Lightmatter
Develops photonic processors and interconnects for AI data centers
Company facts
- Founded
- 2017
- Employees
- 315
- Headquarters
- United States
- CEO
- Nicholas Harris, Darius Bunandar
- Type
- Private
- Funding
- Series D · $400M · 2024
About
Lightmatter is a pioneering company in the field of photonic computing, developing specialized chips and chip communication technology, known as interconnects, that leverage light for data processing and communication. Founded in September 2017 by Nicholas Harris, Darius Bunandar, and Thomas Graham, Lightmatter aims to revolutionize computational power and energy efficiency, particularly for artificial intelligence (AI) workloads and data-intensive applications. The company's mission is to accelerate computing advancements to address significant global challenges while minimizing the environmental impact associated with traditional computing. As silicon-based chips approach their physical and economic limitations, Lightmatter's photonic technology offers a path forward by enabling faster data transport and reduced energy consumption, which is crucial for the infrastructure needed for large-scale chip interconnection in modern data centers. Lightmatter's core product offerings include Envise, a photonic AI accelerator, and Passage, a photonic interconnect platform. Envise is designed to handle the massive matrix multiplications required in deep learning, offering a more energy and cost-efficient solution for large AI workloads compared to traditional electronic GPUs. Passage provides a comprehensive photonic interconnect solution that integrates with existing data center infrastructure to enable higher bandwidth data movement and reduced power consumption for chips. This technology is designed to allow for the scaling of AI infrastructure by providing dense optical interconnects between chiplets, complete with integrated lasers and switching for dynamic network reconfiguration. The company's Idiom software stack complements its hardware by enabling customers to maximize performance on Envise through neural network model compilation, execution, debugging tools, profiling capabilities, and model optimization features. Lightmatter's innovations are positioned to address the increasing demand for energy-efficient and cost-effective data processing solutions from data centers, technology companies developing AI applications, and cloud service providers.