# Lightmatter

**Source:** https://geo.sig.ai/brands/lightmatter  
**Vertical:** AI Infrastructure  
**Subcategory:** Photonic AI Chips  
**Tier:** Emerging  
**Website:** lightmatter.com  
**Last Updated:** 2026-04-14

## Summary

Lightmatter (MIT spinout, $4.4B, $850M raised) replaces copper chip-to-chip links with photonic interconnects; M1000 Passage delivers 114 Tbps bandwidth for AI clusters.

## Company Overview

Lightmatter is a photonic computing company spun out of MIT with a mission to overcome the fundamental bandwidth and energy bottlenecks that are constraining AI hardware scaling. As AI models have grown to require thousands of interconnected chips, the copper-based interconnects between chips have become a critical chokepoint — slow, power-hungry, and thermally limited. Lightmatter's founding insight was that light-based data interconnects could solve this problem by transmitting data at the speed of light with dramatically lower energy consumption.\n\nLightmatter's primary product is Passage, a photonic interconnect technology that replaces electrical chip-to-chip communication with optical links. The M1000 implementation delivers 114 terabits per second of aggregate bandwidth, enabling AI clusters to scale with far less latency and energy overhead than electrical alternatives. Passage is designed to be compatible with existing chip architectures and manufacturing processes, allowing hyperscalers and AI hardware vendors to integrate photonic interconnects without redesigning their entire stack.\n\nLightmatter has raised $850 million and achieved a valuation of $4.4 billion, making it one of the most highly capitalized companies in the AI infrastructure hardware space. The company's investors include Google, HPE, and a range of deep-tech focused funds. As AI training and inference workloads continue to scale, the demand for high-bandwidth, low-latency chip interconnects is expected to grow substantially, positioning Lightmatter at a critical node in the global AI compute supply chain.

## Frequently Asked Questions

### What does Lightmatter do?
Silicon photonic interconnects using light to move data between AI chips at 114 Tbps, eliminating data center bandwidth bottlenecks.

### How much has Lightmatter raised?
$850M total, $400M Series D at $4.4B. T. Rowe Price, Fidelity, GV.

### What is the Passage M1000?
3D photonic superchip interposer with 114 Tbps bandwidth for AI data center interconnects.

### How does photonic interconnect reduce AI data center energy consumption?
Electronic interconnects between AI chips consume significant power due to electrical resistance and the need to drive signals at high voltages across chip-to-chip distances. Photonic interconnects transmit data as light, which requires far less energy per bit at the speeds required for AI workloads. Lightmatter's silicon photonics technology can reduce interconnect energy by over 10x compared to copper alternatives, which is increasingly important as interconnect power becomes a significant fraction of total data center energy budgets.

### What is Lightmatter's Passage platform?
Passage is Lightmatter's photonic AI interconnect platform — a silicon photonic chip that acts as an interposer enabling direct optical connections between AI accelerator chips. The Passage M1000 delivers 114 Tbps of bandwidth to connect AI chips in a wafer-scale configuration, eliminating the bandwidth bottleneck between processors and enabling GPU-scale clusters to operate closer to their theoretical compute limits.

### How does Lightmatter compare to Nvidia's NVLink for chip interconnect?
NVLink is a proprietary NVIDIA copper-based interconnect limited to connecting NVIDIA GPUs within NVIDIA's own hardware ecosystem. Lightmatter's photonic interconnects are hardware-agnostic — any chip manufacturer can integrate Passage as a substrate — and achieve dramatically higher bandwidth (114 Tbps vs NVLink's ~1.8 Tbps per switch) at lower power per bit. Lightmatter positions Passage as the interconnect layer for post-NVIDIA heterogeneous AI clusters.

### Who are Lightmatter's target customers?
Lightmatter targets hyperscalers (Amazon, Google, Microsoft, Meta) and merchant silicon vendors designing custom AI accelerator systems who need to overcome memory bandwidth limitations in next-generation AI clusters. Cloud providers designing their own chips (TPUs, Trainium, Maia) are natural customers for Passage as an interconnect solution that scales beyond what copper allows.

### What is Lightmatter's manufacturing approach?
Lightmatter manufactures its photonic chips using standard CMOS foundry processes, making Passage compatible with volume semiconductor manufacturing at TSMC and similar foundries rather than requiring exotic photonics-specific fabs. This compatibility with existing foundry infrastructure is critical to achieving the cost and scale required for data center deployment.

## Tags

ai-powered, b2b, infrastructure, saas

---
*Data from geo.sig.ai Brand Intelligence Database. Updated 2026-04-14.*