if it works, because partial solutions are exactly what construction has been drowning in.

So yes, there is a beautiful-overreach risk here. But it is the good kind, the kind rooted in noticing that the industry’s inefficiency is systemic and deciding not to pretend a thin workflow layer will fix it. I would rather watch founders overreach in the direction of integrated housing delivery than in the direction of “AI for calendar synergies” or yet another startup promising robot coworkers for knowledge workers.

Verdict: a promising little rocket, wearing steel-toe boots

My verdict is that All3 feels like a promising little rocket with unusually heavy hardware attached. The company has the right kind of early-stage weirdness: not weird because the market is imaginary, but weird because the founders are trying to compress three separate industries into one integrated system and politely call that a startup. That is deranged in the traditional venture-backed sense. It is also kind of great.

I do not know if Mantis becomes the symbol of a new construction stack or simply the most photogenic part of a very difficult operating business. I do know this: if you are going to raise a seed round in 2026, it helps to show up with something more persuasive than a chatbot for PDFs. A robot that might actually help build homes faster? That, at minimum, earns my attention and a cautiously affectionate eyebrow raise.

All3 may yet discover that housing is harder than robotics, and robotics is harder than venture math, and venture math is already one of humanity’s least stable sciences. But there is a real problem here, a real thesis, and enough technical specificity to keep this from collapsing into pure mood-board capitalism. In Silicon Valley terms, that qualifies as almost suspiciously wholesome.