This London Startup Is Solving AI's Biggest Physical Bottleneck

Jan de Vries · 2026-08-24

You know how everyone's talking about AI taking over the physical world? Factories, robots, self-driving cars… it sounds like the future, right? But there's a massive problem hiding underneath all that promise, and a London-based startup called Embedd just raised $2.7 million to tackle it head-on. Let's break it down. The next wave of AI isn't just about chatting with a bot. It's about powering actual machines. The problem is, every time a hardware engineer tweaks a chip or adds a new sensor, software teams have to start almost from scratch. It's a nightmare of complexity, and it's slowing everything down. Embedd's founders, Michael Lazarenko, Maxim Gorinov, and Valentin Gololobov, know this pain intimately. They're Ukrainian tech entrepreneurs who built a hardware company, only to watch it get hammered by the global chip shortage and then disrupted by war. They kept hitting the same wall: every new component meant rewriting huge swaths of software. They realized they'd stumbled onto a bottleneck that's about to get a whole lot worse. ### The Universal Language Problem for Smart Machines Think about any intelligent machine—a robot, a smart factory line, an autonomous vehicle. Before it can do a single thing, the software running it needs to talk to dozens, sometimes hundreds, of different chips. Sensors, processors, controllers… you name it. And here's the kicker: none of them speak the same language. It's like trying to run a meeting where everyone shows up speaking a different dialect. Engineers traditionally have to become translators. They spend weeks buried in thousands of pages of technical documentation, hand-writing code just to get these components to introduce themselves to each other. It's tedious, error-prone, and it sucks up insane amounts of time and talent. Embedd's solution is elegantly simple, yet incredibly powerful. They're building the universal translator. ### How Embedd's Digital Twins Cut Through the Chaos Instead of armies of engineers manually bridging the gap, Embedd creates a **digital twin** of the hardware. They feed this perfect digital replica to their AI agents, giving them the full context of how every chip and component works. The AI then handles the integration automatically, generating all the necessary code so the hardware can seamlessly plug into the software ecosystem above it. - It automates the grunt work that bogs down development. - It reduces human error in critical integration layers. - It turns a process that can take months into something that takes days. The results they're reporting are pretty staggering. They claim to help customers deliver production-ready software for new chips up to **six times faster**. Since their commercial launch, they've already locked in contracts with several semiconductor players, including a notable one with Microchip Technology to enable Zephyr RTOS support. As CEO Michael Lazarenko put it: *"The promise of physical AI is enormous, but today's hardware fragmentation is slowing innovation. This funding enables us to expand our platform and help more semiconductor companies bring their devices into emerging software ecosystems."* The $2.7 million pre-Seed round, led by Seedcamp with a crowd of other top-tier funds like Connect Ventures and Cocoa joining in, is a strong vote of confidence. It signals that smart money sees this bottleneck as a critical chokepoint for the entire physical AI revolution. So, what's the big picture here? Embedd isn't just another AI tool. They're building the foundational plumbing—the software infrastructure—that will allow the physical AI world to actually get built, and built fast. By removing this fundamental friction, they're not just helping semiconductor companies; they're accelerating the arrival of the intelligent, automated world we've been promised. It's a classic case of solving a painful, unsexy problem to unlock something truly transformative. And now, with fresh capital and a clear mission, they're charging ahead to connect the chips to the future.