HL Series
The Hailo-8 edge AI processor delivers up to 26 TOPS of computing performance, providing a significant performance advantage over conventional edge AI processors. It also offers outstanding power and area efficiency, exceeding that of major competing solutions. Even with all required memory integrated, the device can be implemented in a form factor smaller than a coin.
Thanks to its dedicated architecture optimised for neural networks, the Hailo-8 enables large-scale deep learning applications to run more efficiently and reliably on edge devices. This approach allows system designers to achieve high AI performance while reducing overall system cost, making it an ideal solution for demanding edge AI applications.
Most AI accelerators rely on external DRAM to provide the memory bandwidth and capacity required for neural network processing, which increases both system cost and design complexity.
The Hailo-8 adopts an architecture in which all required memory is fully integrated on the processor die. This design delivers the high and deterministic bandwidth needed for AI inference while reducing system cost and simplifying hardware design
Independent of DRAM price fluctuations, the Hailo-8 is capable of delivering up to 26 TOPS of AI performance without external memory, ensuring both high reliability and cost efficiency for edge AI systems.