Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Model Context Protocol
AI Agents
AI Models
AI Foundations
AI Hardware
GenAI Applications
AI in Industries
Data Security
Identity & Access Management
Security Tools
Network Security
Threat Detection and Response (TDR)
Web Proxies
Web Data Scraping
Data Collection
Data Science
Synthetic Data
E-Commerce
Workload Automation
IT Automation
Process Improvement
RMM
Managed File Transfer
Observability
The demand for low-latency processing has driven innovation in edge AI chips. These processors are designed to perform AI computations locally on devices rather than relying on cloud-based solutions.
Based on our experience analyzing AI chip makers, we identified the leading solutions for robotics, industrial IoT, and embedded systems.
*TOPS = Tera Operations Per Second. These are maximum quoted values by vendors.
**Kria K26 performance varies depending on the FPGA configuration.
NVIDIA Jetson AGX Orin delivers 275 TOPS, positioning it as the highest-performance edge AI module currently available. The module is built on NVIDIA’s Ampere architecture and is designed for robotics and autonomous systems that require significant on-device processing capabilities.
Key specifications:
The power consumption range of 10-60W provides flexibility for different deployment scenarios. Lower power modes can extend battery life in mobile robotics applications, while maximum performance mode supports multiple concurrent AI workloads.
NVIDIA’s software ecosystem represents a significant advantage. Models developed for NVIDIA datacenter GPUs can be deployed on Jetson with minimal modification. This compatibility reduces development time for teams working within the NVIDIA ecosystem.
Axelera’s Metis AI platform delivers up to 214 TOPS for high-throughput vision inference workloads. The platform uses Digital In-Memory Computing (D-IMC) architecture to improve throughput and efficiency.
Key specifications:
The D-IMC architecture performs computations directly within memory arrays, reducing data movement between memory and processing units. This approach addresses the memory bandwidth bottleneck that limits performance in traditional architectures.
Axelera targets applications that require the simultaneous processing of multiple video streams. The high throughput enables real-time analysis of dozens of camera feeds from a single device.
Use cases:
Axelera received €61.6 million in funding from the EuroHPC Joint Undertaking in March 2025, supporting the development of their Titania chiplet for deployment by 2028.
Renesas RZ/V2H delivers up to 80 TOPS using sparse computation, or 8 TOPS at INT8, targeting robotics and factory automation that combine vision AI with real-time control on a single chip. The MPU integrates the third-generation DRP-AI3 accelerator alongside both application and real-time processor cores.
Key specifications:
The DRP-AI3 uses INT8 quantization and hardware-supported unstructured pruning to achieve its rated performance, enabling the chip to run object recognition without a heat sink. A separate dynamically reconfigurable processor handles image processing such as OpenCV operations, offloading that work from the CPU cores.
The combination of application cores and real-time cores allows the RZ/V2H to process image recognition results and apply them to mechanical control on the same device. This suits applications where vision inference and motion control run together.
Use cases:
NVIDIA Jetson Orin Nano Super delivers 67 TOPS with sparse operations or 33 TOPS with dense operations, positioning it below the Jetson AGX Orin as a lower-cost entry point to the same software stack.
Key specifications:
The 67 TOPS figure represents a 1.7 times increase over the previous Jetson Orin Nano, achieved through a power mode that raises the GPU, CPU, and memory clocks. Existing Orin Nano owners can apply the increase through a software update rather than buying new hardware.
The memory bandwidth and software compatibility allow the module to run small language models, vision transformers, and vision-language models on the device. Models developed for other Jetson and NVIDIA platforms transfer with minimal modifications, reducing development time for teams using the NVIDIA ecosystem.
Use cases:
EdgeCortix SAKURA delivers 60 TOPS with power consumption under 10W, targeting edge AI servers and high-performance vision applications. The platform features a reconfigurable architecture that adapts to different AI workloads.
Key specifications:
The SAKURA platform’s reconfigurable architecture allows optimization for different neural network topologies without hardware changes. This flexibility enables the deployment of emerging model architectures without requiring chip replacements.
Use cases:
SiMa.ai’s MLSoC (Machine Learning System-on-Chip) delivers over 50 TOPS while maintaining power consumption below 5W. The chip targets embedded vision applications requiring high performance in power-constrained environments.
Key specifications:
SiMa.ai designed the MLSoC specifically for computer vision workloads. The sub-5W power envelope enables deployment in battery-powered devices that require sustained high-performance inference.
Use cases:
Hailo-10H delivers 40 TOPS at INT4, or 20 TOPS at INT8, while consuming 2.5W typical, extending Hailo’s vision-focused hardware to generative AI workloads. The accelerator uses the company’s second-generation neural core architecture in an M.2 module form factor.
Key specifications:
A direct DDR interface, which the Hailo-8 lacks, enables the Hailo-10H to load larger models for language and vision-language inference. Hailo reports that the chip runs 2-billion-parameter models at around 10 tokens per second, generates images with Stable Diffusion 2.1 in under five seconds, and runs object detection on a 4K video stream for conventional vision tasks. These figures are vendor-reported rather than independently measured.1
The Hailo-10H is qualified to AEC-Q100 Grade 2 for automotive use, with vehicle production targeted for 2026. It became available to order in mid-2025 as standalone M.2 modules.
Use cases:
Hailo-8 delivers 26 TOPS while consuming 2.5-3W, representing one of the highest performance-per-watt ratios among edge AI chips.
Key specifications:
The chip supports standard neural network layers and can run models developed in TensorFlow, PyTorch, and ONNX. Hailo’s compiler.
Ambarella’s CV5 system-on-chip delivers over 20 TOPS specifically optimized for computer vision in automotive and camera applications. The chip combines AI processing with advanced image signal processing (ISP) capabilities.
Key specifications:
The CV5’s integrated ISP handles complex image preprocessing, reducing the computational burden on the AI engine. This integration improves overall system efficiency for vision-based applications.
Use cases:
Qualcomm’s Robotics RB5 platform integrates 5G connectivity with edge AI processing, delivering approximately 15 TOPS through its Qualcomm AI Engine. The platform targets autonomous robots and drones that require both high-bandwidth connectivity and on-device AI processing.
Key specifications:
The integration of 5G offers high-bandwidth, low-latency connectivity for applications that require real-time cloud communication.
The RB5 platform supports up to 7 concurrent camera inputs. This multi-camera capability supports 360-degree perception systems for autonomous mobile robots.
Use cases:
Kneron’s KL730 AI SoC delivers 7 TOPS with ultra-low power consumption, targeting IoT and smart home applications. The chip emphasizes edge processing for privacy-sensitive applications.
Key specifications:
The KL730’s low power consumption enables always-on AI processing in battery-powered devices. The chip supports face recognition, object detection, and gesture recognition with minimal energy draw.
Use cases:
The RK3588 is an 8-core SoC featuring a 6 TOPS neural processing unit. The chip targets single-board computers and edge devices requiring moderate AI performance alongside general-purpose computing capabilities.
Key specifications:
The 6 TOPS NPU handles neural network inference for computer vision, natural language processing, and audio processing tasks.
Use cases:
The RK3588’s general-purpose computing capabilities make it suitable for applications where AI inference is one component of a larger system. Organizations building edge devices that combine AI with web servers, databases, or other software services have adopted this SoC.
NXP’s iMX 8M Plus features a 2.3 TOPS neural processing unit, designed specifically for industrial IoT applications. The processor prioritizes reliability, security, and long-term availability over maximum performance.
Key specifications:
The inclusion of a Cortex-M7 real-time core enables deterministic processing for time-critical control loops. This architecture supports applications that combine AI-based decision-making with real-time control, such as industrial robots and automated manufacturing equipment.
NXP’s EdgeLock security features provide hardware-based secure boot, encrypted storage, and secure key management.
Use cases:
Renesas RZ/V2L delivers 1.0 TOPS optimized for industrial vision applications with extremely low power consumption. The chip targets factory automation and quality inspection systems.
Key specifications:
The DRP-AI architecture provides flexibility for different vision algorithms while maintaining low power consumption. This design suits industrial environments requiring long-term reliability and deterministic performance.
Use cases:
The Kria K26 System-on-Module combines a Zynq UltraScale+ MPSoC with FPGA fabric, enabling adaptive edge AI solutions. The FPGA architecture enables customization of the processing pipeline for specific computer vision and sensor-fusion workloads.
Key specifications:
AMD provides pre-built vision AI applications through the Kria KV260 Vision AI Starter Kit. These applications include smart camera implementations with capabilities for object detection, classification, and tracking.
Advantages:
Limitations:
Edge AI chips face a tradeoff between performance and power consumption.
High Performance (>50 TOPS):
These solutions target applications where AI performance is the primary requirement. Use cases include autonomous vehicles, industrial robotics, and multi-camera video analytics systems.
Balanced Performance (15-30 TOPS):
Balanced solutions optimize the performance-per-watt ratio. These chips are suitable for applications where both performance and power consumption are constrained, such as battery-powered robots and smart cameras.
Low Power (<10 TOPS):
Low-power solutions prioritize energy efficiency over raw performance. IoT devices, battery-powered cameras, and embedded systems with limited thermal budgets typically use these chips.
The selection of appropriate hardware depends on:
Software support has a significant impact on the practical performance and development time for edge AI deployments.
NVIDIA Jetson supports the full CUDA ecosystem. Models developed for NVIDIA data center GPUs can be deployed with minimal modification. This compatibility reduces development time for teams using NVIDIA hardware.
Google Edge TPU requires TensorFlow Lite models with int8 quantization. While this limitation ensures optimal performance on the TPU, it requires model conversion and validation steps. Organizations not using TensorFlow may face additional development work.
Intel Movidius integrates with the OpenVINO toolkit, which supports multiple model frameworks. The toolkit’s optimization capabilities can significantly improve inference performance, but require learning Intel-specific tools.
AMD Xilinx Kria demands FPGA development expertise for custom implementations. While pre-built vision AI stacks reduce this requirement, organizations that seek custom processing pipelines require specialized skills.
Qualcomm, Hailo, and other vendors provide their own SDKs and model compilers. Development teams should evaluate these tools during the selection process to understand the required effort for model deployment and optimization.
Edge AI chips are available in multiple form factors to address different integration requirements:
System-on-Module (SoM):
SoM provides a complete computing module that can be integrated into custom carrier boards. This approach reduces hardware design complexity while enabling customization of I/O interfaces.
M.2 and PCIe Cards:
M.2 and PCIe form factors enable adding AI acceleration to existing systems. This approach is suitable for applications that upgrade existing hardware platforms with AI capabilities.
USB accelerators:
USB accelerators provide the simplest integration path. These devices are suitable for prototyping, development, and applications where the host system has available USB ports and sufficient bandwidth.
Integrated SoC:
Integrated SoCs combine CPU, GPU, and NPU in a single chip. This integration reduces board complexity and cost for products designed around the specific SoC.
Robotics and autonomous systems: NVIDIA Jetson AGX Orin or Qualcomm RB5 provide the performance required for real-time navigation, object detection, and path planning. The choice depends on whether 5G connectivity is a requirement.
Industrial IoT and factory automation: NXP i.MX 8M Plus or AMD Xilinx Kria K26 address the security and real-time processing requirements common in industrial applications. The Kria platform suits applications requiring custom sensor interfaces or deterministic latency.
Smart cameras and video analytics: Hailo-8 or Axelera Metis deliver the performance-per-watt ratio required for always-on video processing. Hailo-8 suits single or few-camera deployments, while Axelera Metis targets multi-camera systems.
Battery-powered IoT devices: Google Edge TPU provides the lowest power consumption for applications where battery life is the primary constraint. The 2W power consumption enables extended operation from small batteries.
Drones and AR devices: Intel Movidius Myriad X or SiMa.ai MLSoC balance performance with power consumption for airborne and wearable devices. The weight and thermal constraints in these applications favor efficient solutions.
Automotive applications: Ambarella CV5 or Qualcomm platforms offer the necessary automotive-grade certifications and performance for ADAS and autonomous driving applications.
Development and prototyping: Intel Neural Compute Stick 2 or Google Coral USB Accelerator enable quick evaluation of edge AI capabilities without hardware modifications. These USB devices are suitable for proof-of-concept projects and algorithm development.
Specialized AI chips, including cutting-edge AI chips and other AI accelerators, are designed to run AI models, AI algorithms, and deep neural networks directly on local devices. This shift toward processing data locally reduces cloud or data center overhead. It lowers cloud dependence, which is crucial for real-time data processing, analytics, and decision-making in edge AI applications.
By keeping sensitive data on local devices, organizations can improve security while enabling AI at the edge for various use cases, including object detection, anomaly detection, predictive maintenance, face recognition, and smart city applications. Specialized edge AI technology also enables low-power consumption, low-power computing, and reduced operational costs, which are important factors in embedded AI hardware and AI devices used in robotics, industrial IoT, and other edge environments.
Edge AI technology runs machine learning models, generative AI, and other AI applications directly on specialized hardware such as AI accelerators or a single chip (e.g., a single Metis chip). Unlike cloud AI, which depends on remote servers, AI at the edge focuses on local processing, where data is processed locally using AI inference.
This architecture reduces latency, improves decision-making, and enhances AI capabilities for time-critical uses like real-time monitoring, real-time processing, and managing safety hazards in business operations. Running AI on edge devices also reduces operational expenses, optimizes bandwidth usage, and helps organizations improve efficiency, optimize operations, and boost operational efficiency, especially in environments where continuous connectivity to a remote data center is not guaranteed.
AI accelerators and cutting-edge AI chips enable a wide range of typical applications that rely on AI inference, machine learning, and artificial intelligence running outside the cloud. These include object detection in smart cameras, detecting anomalies in industrial systems, predictive maintenance for equipment, and natural language interfaces on local devices.
Industries such as robotics, autonomous systems, industrial automation, and smart cities benefit from bringing AI closer to sensors for real-time decision making. With low-power consumption designs and support for different models of AI workloads, including large language models and vision-based workloads, edge systems become more cost-effective and help organizations reduce operational expenses. Whether using central processing units with integrated NPUs or advanced AI-specific architectures with minimal reliance on external memory, edge solutions allow AI to run efficiently on a single chip and enable next-generation AI at the edge deployments.
Pick the format that matches where you're publishing. Pasting the link version into your CMS preserves the backlink.
HTML, for blog posts, LinkedIn articles & newsletters. Recommended.
For academic papers and analyst reports following APA 7th style.
For LaTeX documents and academic reference managers.
Results and timestamps of 15 data points. Download the data used in this article as a ZIP file containing one CSV file.
Replaced the Google Coral Dev Board with the Axelera Metis AI Platform in the list of edge AI chip makers.
Your email address will not be published. All fields are required. Comments are left in their original language.
We follow ethical norms & our process for objectivity. This research does not mention any subscribers of AIMultiple's benchmarking services.
