Sustainable AI Hardware Platforms

An IEEE study highlights a new and more sustainable approach to powering artificial intelligence (AI) systems by using light-based technology to reduce energy consumption. As AI becomes more widely used across industries, the demand for powerful, energy-intensive computing continues to grow. Traditional systems rely on graphics processing units (GPUs), which are effective but require substantial energy to operate.

The study, published in the IEEE Journal of Selected Topics in Quantum Electronics and led by Dr. Bassem Tossoun of Hewlett Packard Labs, introduces "photonic integrated circuits (PICs) as an energy-efficient alternative; these circuits use optical neural networks (ONNs), which transmit data using light rather than electricity, allowing for faster processing with minimal energy loss."

To build the platform, researchers used "a combination of silicon photonics and III-V compound semiconductors, enabling integrated lasers and amplifiers that boost performance and scalability."

Image Credit: