CEVA
CEVA is a semiconductor IP company that licenses signal processing platforms, Artificial Intelligence (AI) processors, wireless connectivity IP, and related software and tools for integration into system-on-chip (SoC) designs.
- Licensable Data Stewardship Program (DSP) and AI processor IP for audio, voice, imaging, vision, sensing, and general signal processing workloads (compute IP)
- Wireless connectivity IP for Bluetooth, Wi‑Fi, Narrowband Internet of things (NB-IoT), UWB, and other short-range and cellular Internet of Things (IoT) standards (connectivity IP)
- Sensor fusion, perception, and localization software stacks for edge devices in automotive, mobile, consumer, and industrial markets (edge software)
- Development tools, software libraries, and SDKs supporting integration, programming, and optimization of CEVA-based SoCs (developer tooling)
- Customer support, design services, and integration assistance for OEMs and semiconductor vendors adopting CEVA IP (professional services)
More About CEVA
CEVA focuses on licensable intellectual property (IP) cores and platforms that semiconductor companies and OEMs integrate into custom system-on-chip (SoC) designs for applications across mobile, consumer electronics, automotive, industrial, and IoT domains. Its portfolio centers on digital signal processing (DSP) and AI accelerators (compute IP), along with wireless communication and connectivity IP (connectivity IP) and associated software frameworks. Customers embed CEVA IP into SoCs used in devices such as smartphones, wearables, smart home products, automotive Advanced Driver Assistance System (ADAS) units, and industrial sensors.
On the compute side, CEVA offers DSP and AI processor architectures (AI infrastructure) optimized for parallel, low-power execution of signal processing and Machine Learning (ML) workloads at the edge. These cores are typically used for audio and voice processing, sensor fusion, computer vision, imaging, radar and lidar signal chains, and other real-time tasks. The architectures commonly support standard toolchains based on C/C++, intrinsics, and proprietary instruction set extensions, along with support for common Neural Network (NN) frameworks via conversion tools and optimized libraries.
In wireless connectivity, CEVA supplies IP for Bluetooth, Wi‑Fi, NB-IoT, UWB, and other radio technologies (wireless connectivity). This IP often includes baseband controllers, RF-related hardware blocks, protocol stacks, and reference software that comply with 3rd Generation Partnership Project (3GPP), IEEE 802.11, Bluetooth Streaming Inference Gateway (SIG), and other relevant standards. Enterprise and device manufacturers use this IP to integrate standards-based wireless connectivity into SoCs for products such as access points, client devices, IoT endpoints, and automotive telematics units, instead of sourcing discrete chipsets.
CEVA also develops software stacks and reference solutions (edge software) that run on its DSPs and AI cores, including sensor fusion algorithms, embedded perception and localization modules, and application-specific frameworks for audio, voice, and vision workloads. These software offerings are designed to shorten development cycles and provide pre-validated building blocks that map to CEVA’s hardware IP. Toolchains, SDKs, simulators, and debuggers (developer tooling) support system architects, firmware developers, and algorithm engineers in profiling, optimizing, and validating workloads on CEVA-based SoCs.
In enterprise and institutional environments, CEVA’s IP is typically used indirectly through silicon vendors and tier-1 OEMs who integrate the IP into chips and modules. This places CEVA within marketplace categories such as compute IP and AI accelerators for edge inference, wireless connectivity IP for standards-based communications, and embedded software stacks for perception and sensor fusion. The technical and business relevance of CEVA’s offerings stems from their role in enabling SoC designers to implement domain-specific compute and connectivity capabilities without building these complex subsystems from the ground up, while adhering to industry protocols and power, performance, and area constraints common in edge and IoT deployments.