Treble raises $18M to give physical AI a sense of hearing

Treble (Reykjavík) has closed $18 million (~€15 million) in a Series A-2 led by Paladin Capital Group, with existing backers KOMPAS VC, Frumtak Ventures, and the European Innovation Council (EIC) Fund reinvesting. The round, announced 17 September 2026, brings total funding to approximately €36 million for the acoustic simulation company founded in 2020 by acoustic engineers Finnur Pind (CEO) and Jesper Pedersen.

Treble's platform does one thing: simulate how sound behaves in physical spaces with physics-level accuracy. That sounds narrow. It is not.

What acoustic simulation is actually for

The use case that built the company was professional audio — architects and acousticians using Treble to model how a concert hall or open-plan office would sound before anything was built. Those customers remain. But the round reflects something larger: robotics and physical AI have a hearing problem.

Visual AI has had synthetic training data for years. LiDAR simulators, photorealistic rendering engines, synthetic image pipelines — companies build robots that can see in simulated worlds before they ever touch a real one. Audio has lagged. A robot operating in a warehouse needs to understand the acoustic environment it inhabits: the sound of a conveyor belt, the echo off a concrete wall, a human voice at three metres. Until simulated acoustic data is available at scale, that training either doesn't happen or happens slowly in the physical world.

Treble's platform generates that data. It builds acoustic digital twins — parametric models of physical spaces — and uses them to produce synthetic audio datasets at a cost and speed that physical recording cannot match. Amazon and Logitech are among its current customers; both are building voice-interface products that need to work in the full range of environments their users inhabit.

Finnur Pind: "Audio and voice interaction is becoming a fundamental interface for the next generation of intelligent products, but the physical world is acoustically complex."

From design tool to AI infrastructure

The Series A-2 label signals that this is an extension of Treble's 2024 Series A rather than a new stage. That matters: the company is not re-pitching its story, it is scaling a thesis already validated by paying customers. The €36 million total raised is a meaningful sum for an Icelandic deeptech company at this stage, and Paladin's lead is notable — the Washington DC-based fund specialises in dual-use technology and brings a customer network that extends across commercial and government environments where acoustic intelligence matters.

The US expansion is the stated priority. Treble's platform is already internationally adopted, but building dedicated commercial infrastructure in North America — where its largest technology customers are headquartered — is the next scaling step.

The broader physical AI sector is moving fast: Logibot, also closing today, is deploying humanoid robots in logistics; Magentic is running AI agents through industrial supply chains. Both ultimately depend on machines that can reliably interpret their physical environment. Treble is building part of the sensory stack that makes that reliable.