Lightmatter is a company specializing in photonic computing and optical interconnect technologies designed to accelerate Artificial intelligence and Machine learning workloads. Founded in 2017, the company develops hardware that uses light, rather than electrons, to perform computations and move data, aiming to address the energy and speed limitations of conventional electronic processors in data centers. Its technology targets applications involving Deep learning and Large language model inference and training, where data transfer and processing demands are exceptionally high.
The company's core premise is that as AI models grow in size and complexity, the energy required to power traditional semiconductor-based systems becomes unsustainable. Lightmatter's approach leverages photonics to reduce power consumption and increase computational throughput, positioning itself as an alternative to conventional chip architectures from companies like Intel, AMD, and NVIDIA.
History and Founding
Lightmatter was co-founded in 2017 by Nick Harris, Darius Bunandar, and Thomas Graham, who met while conducting research at MIT CSAIL. The founders sought to commercialize research on photonic computing, which had been explored in academic settings for decades. The company emerged from stealth in 2019 with a focus on building a photonic processor for AI inference.
In 2021, Lightmatter introduced its first major product line, Envise, a photonic accelerator for AI inference, and Passage, an optical interconnect technology for connecting chips and memory. The company has since expanded its product portfolio and raised significant venture capital funding, including a $154 million Series C round in 2021 and a $400 million Series D round in 2024, valuing the company at over $4 billion.
Technology and Products
Lightmatter's primary technology is based on silicon photonics, which integrates optical components onto silicon chips. The company's Envise processor uses photonic arrays to perform matrix multiplication, a fundamental operation in Neural network computations. By encoding data in light, the processor aims to achieve higher bandwidth and lower latency compared to electronic counterparts.
The Passage interconnect technology addresses the bottleneck of data movement between processors and memory. It uses optical waveguides to transmit data at high speeds, potentially enabling more efficient scaling of AI systems across thousands of chips. This is particularly relevant for training Transformer (architecture) models, which require massive parallel processing.
In 2024, Lightmatter announced a partnership with TSMC to manufacture its photonic chips, leveraging TSMC's advanced packaging and fabrication capabilities. The company also collaborates with Broadcom on optical networking components.
Applications and Market Position
Lightmatter targets data center operators and cloud service providers, including potential customers like Amazon Web Services, Google Cloud, and Microsoft Azure. Its technology is positioned for both AI training and inference, with a particular emphasis on reducing the energy footprint of large-scale AI deployments. The company claims its photonic processors can achieve orders of magnitude improvement in energy efficiency compared to electronic processors.
The market for AI accelerators is dominated by established players such as NVIDIA and Cerebras, but Lightmatter differentiates itself through its photonic approach. As of 2025, the company has not announced a commercial deployment of its Envise processor at scale, but it has secured partnerships with research institutions and is developing reference designs for data centers.
Challenges and Future Outlook
Photonic computing faces several technical challenges, including the difficulty of integrating optical components with existing electronic systems and the need for specialized manufacturing processes. Lightmatter's reliance on TSMC for fabrication is a strategic move to overcome these hurdles, but it also ties the company's production timeline to TSMC's capacity.
The company's success depends on the continued growth of AI workloads and the willingness of data center operators to adopt new hardware architectures. With the rise of Generative AI and OpenAI-style models, demand for efficient computing is expected to grow, potentially benefiting Lightmatter. However, competition from electronic accelerators, which continue to improve in efficiency, remains a significant factor.
Leadership and Team
Lightmatter is led by CEO Nick Harris, who co-founded the company and previously worked on photonic research at MIT. The leadership team includes experts in photonics, semiconductor manufacturing, and AI systems. The company has attracted talent from major technology firms and research institutions, including Xerox PARC and Nokia Bell Labs, reflecting its focus on advanced hardware innovation.
As of 2025, Lightmatter employs over 300 people and maintains offices in Boston, Massachusetts, and Fremont, California. The company continues to publish research on photonic computing and collaborates with academic institutions to advance the field.