$2.7m · Pre-Seed · Software · London, UK
London-based Embedd, which builds software infrastructure for semiconductor integration, has raised $2.7m in pre-seed funding led by Seedcamp. Cocoa, Connect Ventures, 2100 Ventures, Vesna Capital, U.ventures, Underline Ventures, Common Magic and Roosh Ventures also participated, according to EU-Startups. The company says the capital will support continued platform development and wider work with semiconductor companies.
Turning chip documents into production code
Before a processor or peripheral can be useful inside a robot, vehicle or medical device, engineers need low-level software that lets the wider system communicate with it. That work typically means interpreting long datasheets, mapping registers and pins, writing drivers and board-support packages, then debugging them against real hardware. Changes in components or operating systems can force teams to repeat part of the integration.
Embedd turns chip documentation into a structured digital twin of the hardware. Its device-tree configurator models peripheral maps, registers and pin assignments, while its driver-generation product produces code for the chosen architecture and operating-system target. The company says this can move production-ready integration from months to weeks; EU-Startups reports Embedd's claim that customers have delivered chip software up to six times faster.
The commercial mechanism is not simply selling another coding assistant. Semiconductor vendors need their chips to work across more software ecosystems, while device manufacturers want to change components without restarting the low-level stack. A reusable hardware model can shorten that enablement work on both sides. Since a commercial launch in April 2026, Embedd says it has signed multiple semiconductor companies, including Microchip Technology for Zephyr support.
Determinism is part of the product
Embedd's official product material emphasises deterministic generation: code is produced from a defined hardware model, with repeatable and inspectable output. That distinction matters for embedded systems where engineers need to review behaviour, reproduce builds and support functional-safety workflows. AI helps structure the source material, but the sellable result has to behave like engineering infrastructure rather than an unpredictable code suggestion.
The execution challenge is coverage. Each additional chip family, device class and target environment expands the platform's usefulness, but also creates more models to validate and maintain. Partnerships with semiconductor companies can supply technical access and distribution, yet Embedd will still need to show that its library grows without turning each integration into a bespoke engineering project.
The product came from a problem the founders had lived through. EU-Startups reports that Michael Lazarenko, Maxim Gorinov and Valentin Gololobov repeatedly rewrote software for replacement components after chip shortages and disruption from Russia's invasion of Ukraine affected their previous hardware company. The new round gives Embedd room to turn that experience into a broader platform; adoption across more production programmes will show whether the digital-twin approach can become a common layer between chips and physical-AI applications.
