CEVA, Inc. plays a low-profile role in the semiconductor value chain: the company does not manufacture the chips found in electronic devices, but develops some of the architectures they use. This U.S. company supplies intellectual property blocks, or IP, for integrated circuit designers. Its technologies cover wireless communications, audio processing, sensor data processing and embedded artificial intelligence, among other areas.
A history rooted in signal processing
CEVA was founded in 2002 through the merger of DSP Group’s intellectual property licensing business and Parthus Technologies. These origins bring together expertise in digital signal processors, or DSPs, and technologies for embedded electronic systems. Listed on Nasdaq under the ticker symbol CEVA, the company follows the IP supplier model: its teams design reusable elements rather than finished components.
As the company has developed, its portfolio has expanded beyond DSPs to cover more connectivity, sensing and computing functions. This evolution has accompanied the proliferation of connected devices and the shift of certain processing tasks to the devices themselves.
Licensed technologies for chip designers
CEVA’s customers can integrate its blocks into systems on a chip, alongside general-purpose processors, memory and other functions. The benefit is access to technologies that have already been developed and documented, while retaining responsibility for the component’s overall architecture. The offering also includes the software and development tools needed to use them.
In connectivity, CEVA offers technologies related to Bluetooth, Wi-Fi and ultra-wideband, or UWB, among others. In embedded computing, its DSPs and neural network acceleration architectures address audio processing, vision and data analysis needs. These building blocks target a variety of markets: consumer electronics, connected devices, automotive and industrial equipment.
The business model combines licensing revenue with royalties based on shipments of products incorporating its technologies. CEVA therefore operates across two time horizons: its customers’ design decisions, followed by the ramp-up in volumes of chips actually brought to market.
What comes next?
The development of embedded AI opens up opportunities for suppliers capable of balancing performance, power consumption and memory constraints. For CEVA, the challenge is to secure adoption of its architectures in devices where local processing can reduce latency and data exchanges with the cloud.
This trajectory will, however, depend on manufacturers’ technology choices, competition among IP suppliers and in-house development by major companies. Software quality, ease of integration and long-term support will matter as much as hardware capabilities. Company website: ceva-ip.com.