Biio
DTU-based startup (founded 2025) building digital in-memory-compute chips for energy-efficient edge-AI inference, initially targeting medical devices. Danish Tech Challenge 2025 participant.
Semiconductor activity in Denmark
Biio (Biio ApS, founded 2025, based at DTU in Kongens Lyngby) develops edge-AI inference silicon built on digital in-memory compute: weights are stored where the arithmetic happens, removing the off-chip memory bottleneck while staying bit-exact and deterministic (no analog drift or spiking variance). The architecture targets standard CMOS nodes (7/14/22 nm) without exotic memories, and scales from 0.2 to 60 TOPS and from 4 million to 1 billion parameters — wearables through robotics — with a single ML software stack. First applications are on-device analysis for medical devices (cardiac, motion and neurotechnology) without cloud dependence. The team cites prior tenures at AMD, Intel and Demant. Selected for DTU Science Park's Danish Tech Challenge 2025 hardware programme.
Key facts
- Founded 2025, based at DTU, Kongens Lyngby
- Digital in-memory compute for edge-AI inference
- 0.2–60 TOPS, 4M–1B parameters, wearable → robotics
- Initial focus: medical devices (cardiac, motion, neurotech)
- Danish Tech Challenge 2025 (DTU Science Park)
- Director Satyam Goel; CVR 45506983
Technology focus
- Digital in-memory compute (bit-exact, deterministic)
- Edge-AI inference accelerators
- Standard CMOS (7/14/22 nm), no exotic memories
- Ultra-low-power medical and wearable devices