All3 raises $25M to deploy construction robots on live German sites

A robot that carries over 100 kilograms, reaches four meters high, and never gets tired or demands a union contract just became real on a construction site in Germany. On April 29, 2026, All3, a London-founded robotics startup with $25 million in fresh seed funding, announced that its Mantis autonomous legged robot will deploy on active multi-family residential construction sites across Germany starting later this year. This is not a pilot program or a controlled lab demo. All3 has already processed over 100,000 square meters of residential projects through its AI-powered design software and has customers lined up. The startup claims that a fleet of Mantis robots can complete a seven-story, 5,000-square-meter building in roughly three months, with cost savings of up to 30%, timeline cuts of up to 50%, and 25% less embodied carbon versus traditional methods.

Why this matters: construction is the world's largest industry by value at $6.7 trillion annually, and it has barely improved productivity in 50 years. That is not hyperbole. A building site in 2026 still moves like one from 1976, the same mix of skilled trades, hand-offs, scheduling chaos, and dependency on labor that cannot scale. Germany alone faces a shortage of roughly 700,000 homes, and the labor constraint is the binding constraint. Every other major manufacturing sector has experienced a productivity revolution. Construction has not. All3 is betting it can change that, and the round structure suggests investors believe the team can actually execute. RTP Global led the round, with SuperSeed, Begin Capital, s16vc, and VNV Global also investing. A $25 million seed is disciplined for a hard-tech company with hardware, software, and factories, it implies the investors expect near-term revenue, not years of R&D before commercial viability.

The specifics matter. Mantis is a legged robot designed to handle installation, fastening, finishing, and inspection tasks using what All3 calls 'embodied AI.' The robot can carry payloads exceeding 100 kilograms and extend to four meters in height, which puts it into the range of human-height reaching tasks, drywall installation, electrical roughing, HVAC placement, the middle-tier labor that is expensive and hard to recruit. All3 plans to deploy these as fleets across its first commercial projects, coordinating multiple robots on the same site working concurrent tasks. The technical challenge is substantial: real construction sites are chaotic. Material placement is irregular, worker presence is dynamic, and tolerances matter. The two arXiv preprints cited in All3's research context, OmniRobotHome on multiadic human-robot interaction and the Science Robotics paper on foundation models for robotic deployment, reflect the actual technical problems All3 must solve to move from one-robot demos to five-robot coordination on a live construction site.

What made this possible right now: All3 was founded in 2023 by Rodion Shishkov and Slava Bocharov, the co-founders of Samokat, a dark store rapid delivery platform that became Russia's largest food delivery company and sold for $1.5 billion within five years. That track record is not footnote material. Samokat required solving real-time inventory coordination, last-mile logistics, and unit economics in a capital-constrained market, all under deadline pressure with a window to prove it before competitors shut you down. Both founders brought that operational discipline to All3. They chose timber composite structural systems as the material strategy, positioning themselves as a lower-carbon alternative to concrete, which accounts for 7–8% of global emissions. That is not virtue signaling. It means their robots are fabricating and assembling a different building system than competitors like ICON (concrete 3D printing for low-rise) or Fastbrick Robotics (bricklaying automation for single-family homes). By redesigning the entire construction sequence around what robots can do, not retrofitting robots onto existing construction methods, All3 created scope of differentiation that is harder to replicate than a single automation step.

Who wins and who loses: Germany's construction labor shortage is real, and All3's timing is precise. The country faces an acute housing crisis, and traditional labor recruitment will not solve it. General contractors and developers in Germany have an immediate economic incentive to adopt anything that shortens timeline and reduces cost. All3's claim of 30% cost savings and 50% timeline reduction would be worth millions on a single 5,000-square-meter project. That is not vaporware math, that is project margin. The losers are less obvious but significant: construction trades that depend on physical labor for residential assembly (electricians, drywall installers, HVAC technicians in the middle-tier task stack) will face displacement in markets where All3 can scale. The broader question is whether All3 can actually deliver on the claimed performance benchmarks at the cost structure it implies. Legged robots are mechanically more complex than wheeled platforms, and coordinating fleets in real-world environments without purpose-built infrastructure is genuinely hard. The fact that All3 has signed customers and processed 100,000 square meters through its design software suggests the team has moved past the 'can we build it' phase, but the 'can we do it reliably at the cost we quoted' phase is where most hard-tech startups fail.

The real read: All3 is the strongest commercial robotics play I have seen in eighteen months because it solves a trillion-dollar problem (construction productivity) with a team that has already proven it can scale an operationally complex business in a capital-constrained environment (Samokat at $1.5B exit in five years). The $25 million seed is not a validation bet, it is an execution bet. The investors (RTP Global, SuperSeed, Begin Capital) are backing a team with a product in the field and customers paying, not a technology bet on whether robots can work on construction sites. The real risk is not whether Mantis works in controlled trials. It is whether the cost structure holds when you are deploying fleet-coordinated robots at scale on irregular, chaotic real-world sites with human labor still on-site. If All3 can hit 50% timeline reduction and 30% cost savings repeatedly across a dozen projects in 2026 and 2027, the unit economics alone will attract capital at a velocity that most robotics companies never see. If the benchmarks slip by 20%, the entire investment thesis craters because construction is a thin-margin, low-velocity business where timeline and cost are the only language that matters. I would watch for the first public completion milestone rather than investor announcements.

What to watch: (1) Completion timelines and cost variance on the first three commercial Mantis deployments in Germany versus the estimated three-month / 5,000-square-meter benchmark, actual data, not marketing claims. If All3 delivers two projects on time and under the claimed cost structure by Q4 2026, the competitive response from Skanska, Bovis, and other large GCs will be immediate. (2) The scale of the 2027 pipeline: All3 claims 100,000 square meters processed through design software already; watch whether that translates to 500,000+ square meters of signed contracts or LOIs by mid-2027. That would indicate real demand beyond the Germany launch market and would justify a Series A at a much higher valuation. (3) Regulatory or labor union resistance in Germany or other EU markets, construction trades are heavily unionized, and a robot that eliminates 30-50% of assembly labor will face political pushback. If All3 encounters significant union or regulatory friction in 2027, that may limit scaling speed even if the technology works perfectly.