Lightmatter founders Darius Bunandar, Nick Harris and Thomas Graham. [Image: Courtesy of Lightmatter]
Lightmatter, a start-up company seeking to commercialize a chip-based photonic deep-learning system pioneered at the Massachusetts Institute of Technology (MIT), USA, announced in late February that it had secured US$22 million in new series A funding. The new stake, from an investor group led by GV (formerly Google Ventures), brings the one-year-old start-up’s total series A capital raise to US$33 million.
Toward an all-optical AI architecture
Founded in January 2018, Lightmatter grew out of work at the MIT Research Laboratory of Electronics, led by OSA Senior Member Dirk Englund and colleague Marin Soljačić, to create an all-optical neural-net computing architecture centered around a chip-scale 2-D array of programmable Mach-Zehnder interferometers. The start-up’s founders include CEO Nicholas Harris, a co-lead author on the 2017 paper describing the work, chief scientist Darius Bunandar, an MIT Ph.D. candidate, and chief operating officer Thomas Graham, a graduate of MIT’s Sloane School of Business and a former Google employee. Englund serves as the company’s technical advisor.
The new US$22 million add-in to Lightmatter’s funding led by GV, the venture-capital arm of Google parent Alphabet, Inc., also includes two other VC participants, Matrix Partners and Spark Capital. As part of the deal, Tyson Clark, a partner in GV, has joined Lightmatter’s board.
Addressing the “post-Moore’s Law era”
Lightmatter’s business pitch to these investors and others lies in the possibility of neural-net computing to solve the bottleneck problems increasingly threatening Moore’s-law-type growth in the information arena. Harris struck this note in a post on the blog site Medium, which has become a favorite Silicon Valley platform for announcing new business developments.
“While integrated photonic systems are typically considered in the context of communication in datacenters, we’re leveraging them to address the growing demand for compute, largely driven by artificial intelligence, in the post-Moore’s law era,” Harris wrote. “If you’re going to pack more transistors onto the same sized chip, which has been happening for decades, those transistors need to be commensurately more energy efficient; herein lies the problem. At Lightmatter, we are addressing this scaling challenge in the context of artificial intelligence.”
From student competition to Silicon Valley
The Lightmatter engineering team has been busy advancing that work in the year since the company officially got off the blocks. In the post, Harris reported that, eight months after the company’s January 2018 founding, the firm had “taped out a chip with over a billion transistors.” (Tape-out, in IC-manufacturing jargon, refers to the point at which the photomask of the circuit has been fully designed and is ready to send to the manufacturer for production.)
A story on Lightmatter on CNBC.com notes that the company’s three founding executives, who met while Harris and Bunandar were MIT grad students, first encountered GV after the Lightmatter trio won a US$100,000 MIT entrepreneurship competition. GV general partner Erik Nordlander was a judge in the competition. Subsequent visits to Silicon Valley to make their pitch resulted in early seed funding from other investors.
Those early stakes have now been boosted by the US$22 million additional stake from the GV-led group. In a press release accompanying the announcement, GV partner and new Lightmatter board member Tyler Clark said that GV believes that the Lightmatter team’s “theoretical expertise and engineering talent are clear differentiators in the market for artificial intelligence accelerators.”