Photonic Interconnects

Photonic interconnects solve the most immediate bottleneck in AI data center scaling: the bandwidth wall between chips. As AI model sizes grow beyond what fits on a single accelerator, training requires distributing computation across thousands of GPUs connected by a network fabric. Electrical interconnects are hitting fundamental limits — signal integrity degrades with distance, power consumption scales with data rate, and copper traces can't keep pace with the bandwidth demands of trillion-parameter models.

Co-packaged optics (CPO) places optical transceivers directly on or adjacent to the compute die, converting electrical signals to light at the earliest possible point. This eliminates the lossy, power-hungry electrical path to separate pluggable optics modules. Lightmatter's Passage L200 has demonstrated a record 1.6 Tbps per fiber and over 200 Tbps per package — bandwidth that enables fully non-blocking communication between AI accelerators.

The commercial impact is substantial: Lightmatter claims up to 8x faster AI model training with their interconnect fabric, stemming from reduced communication bottlenecks during distributed training. The company's partnership with Qualcomm signals that CPO is moving from specialized AI infrastructure toward broader adoption in mobile and edge computing.

A second architectural axis is emerging above the chiplet-style CPO that Ayar Labs (TeraPHY) and the OIF standard define. Lightmatter's Passage M-series moves optics onto an active photonic interposer — the M1000 is a multi-reticle, >4,000 mm² 3D interposer carrying 114 Tbps of total optical bandwidth across 256 fibers (448 Gbps/fiber), with electro-optical I/O routable virtually anywhere on its surface for the die complex stacked on top, plus the world's first built-in solid-state optical circuit switching on the package. Where a TeraPHY chiplet sits beside the compute die, the M1000 is the substrate the compute die sits on. Commercialization of this 3D-CPO line now has a supply path: Lightmatter's January 2026 partnership with GUC (Global Unichip Corp) co-designs Passage 3D CPO into GUC's ASIC + advanced-packaging flow for AI hyperscalers — the interconnect analogue of Ayar Labs' route through TSMC COUPE/UCIe.

The broader significance is architectural. Photonic interconnects could enable disaggregated computing — separating memory, compute, and storage into independently scalable pools connected by optical fabric, rather than the monolithic architectures that dominate today. Lightmatter explicitly frames the M-series + GUC path as extending the AI cluster scale-up domain "across multiple racks."

Key Claims

• 1.6 Tbps per fiber achieved — Lightmatter Passage L200 sets record for single-fiber bandwidth. Evidence: strong (Lightmatter Passage L200)
• 200+ Tbps per package — Co-packaged optics enabling fully non-blocking chip-to-chip communication. Evidence: strong (Lightmatter Passage L200)
• 114 Tbps total optical bandwidth on a 3D active interposer — Lightmatter Passage M1000: >4,000 mm² 8-tile interposer, 256 fibers at 448 Gbps each, 8-wavelength WDM, 1,024 SerDes, 1.5 kW delivery, world-first on-package solid-state optical circuit switching (GF Fotonix + Amkor). A distinct architectural axis from chiplet-style CPO. Evidence: strong (vendor spec) (Lightmatter Passage M1000)
• Commercial 3D CPO supply path via GUC — Lightmatter + GUC (Jan 2026) co-design Passage 3D CPO into GUC ASIC + advanced packaging for AI hyperscalers, extending cluster scale-up across multiple racks. Interconnect analogue of the Ayar Labs → TSMC COUPE route. Evidence: moderate (press release, no independent benchmark) (Lightmatter–GUC Partnership)
• Up to 8x faster AI training claimed — Reduced communication bottlenecks in distributed training. Evidence: moderate (Lightmatter Passage L200)
• CPO moving beyond AI niche — Qualcomm partnership signals broader adoption trajectory. Evidence: moderate (Lightmatter Passage L200)
• First UCIe-compliant optical chiplet — Ayar Labs standardizes optical I/O into the chiplet ecosystem. SuperNova 16-wavelength light source at 8 Tbps. TSMC COUPE partnership makes it accessible to broader semiconductor industry. Evidence: strong (Ayar Labs UCIe Chiplet)
• $500M Nvidia-backed raise signals market conviction — Nvidia sees optical interconnect as essential for next-gen AI systems. Production samples targeted for 2026. Evidence: strong (Ayar Labs UCIe Chiplet)
• 2026-2028 deployment timeline for AI data centers — Early adopters deploying CPO in 2026; broader 800G/1.6T adoption in 2027; photonic interconnects standard for AI-scale networking by 2028. Evidence: moderate (Photonics Shift)
• Manufacturing readiness is the real bottleneck — Photonic component yield, InP laser supply constraints, and workforce skills gap constrain adoption pace more than the physics. Evidence: moderate (Photonics Shift)

Open Questions

• What is the cost premium of CPO vs. pluggable optics at data center scale?
• Can CPO integration avoid yield losses from combining optical and electronic fabrication?
• Will photonic interconnects enable true disaggregated computing architectures?
• How do optical switching fabrics compare to electrical switches for AI cluster networks? (Lightmatter's M1000 now puts solid-state optical circuit switching on the package itself — independent benchmarks vs. electrical switching pending.)
• Active photonic interposer (Lightmatter M-series) vs. chiplet-style CPO (Ayar Labs TeraPHY / TSMC COUPE): which architecture wins for hyperscale scale-up, and do they converge?

Related Concepts

• Photonic Neural Networks — Compute layer that photonic interconnects serve
• Photonic Tensor Cores — High-density compute requiring high-bandwidth data feeds

Backlinks

Pages that reference this concept:

• Lightmatter
• Ayar Labs
• GUC (Global Unichip Corp)

Changelog

• 2026-06-24 — Added the Lightmatter Passage M1000 active-photonic-interposer architecture (114 Tbps, 256 fibers, on-package optical circuit switching) as a distinct axis from chiplet-style CPO, and the Lightmatter–GUC commercialization partnership (Jan 2026). New entity: GUC.