Bolt a camera above an assembly line, point it at the work rather than the worker, and you can watch a factory lose time it did not know it was losing. That is the entire premise of Almetra, the Berlin company that just raised €16.3 million in a Series A, and the receipts it points to are unusually concrete for an early-stage round: a Bosch e-bike subsidiary reported output up 19% within weeks of switching the cameras on, and ABB reported a 15% productivity gain. The pitch is that most plants already know performance is short of where it should be, and have no objective way to say why.

The round was led by blisce/, the transatlantic fund behind early bets on Spotify and Pinterest, with NAP, Merantix Capital, Robin Capital, Underline and Critical Ventures alongside a group of angels. The money funds product work, a push into the United States, and the build-out of the platform from a passive observer into something that can act on the floor. Almetra was founded in 2022 as Deltia, out of the Berlin AI venture studio Merantix, and rebranded this year to mark a shift in ambition: where Deltia analysed manual production steps, Almetra wants to be the data and control layer for the whole factory, pulling together video, machine data, IT systems and what the operators already know.

The wedge is a camera because a camera asks for nothing

The mechanism is deliberately light. The cameras process video locally and convert it into structured production data, cycle times, output rates, equipment utilisation, without plugging into a plant’s IT systems at all, which is the part that matters to anyone who has watched a multi-year MES rollout stall. Founder and chief executive Maximilian Fischer, who built the company with Silviu Homoceanu, frames the problem plainly: plant teams know they are losing capacity but not where or why, and the platform is meant to surface the optimisation opportunities within the first few weeks rather than the first few quarters. No sensors to retrofit, no stopwatch studies, no engineer standing on the line with a clipboard.

The privacy question is the one that decides whether any of this is allowed in a German plant, and Almetra has clearly built the answer into the product rather than bolting it on. Raw footage is anonymised and, by the company’s account, discarded almost immediately, most data never leaves the site, and only short randomised snippets are retained for root-cause analysis. It tracks processes, not people, with no individual performance ranking, and pitches the works council as an ally rather than an obstacle to be managed. In a country where the works council holds real co-decision power over workplace surveillance, that is not a feature, it is the price of entry, and a rival that gets it wrong does not get a second meeting.

The market underneath is large enough to attract company. The global machine-vision market was valued at about $20.4 billion in 2024 and is projected to reach $41.7 billion by 2030, and Almetra is far from alone chasing it. Sight Machine, which has raised more than $85 million, embeds deeper and targets larger enterprise deployments, while Augury focuses on machine health through vibration and ultrasound sensors. Almetra’s differentiation is the no-integration camera, the argument that the cheapest and fastest way into a factory is the one that asks the IT department for nothing.

The 20% is an average, and the robots are still a promise

It is worth reading the headline numbers carefully, because the company itself reports them carefully. The round 20% productivity figure is best understood as a convenient average of plant-level results that vary by site, ABB at roughly 15%, Viessmann reporting a 20% cut in cycle time, the Bosch e-bike line at 19%, rather than one measured constant. The customer roster is the real signal: Bosch, Siemens Energy, ABB, Viessmann and Aumovio, the unit formerly known as Continental, are not names an early-stage company wins by accident. Acceptance into Google DeepMind’s Robotics Accelerator and a Physical AI fellowship run by AWS, Nvidia and MassRobotics says the same thing in a different language, that a company sitting on continuous video of live factory floors is a plausible data source for the next generation of industrial robots.

That robotics ambition is exactly where the skepticism should sit. Almetra is around 40 people, the leap from telling a plant manager where the time goes to actually closing the loop with robotics is large, and the gap between a productivity dashboard and a control layer is the gap most industrial-AI companies fail to cross. The camera that gets you in the door is also a camera that, once inside, has to justify becoming the system everything else runs through. Plenty of monitoring tools have stalled as expensive ways to confirm what a good floor manager already suspected.

Step back and the geography is the point. Germany is Europe’s manufacturing backbone, and most of its plants still make decisions on incomplete data and end-of-shift output numbers, which is precisely the gap a wave of German industrial-AI money is now chasing, from NEURA Robotics’ very large physical-AI round to Sereact and SPREAD. The bet across all of them is that the data and control layer of the European factory is a category worth owning, and that the company holding the cameras is well placed to own it. The camera was always the cheapest way into the factory. Almetra is wagering it is also the way to everything else inside.