Optical Computing
Silicon photonics, co-packaged optics, photonic AI accelerators, optical neural networks
Integrated Platforms and Techniques for Photonic Neural Networks
Comprehensive review of integrated photonic platforms for neural networks
Integrated Neuromorphic Photonic Computing for AI Acceleration
Photonic tensor cores: 880 TOPS/mm2, 5.1 TOPS/W — 1-3 OOM over digital accelerators
Ultra-compact photonic AI chip operates at the speed of light
Inverse-designed nanophotonic neural network achieving 90-99% classification accuracy on 10K+ biomedical images at picosecond timescales
An integrated large-scale photonic accelerator with ultralow latency
16,000+ component photonic accelerator on single chip demonstrating ultralow latency and faster computing than commercial GPU
Harnessing Photonics for Machine Intelligence
System-level cross-layer benchmarking framework (SimPhony) revealing that peripheral overheads dominate photonic AI energy budgets, and that dynamic Transformer workloads expose fundamental limitations in MZI mesh architectures
Integrated photonic neuromorphic computing: device, architecture, chip, algorithm
Systematic multi-layer review of photonic neuromorphic computing spanning photonic neurons, synapses, network architectures, integrated chips, and optimization algorithms
A fully-programmable integrated photonic processor for domain-specific and general-purpose computing
First single programmable photonic processor handling both NP-complete problem solving (100% accuracy) and general-purpose matrix computation (97% MNIST accuracy) without hardware modification
Quantum photonics on a chip
Comprehensive review of integrated quantum photonic circuits on chip, synthesizing materials platforms, single-photon sources, detectors, and applications across quantum computing, communication, and sensing
Photonic convolutional neural network with pre-trained in-situ training
Fully-optical CNN for MNIST at 94% accuracy with 100-242x better energy efficiency than GPUs, 2132 on-chip tunable parameters, no O/E/O conversions.
Integrated photonic neural network with on-chip backpropagation training
First integrated photonic deep NN trained end-to-end with on-chip gradient-descent backpropagation — all linear AND nonlinear computation on a single photonic chip, robust to fabrication variations (>90% accuracy). Nature 651, 927-932 (2026). Answers the KB's 'training, not just inference' open question.
Inverse-designed nanophotonic neural network accelerators for ultra-compact optical computing
Peer-reviewed Nature Communications paper for the Univ. of Sydney inverse-designed nanophotonic NN accelerator: ~400M parameters/mm2 via subwavelength voxels as trainable DOF; ~89% MNIST / ~90% MedNIST on-chip (90-99% on 10K+ biomedical images), picosecond/speed-of-light, human-hair footprint. Authority upgrade over the prior news-only entry.
Deep neural network inference on an integrated, reconfigurable photonic tensor processor
imec iSiPP50G photonic tensor processor with PyTorch interface; honestly reports 0.022 TOPS/W vs 5.65 TOPS/W electronic — peripherals (DAC/ADC/TIA) dominate the energy budget, not the photonic core
SKYLIGHT: A Scalable Hundred-Channel 3D Photonic In-Memory Tensor Core for Real-Time AI Inference
SKYLIGHT proposes scalable 3D, WDM-enabled, non-volatile photonic tensor core architecture with hundred-channel parallelism for real-time AI inference
LightIn: Versatile Silicon Integrated Photonic Processor for AI Clusters
LightIn 160-component silicon photonic processor — first to realize compute + signal processing + switching + encryption on a single CMOS-compatible chip for AI clusters
Towards Topology-Aware Very Large-Scale Photonic AI Accelerators
Identifies the 'Utilization Wall' — topology-dominated scaling bottleneck in photonic accelerators; symmetric-grid topologies recover up to 6x utilization, cut memory access >40%
Lightmatter Passage L200: Record 1.6 Tbps Per Fiber
Record 1.6 Tbps/fiber; L200 CPO enabling 200+ Tbps/package; 8x faster AI training
Ayar Labs UCIe Optical Chiplet
First UCIe optical chiplet. 16-wavelength SuperNova, 8 Tbps. TSMC COUPE partnership. $500M raise Mar 2026 (Nvidia-backed). Production samples 2026.
Ayar Labs raises $500M to mass-produce CPO chiplets
$500M Series E for CPO mass production; TeraPHY delivers 200 Tbps/package — 5x Nvidia Rubin bandwidth
TSMC COUPE Platform for Silicon Photonics — Alchip + Ayar Labs Demo
First fully integrated COUPE-based optical I/O engine — 100 Tb/s per accelerator via UCIe interface
OIF 3.2T Co-Packaged Module Implementation Agreement
First industry co-packaging standard — 3.2Tb/s module with 8x400G optical interfaces enabling 51.2Tb/s switch bandwidth
Q.ANT NPU 2 Photonic Processor for AI and HPC
Second-gen photonic processor claiming 30x lower energy and 50x higher performance for AI/HPC, shipping early 2026
Lightmatter Unveils Passage M1000 Photonic Superchip, World's Fastest AI Interconnect
Passage M1000 3D photonic superchip — 114 Tbps total optical bandwidth, 256 fibers (448 Gbps/fiber), >4,000 mm2 8-tile 3D active photonic interposer with world-first built-in solid-state optical circuit switching; GF Fotonix + Amkor.
Lightmatter and GUC Partner to Produce Co-Packaged Optics (CPO) Solutions for AI Hyperscalers
Lightmatter + GUC strategic partnership to bring commercial Passage 3D CPO to AI hyperscalers — pairs Lightmatter's 3D-stacked silicon photonics engine with GUC's advanced ASIC design + packaging to extend AI cluster scale-up across multiple racks.
NVIDIA Announces Spectrum-X Photonics, Co-Packaged Optics Networking Switches to Scale AI Factories to Millions of GPUs
NVIDIA's own CPO switch line — Spectrum-X Photonics (Ethernet, 2H 2026) + Quantum-X Photonics (InfiniBand, 2025); 3.5x power efficiency, 4x fewer lasers, up to 400Tb/s; closes the KB's top-flagged gap
Broadcom Announces Tomahawk 6 – Davisson, the Industry's First 102.4-Tbps Ethernet Switch with Co-Packaged Optics
3rd-gen CPO switch BCM78919: 102.4Tbps via 64x1.6TbE, 16 Davisson DR optical engines on TSMC COUPE, ~70% lower power, field-replaceable lasers; shipping since Oct 2025
Lightwave Logic Q1 2026 Earnings Call — Strategic Investment Focus
Q1 2026 revenue $29K (+27% YoY); PDK 1.1 handoff to production May 2026; 4 Fortune 500/Global 500 customers at Stage 3; ~$100M cash; FY26/27 guidance only $0.41M/$0.64M
The Shift to Photonics for Data Center Networks and AI
Industry deployment timeline 2026-2028. Manufacturing, supply chain, integration realities.
Five Key Trends of Co-Packaged Optics (CPO) in 2026
Industry analysis: CPO reduces link power 30W→9W, UCIe optical extends package boundaries, thermal management is critical blocker