Top 50 Analog Neural Network Chip

Irreversible provides analog in‑memory AI inference chips that execute neural‑network multiply‑accumulate operations directly within memristive non‑volatile memory cells. By eliminating digital‑to‑memory data movement, the chips achieve up to a thousand‑fold reduction in energy use, enabling continuous inference on power‑constrained edge devices such as drones, robots, and sensors. The solution includes a hardware‑aware training workflow, digital‑twin simulation, and flexible IP delivery models for integration into ASICs or system‑in‑package modules.

The startup specializes in neuromorphic analog signal processing technology to create ultra-low power AI chips that operate at microwatt levels and microsecond latency. This enables industrial electronics manufacturers to implement neural networks directly at the sensor, effectively addressing the challenges of power constraints in tiny AI applications.

Mythic provides analog compute‑in‑memory AI inference accelerators that integrate compute and weight storage on a single silicon plane, eliminating off‑chip memory traffic. Delivered as standard M.2 cards, the APUs achieve up to 25 TOPS with 3‑4× lower power than comparable digital accelerators, and are compatible with TensorFlow and PyTorch for edge devices such as robots, drones, and smart‑city cameras.

SEMIQA provides an analog in‑memory accelerator that performs matrix‑vector operations directly within configurable synapse crossbars, reducing data movement and energy consumption. The platform delivers sub‑nanosecond latency with >10× lower energy per operation than digital ASICs, supports on‑chip learning, and includes a software stack that maps TensorFlow/PyTorch models to analog kernels for edge and data‑center deployments.

The startup develops deep sub-threshold analog in-memory computation semiconductors that execute data-center-grade AI workloads with significantly lower power consumption and no active cooling. This technology processes data within memory, enabling industries to efficiently handle complex tasks while minimizing energy use.

This company develops proprietary neuromorphic processors utilizing a fully analog pipeline with in-memory computing and spiking neurons. Their chips enable ultra-low power, real-time processing of temporal sensor data for edge AI applications like keyword spotting and wake word detection. The technology achieves high accuracy while consuming significantly less power and using fewer parameters than conventional digital neural networks.

Vellex provides a programmable analog IP core that enables on‑device AI model training for low‑power edge hardware. The core maps gradient‑based optimization onto a physics‑driven circuit, achieving weight updates in nanoseconds with sub‑10 mW power, and is programmed via a code‑to‑circuit compiler that accepts PyTorch, JAX, or TensorFlow models. Vellex also offers a software optimizer and a developer kit for rapid prototyping, allowing semiconductor partners and OEMs to embed continuous learning without cloud dependence.

GEMESYS develops fully analog, brain-inspired AI hardware chips utilizing memristor technology. This integrated circuit enables high-efficiency AI training and inference operations directly at the edge. The company provides a low-power, data-efficient platform to deploy autonomous and adaptable artificial intelligence on mobile devices and sensors.

Innatera develops ultra-low-power neuromorphic processors based on a proprietary analog-mixed signal computing architecture. These processors utilize spiking neural networks to enable high-performance pattern recognition directly at the sensor edge. The technology delivers cognition performance with ultra-low power consumption and short response latency for power-limited applications.

Ambient Scientific develops ultra-low power AI microprocessors, such as the GPX10, specifically designed for on-device edge computing applications. These processors utilize proprietary architecture to accelerate neural networks while consuming minimal power, enabling years of always-on AI functionality from a single battery. This technology reduces reliance on cloud infrastructure, offering lower latency and enhanced data privacy for sensor fusion, audio, and vision tasks.

Reexen develops high-performance neural network processors optimized for sensor-end applications. Their processors enable advanced AI capabilities directly on sensor devices, reducing latency and improving efficiency for real-time data processing.

EnCharge AI develops high-efficiency analog in-memory computing GPUs and digital AI accelerators for edge-to-cloud deployment. Their validated hardware and flexible software offer significant improvements in performance, TCO, and sustainability compared to traditional solutions. The company provides versatile products from chiplets to PCIe cards, enabling seamless orchestration for on-device and cloud AI inference.

Conventional image sensors and vision processors consume significant power and often lack sensitivity across a broad spectral range, limiting their effectiveness in low‑light or adverse weather conditions for autonomous systems and immersive devices. 2DNeuralVision creates a suite of low‑power computer‑vision components built on graphene and transition‑metal dichalcogenide (TMDC) materials. The project delivers a wide‑spectrum 2‑dimensional image sensor that captures visible to short‑wave infrared light while maintaining minimal energy draw.

Mach42 utilizes proprietary neural network technology to accelerate the verification process of analog circuit designs, achieving high accuracy with minimal data input. This platform significantly reduces design cycle times, enabling faster time-to-market for complex simulations in engineering and scientific applications.

Neuron IP provides end‑to‑end ASIC design services for AI and edge workloads, handling architecture definition, RTL development, synthesis, physical design, and tape‑out management. The company also offers a library of configurable silicon IP blocks and chiplet interface IP with automated verification, enabling fabless semiconductor firms and system integrators to accelerate time‑to‑market without extensive in‑house expertise.

Unaware provides a PCIe plug‑and‑play AI accelerator ASIC that runs neural network inference directly on hardware without a host operating system or runtime libraries. The chip’s dataflow architecture, on‑chip weight storage, and secure enclaves deliver over 10 TOPS/W efficiency while protecting model and data privacy, targeting privacy‑focused AI developers, edge‑computing startups, and small research labs.

Type 1 Compute develops neuromorphic processors that enable high-performance, low-power AI inference and continuous learning directly on edge devices. Their architecture emulates biological neural networks to achieve sub-100ms latency and GPU-level compute efficiency at under 1W, removing cloud dependency for real-time decision-making in applications like autonomous systems and medical devices.

SynSense develops mixed-signal neuromorphic processors that achieve ultra-low power consumption and low-latency performance for edge computing applications. Their technology addresses the challenges of high energy use and slow response times in AI systems, enabling efficient real-time processing across various domains such as robotics, smart homes, and autonomous driving.

SilicoSapien offers a neuromorphic AI architecture that mimics biological neural pathways for highly efficient and interpretable AI. Its brain-inspired design reduces energy consumption by up to 1,000x and provides transparent decision-making, ideal for power-constrained environments and critical applications.

SimChip specializes in the design and production of high-end analog integrated circuits, focusing on precision and performance for various applications. The company addresses the need for reliable and efficient analog solutions in industries such as telecommunications and automotive, enhancing signal integrity and reducing power consumption.

Analog Craft provides an intelligent design flow to enhance analog integrated circuit design productivity. The platform utilizes AI and modern compute resources to streamline the process from initial concept through to silicon implementation. This solution is engineered to accelerate the development cycle for analog IC teams.

Ingk is developing a new class of supercomputing semiconductors that utilize non-silicon, non-arithmetic chip designs to achieve processing speeds of up to 10 THz while consuming significantly less power than traditional silicon-based GPUs and CPUs. This technology eliminates the limitations of Deep Neural Networks through self-reorganizing programming, enabling scalable and flexible computing solutions for advanced AI applications.

Develops on-device AI semiconductor solutions, including custom NPUs, SoC ASICs, and specialized modules, optimized for low power consumption and high performance in applications like video analytics, security, and robotics. By enabling real-time AI processing with support for multiple models on a single chip, DEEPX addresses the challenges of latency, privacy, and network costs associated with cloud-based systems. Its scalable architecture and 259 patents ensure cost-competitive, silicon-proven products for global markets.

AIDAChip offers an AI‑native operating system that unifies the entire chip design lifecycle, centralizing specifications, simulations, layout, and verification in a shared knowledge graph. AI‑driven coordination agents automatically propagate changes, generate testbenches, and schedule simulations, cutting tape‑out cycles and reducing engineering overhead for analog, digital, and mixed‑signal projects.

Omni Design Technologies provides high‑speed, low‑power analog‑to‑digital converter and front‑end semiconductor IP for heterogeneous system‑on‑chip designs, offered as reusable IP blocks, chiplet “droplets,” or hard macros. Their 64 GS/s ADC cores and multi‑channel analog front‑ends deliver >10 ENOB with sub‑100 µW power per channel, include on‑chip PVT monitoring, and support DSP‑ready interfaces such as JESD204B/C and high‑bandwidth SerDes. The IP enables fabless and in‑house design teams to accelerate development of AI accelerators, data‑center networking, automotive ADAS, telecom RF, aerospace, and quantum computing products.

BrainChip licenses AI accelerator hardware designs and development tools for on-device intelligence. Their Akida processor IP utilizes sparsity and event-based neural networks to deliver unmatched efficiency for real-time AI applications. This technology reduces latency and power consumption, enabling devices to detect, analyze, and respond to events without cloud dependence.

Qingwei Intelligent Technology designs and manufactures AI chips optimized for high-performance computing tasks in artificial intelligence applications. Their technology addresses the demand for efficient processing power in AI systems, enabling faster data analysis and improved machine learning capabilities.

Aivatech develops 3D-vision AIoT chips that integrate advanced facial and palm vein recognition technologies, RGB-D depth cameras, and vehicle license plate recognition systems. These solutions enhance security and automation in various applications, addressing the need for efficient and accurate identification in smart environments.

Scalinx designs and industrializes high-performance semiconductor chips for analog signal conversion, featuring proprietary SCCORE™ technology that optimizes size, weight, and power consumption. Their solutions include highly configurable data converter cores and agile RF receivers, addressing the need for efficient, low-noise signal processing in communication, defense, and test measurement applications.

Deepsilicon develops software and hardware solutions that optimize neural network performance on-device, achieving 8x less RAM usage, 20x higher throughput, and 100x improved power efficiency. This technology addresses the challenges of high resource consumption and slow processing speeds in running complex AI models.

Singerno Microelectronics specializes in the design of digital-analog mixed-signal chips, enabling precise signal processing for various electronic applications. Their technology addresses the need for efficient power management and improved performance in consumer electronics and industrial systems.

Zettascale Computing Corporation designs energy-efficient, reconfigurable dataflow chips (XPUs) for AI training and inference. These chips adapt their architecture to specific AI models, optimizing dataflow and reducing memory movement for superior energy efficiency and throughput compared to traditional accelerators.

SpiNNcloud provides ultra energy-efficient computing infrastructure specifically optimized for next-generation AI inference workloads. Their brain-inspired chip architecture leverages dynamic sparsity to achieve significantly higher energy efficiency compared to traditional GPUs. This infrastructure enables scalable, low-power AI processing necessary to address growing GenAI energy demands.

Cognichip develops Artificial Chip Intelligence (ACI®), an AI-first platform for semiconductor design. This technology utilizes a physics-informed foundation AI model to fundamentally alter the economics of chip creation. ACI® provides a new level of intelligence and precision for scaling semiconductor development.

Xinyaotu provides a full-stack integrated single-chip solution that combines intelligent sensors and computing with proprietary AI algorithms. This technology enables efficient data collection and processing for applications requiring real-time insights, enhancing operational performance in various industries.

SEMRON develops a 3D-scalable AI inference chip using its proprietary CapRAM™ technology, which integrates compute-in-memory architecture to enhance energy efficiency and parameter density for AI applications. This technology addresses the high costs and power consumption of traditional AI chips, enabling efficient deployment of generative AI models directly on edge devices like smartphones and wearables.

SiliconBee designs custom AI ASICs with dedicated tensor cores delivering up to 200 TOPS per watt, integrated HBM2e memory exceeding 1 TB/s bandwidth, and a low‑latency mesh interconnect for scalable multi‑chip configurations. Its hardware and accompanying software stack provide native support for TensorFlow, PyTorch, and ONNX, enabling high‑throughput training and inference across data‑center and edge form factors. The company serves data‑center operators, AI‑focused cloud providers, and OEMs building edge inference devices that require efficient, high‑performance machine‑learning compute.

炬芯科技 (Actions Semiconductor) designs and supplies integrated System-on-Chip (SoC) solutions for low-power AIoT applications. Their chips deliver high-fidelity audio processing and low-latency wireless connectivity, enabling advanced features like on-device AI inference for smart audio devices and wearables.

Great Sky designs and manufactures novel computing hardware by integrating superconductors, semiconductors, and photonics. This extreme neural architecture is engineered to achieve theoretical limits in computational speed and energy efficiency. The company delivers high-performance intelligence solutions operating at the boundaries of physics.

Zhonghao Xinying develops specialized AI chips focused on enhancing computational efficiency for artificial intelligence applications. Their technology addresses the high energy consumption and processing limitations faced by current AI systems, enabling faster and more cost-effective AI deployment.

ThinkForce is a Shanghai-based manufacturer of artificial intelligence chips designed to enhance processing efficiency in machine learning applications. Their technology addresses the demand for high-performance computing solutions in industries reliant on real-time data analysis.

Key Foundry is a dedicated semiconductor foundry in South Korea that specializes in 8-inch wafer processing for analog, mixed-signal, power, and non-volatile memory (eNVM) applications. The company provides tailored online foundry services and multi-project wafer (MPW) programs to support the diverse manufacturing needs of its clients in the semiconductor industry.

Exigent provides a physics‑infused neural network platform that accelerates lithography pattern simulation, delivering full‑chip results in seconds on GPU/TPU clusters while preserving sub‑nanometer accuracy. The platform also includes a factory‑general‑intelligence engine that ingests sensor data to recommend process adjustments, predict equipment wear, and integrates via RESTful and gRPC APIs with CAD, MES, and ERP systems, offered as cloud SaaS or on‑premise Kubernetes.

Develops contact-free, chip-scale biomagnetic sensors that detect neural activity without direct skin contact, enabling integration into wearables for applications in health monitoring, gesture control, and personalized mental health treatments. These low-power, cost-effective sensors provide unprecedented data on biometrics, emotions, and cognitive states, supporting advancements in consumer devices, prosthetics, and neurodegenerative disorder detection.

GeNeXIC provides end‑to‑end custom ASIC and system‑design services for analog, mixed‑signal, RF, MMIC, and silicon‑photonic blocks, translating client specifications into silicon‑ready solutions. The company handles architecture, layout, verification, first‑silicon validation, and production hand‑off using leading EDA toolchains and a network of global foundries, ensuring performance, power efficiency, and area optimization for fabless and OEM customers.

Agentrys provides an AI‑driven automation platform that orchestrates standard EDA tools across the full chip design flow—from RTL entry and verification to physical implementation and packaging—using autonomous agents and GPU‑accelerated optimization engines. The platform applies reinforcement‑learning, evolutionary search, and surrogate modeling to explore high‑dimensional design spaces and meet power, performance, and area targets with minimal human intervention, while offering a natural‑language interface for task specification. It is delivered under a perpetual software license with optional professional support and can be deployed on‑premise or via secure cloud APIs.

Akhetonics develops an all-optical XPU, a general-purpose processor designed for ultra-low power, high-performance computing and AI applications. This platform integrates digital, analog, and quantum computing within a single photonics architecture, eliminating electronic conversion for data processing. The resulting processors offer significant speed and efficiency advantages by operating data entirely in the optical domain at THz clock speeds.

Unconventional AI develops new computing foundations specifically designed to optimize energy efficiency for artificial intelligence applications. The company focuses on achieving biology-scale efficiency by integrating expertise across AI systems, analog circuits, and computing theory. This approach aims to significantly reduce the power consumption associated with large-scale AI models.

Artemis is an ultra‑low‑power ASIC accelerator that operates at near‑threshold voltage (≈0.275 V) and uses an asynchronous massively parallel architecture to deliver AI inference with about 35 % lower power consumption than comparable HPC accelerators. The platform integrates on‑chip ADCs, vector extensions, and a CXL‑compatible interconnect to enable near‑data processing and coherent memory access for hyperscale data‑center and edge AI workloads.

AnDAPT develops programmable AmP chips that integrate up to 10 power rails and both analog and digital components into a compact 5mm x 5mm package, specifically designed for FPGAs and SoCs. This technology reduces PCB space by over 17% and enables rapid customization of power solutions, significantly decreasing design time and accelerating time-to-market for electronic products.