Lightmatter, a startup based in Mountain View, California, has unveiled new photonics technology aimed at enhancing AI chip performance. Their approach uses optical connections and silicon photonics to transmit information using light, rather than electrical signals.
The quest for faster, more energy-efficient computing has led scientists to explore photonic chips – devices that use light instead of electricity to process information.
Recent advancements have addressed significant challenges in this field, bringing us closer to integrating photonic chips into mainstream artificial intelligence (AI) applications.
The Promise of Photonic Computing
Traditional electronic chips are reaching their physical and performance limits, especially with the increasing demands of AI workloads.
Photonic chips offer a compelling alternative by leveraging photons to transmit and process data, potentially achieving higher speeds and lower energy consumption. This approach is particularly advantageous for AI tasks that require massive parallel processing and high bandwidth.
Recent Breakthroughs
Lightmatter’s Innovations
Lightmatter, a startup based in Mountain View, California, has unveiled new photonics technology aimed at enhancing AI chip performance. Their approach uses optical connections and silicon photonics to transmit information using light, rather than electrical signals.
This technology includes an interposer for connecting AI chips and a “chiplet” to augment AI chips, with production expected in the coming years.
Additionally, Lightmatter introduced a new type of computer chip that uses beams of light to accelerate AI computations and reduce energy consumption. This chip addresses limitations in traditional chip design by employing photonic technology to process data using light, avoiding the scaling challenges of electronic transistors.
TDK’s Spin Photo Detector
Japanese electronics company TDK has announced a significant breakthrough with the development of the world’s first “spin photo detector.”
This device merges optical, electronic, and magnetic elements to achieve response times of just 20 picoseconds, far outpacing current semiconductor-based photo detectors.
TDK plans to distribute samples by March 2026 and targets commercial production within three to five years.
Large-Scale Optical Neural Networks
Researchers have developed a large-scale optical neural network (ONN) featuring over 41 million photonic neurons on a metasurface.
This system surpasses digital electronics in speed and energy efficiency and closes the performance gap with large-scale AI models.
The ONN can process tens of millions of weights in a single operation, demonstrating high-performance solutions for real-world AI challenges, such as accelerating cancer detection analysis.
Industry Impact and Future Outlook
The photonic integrated circuits (PICs) market is projected to surpass $54 billion by 2035, driven by the growing demand from AI data centers, telecommunications, quantum computing, and sensing applications.
Companies like STMicroelectronics are introducing new silicon-germanium PIC production processes in collaboration with major cloud service providers, highlighting the industry’s shift toward photonic solutions.
Co-packaged optics (CPO), which involves integrating photonics directly with electronic components, is gaining traction to enhance performance and energy efficiency. This approach reduces signal loss and power consumption associated with traditional interconnects, making it particularly beneficial within AI data centers.
The recent advancements in photonic chip technology mark significant progress toward integrating light-based computing into mainstream AI applications.
