Lightmatter, which is developing AI chips that use light signals rather than electrical signals, raises $80M Series B, bringing its total raised to about $113M
https://techcrunch.com/...
Context & Ripple Effects
This round is a midpoint in one of the longest-running venture arcs in AI silicon: Boston-based Lightmatter took an $11M Series A from Matrix Partners and Spark in 2018 to build chips that process AI workloads with light rather than electrons, then added a $22M A-1 led by GV a year later. The $80M Series B brings total funding to about $113M — still early-stage money by what came after.
First-order effects
- Lightmatter gets the capital to push its photonic AI chips past prototype toward manufacturable silicon, while Matrix, Spark, and GV double down on a physics-bet that most AI chip investors were still treating as fringe in 2021.
Second-order effects
- The round legitimizes photonics as an investable category: rival Lightelligence had raised a Baidu Ventures-backed seed for photonic AI circuits just days before Lightmatter's Series A, and later backers Fidelity and T. Rowe Price followed the trail this B blazed into a $400M Series D at a $4.4B valuation.
Third-order effects
- The trajectory — from $11M to $850M raised, ending in shipping an interposer and chiplet for optical chip-to-chip data movement — points to photonics graduating from a compute experiment into an interconnect layer of AI infrastructure, with each round buying the years of fab-grade engineering that electrical-switching incumbents would otherwise own outright.
The trend: AI compute is pulling optical interconnect and photonic processing out of the lab through an escalating sequence of ever-larger venture rounds, as data movement between chips becomes the bottleneck worth betting on.