CEO's Column
Search
More
AI Infrastructure

Lightmatter Unveils Passage M1000 Superchip for Advanced AI Infrastructure

ByNeelima N M
2025-04-01.3 months ago
Lightmatter Unveils Passage M1000 Superchip for Advanced AI Infrastructure
Lightmatter unveils the Passage M1000, a 3D photonic superchip delivering 114 Tbps optical bandwidth to accelerate AI model training and deployment. (Image generated using AI)

Lightmatter, a leader in photonic computing, has introduced the Passage M1000, a 3D photonic superchip engineered for next-generation XPUs and high-performance switches. Designed to meet the growing demands of AI data centers, the M1000 offers an unprecedented 114 terabits per second (Tbps) of optical bandwidth, enabling connectivity for thousands of GPUs within a single computing domain.

Spanning over 4,000 square millimeters, the M1000 acts as a multi-reticle active photonic interposer. Unlike conventional chip designs, which restrict input/output (I/O) connectivity to chip edges, the M1000 integrates electro-optical I/O across its entire surface.

This is made possible through a reconfigurable waveguide network that transmits high-bandwidth wavelength-division multiplexed (WDM) optical signals throughout the chip.

Massive Fiber Integration and Compact Footprint

The M1000 supports an unprecedented 256 fiber attachments, each offering 448 Gbps bandwidth, delivering an order of magnitude improvement in total bandwidth compared to conventional co-packaged optics (CPO). Despite this performance leap, the platform maintains a compact footprint, further supporting Lightmatter’s mission to redefine the scalability of AI systems.

Lightmatter collaborated with industry partners GlobalFoundries (GF) and Amkor to support volume production. The chip is built using GF’s Fotonix platform, which seamlessly integrates photonic components with high-performance CMOS logic, creating a monolithic solution ideal for AI workloads.

CEO Nick Harris said, “Passage M1000 is a breakthrough achievement in photonics and semiconductor packaging for AI infrastructure.”

He added, “We are delivering a cutting-edge photonics roadmap years ahead of industry projections. Shoreline is no longer a limitation for I/O. This is all made possible by our close co-engineering with leading foundry and assembly partners and our supply chain ecosystem.”


Also read: Treasure Global to Develop Advanced AI Cloud Infrastructure in Malaysia

Technical Highlights and Availability

The M1000 is equipped with cutting-edge features, including an 8-tile 3D active interposer with a programmable waveguide network, and 1,024 integrated electrical SerDes supporting 56 Gbps NRZ modulation.

It also boasts an 8-wavelength WDM (Wavelength Division Multiplexing) on waveguides and fibers, along with 256 edge-attached optical fibers. Despite its advanced capabilities, the M1000 delivers 1.5 kW of power in a compact 7,735 mm² package, combining high-performance optical and electrical integration in an efficient design.

Together with Lightmatter’s newly announced Passage L200 and the Guide™ light engine, the M1000 aims to accelerate AI model training and deployment. The Passage M1000 is expected to be available in summer 2025.

Related Topics

AI Computing ResourcesAI infrastructureAI Data Centers

Subscribe to NG.ai News for real-time AI insights, personalized updates, and expert analysis—delivered straight to your inbox.