CogniFiber

What CogniFiber does

CogniFiber is an Israeli deep-technology company founded in 2018 and headquartered in Rosh HaAyin, Israel, that develops pure-photonic computing hardware for AI inference workloads. Its core product, the DeepLight processor, is a fully-optical, brain-inspired non-von Neumann neuromorphic processor that performs computation as photons travel through multi-core optical fibers, eliminating traditional memory-access and bus bottlenecks. The technology portfolio includes the Aurora full-system photonic ML/AI prototype (a 4-layer classifier achieving 100M inferences per second), trainable photonic neurons operating at 0.5 Gbps per channel, a 4-bit D2A photonic chip for data compression, and a 32-channel all-photonic MAC operator for pattern recognition, supported by 32 patent applications.

The company's go-to-market is direct enterprise engagement with data center operators and AI infrastructure providers, signaled by a 'BOOK A DEMO' call-to-action on its website. CogniFiber targets four primary use-case segments — Smart Cities, Smart Mobility, Industrial IoT, and Data Center Operators — with secondary applications in cybersecurity and pharmaceutical AI. Its revenue model is hardware sales of DeepLight processors, with no public pricing disclosed; commercial contracts appear to be quote-based enterprise deals. The company has raised $14 million to date across a 2019 seed round, a 2022 Series A, and a 2024 round led by Eastern Epic Capital and Chartered Group, with grant support from the Israel Innovation Authority.

Leadership combines commercial and academic credentials: Dr. Eyal Cohen serves as CEO and Co-Founder, bringing experience across neuroscience (Weizmann Institute), hardware design (Mellanox, Saifun), and photonic computing research (Bar Ilan, Hebrew University); Prof. Zeev Zalevsky serves as CTO and Co-Founder, a Bar Ilan University Full Professor and Dean of Engineering with 570+ peer-reviewed papers, approximately 100 patents, and prior co-founder/CTO roles at multiple exits (Civcom, XCeed, Zsquare Medical, ContinUse Biometrics). The broader team of 1–10 employees includes a VP R&D, Head of Products, AI Scientist, Lead R&D Physicist, Senior FPGA Engineer, and Optical Lab Manager.

Name

CogniFiber

Legal name

CogniFiber

Website

https://cognifiber.com

Company type

Private

Founded year

2018

Operating status

Operating

Headcount range

1–10 employees

Short description

CogniFiber is an Israeli fabless deep-tech company developing pure-photonic neuromorphic processors (DeepLight) that perform AI inference inside multi-core optical fibers, targeting enterprise data center operators and AI infrastructure providers.

Ownership category

akta.pro rank

Product category

Photonic AI Processors

NAICS

Semiconductor and Related Device Manufacturing (334413), Semiconductor and Other Electronic Component Manufacturing (3344)

SIC

Semiconductors & Related Devices (3674), Electronic Computers (3571)

akta.pro primary industry

AI Inference Hardware & Edge Accelerators (HDAAAAAC)

akta.pro secondary industries

Compute & Acceleration ICs (CPU/GPU/AI/FPGA/ASIC) (HDAHABAC), AI Accelerators (GPUs/TPUs/NPUs/ASICs) (HDAAAAAA)

Keywords

• Photonic computing
• AI inference processors
• Optical fiber computing
• Neuromorphic processors
• Photonic AI hardware

Headquarters

HQ city

Rosh Ha'ayin

HQ country

Israel

HQ region

Middle East

Offices1 record

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Technology or R&D, Personnel, Operations, Marketing or Sales, Infrastructure

Revenue model

1. Hardware Processor Sales: Sale of DeepLight photonic processors to enterprise customers for AI infrastructure deployment in data centers. The company is developing processors for integration into AI computing infrastructure.

Go-to-market motion1 record

Distribution channels1 record

Marketing channels3 records

Core offering

CogniFiber develops and sells the DeepLight pure-photonic supercomputing processor, which uses multi-core optical fibers to perform AI computation at the speed of light as photons travel through the fiber. The technology implements a brain-inspired, non-von Neumann architecture that eliminates traditional memory access and bus bottlenecks, enabling accelerated AI inference for data center, smart city, smart mobility, and industrial IoT deployments.

Product overview

CogniFiber is an Israeli company developing pure photonic computing technology for AI workloads. The core product is the DeepLight photonic supercomputing processor, which uses multi-core optical fibers to process data at the speed of light. The technology portfolio includes the Aurora full-system photonic ML/AI prototype, trainable photonic neurons (0.5Gbps per channel), a 4-bit D2A photonic chip for data compression, and a 32-channel all-photonic MAC operator for pattern recognition. Together, these components demonstrate an end-to-end photonic computing system that eliminates traditional von Neumann bottlenecks by performing computation directly within the fiber during data transmission.

Differentiator

Problem solved

Functional benefit

Brands

• DeepLight: Pure-photonic supercomputing processor technology that processes data at the speed of light using multi-core optical fibers for AI inference.

Products and services

• DeepLight Photonic Processor A pure-photonic supercomputing processor that uses multi-core optical fibers to process AI data at the speed of light as photons travel through the fiber, eliminating traditional memory access bottlenecks and von Neumann architecture limitations. Targeted at data center operators, AI infrastructure providers, and developers of smart city, smart mobility, and industrial IoT systems.
• Aurora Full-System Photonic ML/AI Classifier Full-system photonic ML/AI prototype featuring a 4-layer neural network architecture that achieves 100M inferences per second, demonstrating the company's in-fiber photonic processing capabilities. Intended for customers evaluating photonic AI inference for pattern recognition and classification workloads.

Quantifiable outcome

• 1000X faster AI inference than current processors
• +6 more outcomes

Segments5 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability3 records

Feature6 records

Partnerships

One partnership is on record.

• NTT Innovation Laboratory IsraelstrategicStrategic or Co-development PartnerNTT Innovation Laboratory Israel participated in a joint initiative with Ignite the Spark creating an innovation map of Israeli energy-tech startups. CogniFiber is featured among approximately 60 Israeli startups developing energy solutions for data centers. NTT is the world's third-largest data-center service provider and positions itself as a potential partner for Israeli innovation.

Scale indicators8 records

Recent moves6 records

Expansion highlights5 records

Direct peers

• Lightelligence: Lightelligence develops optical computing chips for AI workloads, pursuing photonic matrix multiplication and inference acceleration that directly competes with CogniFiber's in-fiber photonic ML approach.
• Celestial AI: Celestial AI combines photonic fabric with memory for AI compute and interconnect (Photonic Fabric), directly addressing the same AI inference and data center performance-per-watt problem CogniFiber targets with photonic processors.
• Lightmatter: Lightmatter builds photonic AI processors (Envise, Passage) using silicon photonics for AI inference and interconnects, directly competing with CogniFiber's DeepLight in the photonic AI accelerator category and serving the same data center and AI infrastructure customers.
• Luminous Computing: Luminous Computing is developing a photonic AI supercomputer aimed at large-scale AI workloads, making it a direct architectural competitor to CogniFiber's pure-photonic DeepLight supercomputing processor.
• Fathom Computing: Fathom Computing focuses on optical/photonic computing for AI, pursuing similar brain-inspired photonic architectures as CogniFiber to overcome the memory and data movement bottlenecks of von Neumann systems.

Broad incumbents

• Intel (Habana Labs): Intel's Habana Labs unit produces Gaudi and Greco AI accelerators for data center training and inference, representing a broader incumbent offering competing for the same AI infrastructure budgets CogniFiber targets.
• Nvidia: Nvidia is the dominant incumbent supplier of AI training and inference GPUs and is the explicit benchmark in CogniFiber's 1000X performance claim. Its data center platform is the primary incumbent solution CogniFiber aims to displace for inference workloads.

Emerging players

• Syntiant: Syntiant produces ultra-low-power edge AI processors for always-on sensor inference, overlapping with CogniFiber's Smart Mobility and Industrial IoT segments where low-power AI inference at the edge is the key value proposition.
• Groq: Groq builds the LPU, a deterministic AI inference accelerator targeting low-latency, high-throughput inference in data centers — a closely comparable end-application (AI inference acceleration) though implemented in silicon rather than photonics.
• Mythic AI: Mythic develops analog compute-in-memory AI inference processors for edge devices, sharing CogniFiber's neuromorphic/analog-philosophy roots and overlapping in edge AI inference use cases (IIoT, smart mobility).

Market position

Strengths5 records

Weaknesses5 records

Competitive moat3 records

Key risks6 records

Key highlights6 records

Customer concentration

Revenue estimate

Valuation estimate

Number of profiles

Profiles8 records

Funding overview

Funding rounds3 records

Investors2 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about CogniFiber