3E8 leverages light and optics to build the next-gen AI accelerator. By using on-chip light processing and programmable photonic memories, neural networks computations can be accelerated by several orders of magnitude for the same energy consumption and chip cost.
In pursuit of 1000x
At 3E8 we pursue 100-1000x improvement in AI computing speed and energy efficiency. Fiber optics has transformed telecom over the past decade and enabled a 1000x improvement in communication bandwidth. Photonics is capable of achieving a similar leap in AI computing for the exact fundamental reasons.
3E8 develops lightspeed photonic AI silicon, systems and software to achieve this leap in AI compute efficiency.
3E8 builds on decades of research at the George Washningtom University and brings together several innovative photonic technologies that, collectively, offer a unique solution to the problem of increased energy consumption of AI computations.
3E8 photonic memory and digital to optical conversion technologies enable lightspeed matrix multiplication operations with superior energy efficiency. Combined together, these technologies enable a computational efficiency of more than 50 TOPS/W.