Iceland's Treble raises $18M to simulate sound for physical AI and voice devices

Reykjavík-based Treble Technologies has raised $18 million, about €15 million, in a Series A-2 round to push its sound-simulation platform deeper into voice AI and physical AI. Paladin Capital Group led the round, with participation from existing investors KOMPAS VC, Frumtak Ventures, and the European Innovation Council Fund, Treble said in a September 17 GlobeNewswire release. The company says the financing brings total funding to date to €36 million. TechCrunch, writing a day earlier, also listed Omega ehf among participants and said Treble counts Amazon and Logitech as customers.

Acoustic twins for a world that still sounds messy

Treble builds physics-based models of how sound moves through rooms, materials, vehicles, and devices, then turns those models into digital acoustic twins and labeled synthetic datasets. The pitch is that speech and audio models that look strong in a quiet lab fall ap Wearables, head-mounted devices, humanoid robots, and autonomous vehicles make that harder still because the listener and the noise field keep moving. Paladin vice president François Ruether told TechCrunch that as more products depend on understanding sound, shared simulation infrastructure becomes more valuable across voice AI, wearables, robotics, and physical AI, while customers keep ownership of their models and workflows.

From consumer audio into robots that can hear around corners

Founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen, Treble previously raised about $12 million in 2024, according to TechCrunch. The new capital is meant to extend the platform across consumer and enterprise audio products and to increase focus on physical AI, including robotics, automotive, and drones that need sound-based functions rather than vision alone. Ruether's GlobeNewswire comments go further: robots that only see what is in front of them miss critical cues such as a fall in another room or a crash around a corner, and Treble's simulation stack is meant to help models fuse visual and auditory inputs.

Treble has also partnered with Hugging Face on a far-field automatic speech recognition leaderboard that uses simulated homes, offices, and classrooms rather than clean lab audio, a signal that the company wants its synthetic environments to become industry test infrastructure, not only a private training tool.

The odd angle is what Treble is selling: not another voice model, but the acoustic weather system those models train against. In a market obsessed with larger multimodal brains, an Icelandic simulation company is raising on the bet that physical AI will stall without better fake rooms, fake noise, and fake microphones before anything ships in silicon.art when reverberation, competing talkers, moving microphones, or unfamiliar rooms show up. Instead of months of physical prototype testing and field recording, Treble argues teams can generate hard scenarios in days and compare product configurations under controlled conditions.

Co-founder and CEO Finnur Pind framed the problem in the Tech.eu write-up: audio and voice are becoming a core interface for intelligent products, but the physical world is acoustically complex.