Exclaim Robotics has raised $4.95 million to develop autonomous machines for routine maintenance inside AI data centres, targeting the physical repair work that remains manual even as most computing operations are managed remotely.

The Zurich company’s pre-seed round was co-led by Founderful and Playfair Capital. It will finance early prototypes, robotics hardware, and recruitment as Exclaim begins developing machines that can navigate a data hall, locate an equipment rack, and work across its front face without requiring the building to be redesigned around the robot.

Its initial tasks are deliberately narrow: cleaning and replacing optical connectors and small form-factor pluggable transceivers, managing copper network cables, and replacing failed drives. These are repetitive jobs, but they demand precise handling in spaces packed with cables and sensitive equipment, where an incorrect movement can interrupt services or damage expensive hardware.

Founder and chief executive Helen Oleynikova has spent 15 years working in robotics at organisations including Willow Garage, Google, Microsoft, Nvidia, and ETH Zurich. She co-created the nvblox mapping and collision-avoidance library used within Nvidia’s Isaac ROS platform and is joined by founding engineer Patrick Pfreundschuh, whose research has included LiDAR navigation and autonomous systems for difficult environments.

Automation reaches the physical rack

Data centres already automate provisioning, failover, scheduling, monitoring, and network configuration, but replacing a failed component still generally requires someone to enter the hall and touch the equipment. Exclaim is betting that higher rack densities, more demanding cooling systems, and changes in power distribution will make that manual exception more expensive and hazardous.

The company expects future AI racks to use 800-volt direct-current distribution, placing technicians close to equipment with a greater arc-flash risk. Oleynikova said that maintaining such systems by hand “gets more dangerous every year”, although the pace and breadth of that transition will vary among hardware suppliers and operators.

Exclaim plans to train its robots in simulation before repeating tasks on physical equipment. Where the system lacks sufficient confidence, it is intended to stop and transfer control to a remote human operator rather than attempt the repair autonomously.

That hybrid model is likely to be essential in early deployments. Data-centre racks may follow broad standards, but cabling, labelling, component placement, and maintenance procedures differ among sites, while any robot operating near live production equipment will need clear controls for safety, access, audit records, and emergency intervention.

An early company in a difficult market

Robotics is already entering the construction side of the data-centre industry, where machines drill concrete, mark layouts, inspect structures, and move materials. Exclaim is pursuing a harder operating environment because its equipment would work after a facility has gone live, when interruptions carry direct commercial consequences and every action must coexist with security and availability requirements.

Its emergence also adds another specialist to Europe’s expanding robotics market, where companies are applying autonomy to manufacturing, logistics, inspection, infrastructure, and hazardous work. A technically experienced founding team can reduce research risk, but it does not remove the long timelines and capital requirements associated with producing reliable hardware.

Exclaim has not disclosed a completed prototype, paying customer, trial site, revenue, or timetable for commercial deployment. Two further founding engineers are expected to join in 2027, which underlines how early the business remains despite the specificity of its target market.

Data-centre operators will ultimately require evidence that a robot can identify the correct component, manipulate it without damaging neighbouring equipment, recover safely from errors, and integrate with existing maintenance and access systems. Labour savings alone will not justify deployment if the machine creates a new source of downtime or needs frequent human supervision.

The company also argues that robotic maintenance could permit warmer data halls, reduce physical access, and support facilities in remote locations. Those benefits depend on the robot becoming reliable enough to remove routine human presence rather than merely accompanying technicians.

Exclaim’s funding buys the engineering time needed to answer those questions. Its proposition starts from a genuine operational gap — software cannot replace a failed drive or reseat an optical connector — but the commercial threshold will be much higher than proving that a robot can complete the movement once. A more useful milestone will arrive when it can repeat the work safely inside an operating facility whose owner is prepared to trust it.