20 Dec 2023
Silicon photonics startup planning to open Toronto office as company headcount grows rapidly.
Lightmatter, the Boston-based startup developing silicon photonics computing hardware aimed at generative artificial intelligence (AI) and high-performance computing applications, says it has raised a further $155 million in venture funding.
The “series C-2” support, led by Google Ventures and Viking Global Investors, follows a similar-sized effort earlier this year, and brings its total fundraising so far to $420 million.
The funding will see the firm, which is currently looking to hire extensively, further expand its headcount, with plans to open a new office in Toronto next year. Lightmatter already has sites in Boston and Mountain View, California.
Co-founder and CEO Nick Harris commented: “Lightmatter is positioned to be a key driver in powering the next generation of computing systems that will further enable AI innovation.
“Through photonic technologies, Lightmatter is ensuring the steady progress in computing performance continues, despite growing power consumption challenges and slowing progress with transistor scaling.”
Erik Nordlander, general partner at Google Ventures, added: “Lightmatter is harnessing the power of silicon photonics to meet [the generative AI] challenge, unlocking performance bottlenecks, increasing bandwidth, and allowing AI models to increase in size and scale.”
Among the products introduced by Lightmatter earlier this year is its “Passage” photonic interconnect, which is said to offer a dramatic (40x) improvement in interconnect density, allowing side-by-side integration of silicon photonics with electronics.
“Attaching optical fibers to chips is extremely expensive,” points out the firm. “Passage has built-in photonic interconnects and no fibers to attach, which means reduced complexity.
“By removing the need for driving long wire traces between computer chips and optical transceivers, Passage achieves 5x more energy efficient communications with 200x more bandwidth - reducing operating costs for high-performance computing systems.”
Meanwhile the “Envise” hardware is described as “the world’s first photonic computing platform”. Lightmatter says it is able to run the largest neural networks developed to date, while consuming a relatively low power of 3 kW.
Behind the silicon photonics hardware was earlier work focused on programmable Mach-Zehnder interferometers.
Alongside MIT-trained Harris, Lightmatter’s co-founders included chief scientist Darius Bunandar, and COO and former Google employee Thomas Graham..
The firm says that Danner Stodolsky, previously Google’s VP of engineering, has now joined as its VP of data center architecture, and will lead the design, simulation and evaluation of large-scale systems based on Lightmatter's technology.
Lightmatter is also part of a $4.8 million project funded by the US Intelligence Advanced Research Projects Activity (IARPA) to develop a photonic processor specifically for autonomous vehicles.
Harvard University, Boston University, and MIT are all part of the effort, which is looking to produce a system capable of supporting "Level 4" autonomy, where no human control of a vehicle is needed.