HAETAE (Heterogeneously Integrated Multi-material Photonic Chiplets for Neuromorphic Photonic Transfer Learning AI Engines) is a three-year collaborative research program established under the EU–Korea Digital Partnership.

The program is dedicated to the development of an advanced photonic neural-network computing architecture enabled by the heterogeneous integration of multiple materials, including silicon, indium phosphide (InP), silicon nitride (Si₃N₄), and SiGe.
Key technological components comprise micro-transfer printing (uTP), multi-chiplet bonding methodologies, low-loss photonic circuit engineering, and non-volatile MEMS-based weight elements. Together, these technologies facilitate the realization of a photonic computing platform specifically optimized for transfer-learning-oriented large-scale AI models.

The HAETAE platform demonstrates ultra-low energy consumption (18.5 fJ/MAC), exceptional area efficiency (2.45 TMAC/s/mm²), and high computational throughput (4.1 PMAC/s).

Through validation across multiple application domains—including LLM-based AI inference, cybersecurity threat analytics, and optical-communication digital signal processing—the program substantiates the feasibility and future potential of next-generation photonic AI computing systems.