Optical Interconnect Startup Quintessent Raises $40M for AI Cluster Lasers
Quintessent secured $40M in Series A funding to ramp up production of its quantum-dot comb lasers for AI data center networks.
Data center operators link AI chips in servers using fiber-optic cables, through which data travels as light. Conventional network lasers generate that light with microscopic structures called quantum wells, typically nanosheets made of two materials with differing optical properties. Quintessent’s device, called a comb laser, instead uses quantum dots — spherical nanostructures made of a single material — which the company says are more durable and efficient than existing light sources.
Quantum dots are commonly used in high-end televisions. When hit by ultraviolet light, electrons in the dots gain energy and then release light as they return to their original energy level. Quintessent makes its quantum dots from gallium arsenide, a material through which electrons move several times faster than through silicon, enabling quicker switching between energy levels and higher-speed laser pulses that boost network performance.
The company says its comb laser requires fewer supporting components than rival products, reducing costs, and uses 40% less power while operating more reliably at high ambient temperatures. It can generate light beams at eight different wavelengths, increasing fiber-optic link throughput, and the technology can be extended to support more wavelengths and spectrum ranges.
“For the past several years, we have focused on building technologies designed to make optical connectivity fundamentally simpler to deploy and scale,” said Quintessent co-founder and Chief Executive Officer Alan Liu. “With today’s news, Quintessent is starting its transition from tech development to a product-focused company.”
The funding will be used to ramp up manufacturing of its comb laser and make reliability improvements. The company also plans to develop additional optical products, including signal amplifiers to improve connection reliability.