Ceva's AI Surge: Powering the Physical AI Revolution at the Edge

ROCKVILLE, MD – February 17, 2026 – By Carol Moore

Ceva, Inc., a key player in licensing the silicon and software IP that powers smart devices, has announced a landmark year for its artificial intelligence business. In 2025, the company secured 10 new licensing agreements for its NeuPro™ neural processing units (NPUs), with AI-related deals now accounting for over 20% of its total licensing revenue. This financial milestone signals more than just a successful year for Ceva; it marks a significant inflection point for the entire technology industry as AI processing moves decisively from the distant cloud to the devices in our hands, homes, and vehicles.

The surge in demand for on-device AI is fueling Ceva's vision of "Physical AI," a concept where connectivity, environmental sensing, and intelligent inference converge within a single device. This shift is creating a new paradigm for everything from personal computers and smartphones to industrial robots and autonomous cars, all requiring power-efficient, localized intelligence.

The Inevitable Shift to On-Device Intelligence

For years, heavy-duty AI processing has been the domain of massive, power-hungry data centers. However, the proliferation of Internet of Things (IoT) devices and the demand for real-time responsiveness have exposed the limitations of a cloud-only approach. Sending data to the cloud for analysis introduces latency, consumes network bandwidth, raises privacy concerns, and is often impractical for battery-powered devices.

The industry's solution is edge AI, where processing occurs directly on the device. This approach offers near-instantaneous results, enhances data privacy by keeping sensitive information local, and significantly reduces power consumption. Ceva is positioning itself at the heart of this transition. The company's IP portfolio, which includes not only AI processors but also essential connectivity tech like 5G, Wi-Fi, and Bluetooth, provides the foundational building blocks for these next-generation smart edge products.

"2025 marked a pivotal year for our AI strategy," commented Amir Panush, Chief Executive Officer of Ceva, in a statement. "The breadth of NPU and AI DSP licenses we signed—from microcontrollers and PCs to automotive and consumer devices—demonstrates that AI at the edge is moving from experimentation to deployment."

Strategic Wins Cement Market Position

A key highlight underscoring this trend was Ceva's strategic agreement with what it describes as "one of the world's leading PC original equipment manufacturers." While the OEM's name remains confidential, the deal is a major validation of Ceva's high-performance NeuPro NPU architecture. This partnership will see Ceva's IP become a foundational component in the OEM's next-generation on-device AI compute architecture. The move reflects a broader push in the PC market, driven by giants like Microsoft and Apple, to integrate dedicated AI capabilities into operating systems, creating a new class of "AI PCs" that can handle complex tasks like real-time translation and advanced content creation without relying on the cloud.

Further demonstrating the breadth of its AI adoption, Ceva also signed a comprehensive portfolio license with Microchip Technology, a global leader in microcontrollers and connectivity solutions. By adopting Ceva's NPUs, Microchip plans to embed AI capabilities across its vast product families. This partnership is significant because it brings sophisticated AI inference to a wider range of smaller, power-constrained devices used in industrial, automotive, and consumer applications. It signals a trend toward standardizing on proven NPU architectures as AI becomes a core, expected feature across the entire semiconductor landscape.

Additional wins with companies like ALi Corp for consumer electronics and Nextchip for automotive Advanced Driver Assistance Systems (ADAS) further illustrate the diverse market penetration of Ceva's AI technology, from ultra-low-power audio processing to high-performance computer vision.

A Financial Inflection Point and Future Royalties

The strategic wins of 2025 have translated into a material shift in Ceva's revenue mix. With AI now contributing more than a fifth of licensing revenue, the company has established a powerful new growth engine. This success is built on a dual-revenue IP model: an upfront licensing fee, followed by ongoing royalties paid on every chip shipped that contains Ceva's technology.

The 54 total licensing agreements Ceva signed in 2025, including the 10 NPU deals, represent an estimated $125 million in aggregate lifetime royalty opportunity. This provides investors with a clearer long-term view of the company's revenue potential. According to the company, six of these new AI customers are expected to have finished silicon designs ready by the end of 2026, with initial royalty streams projected to begin flowing as early as 2027.

This AI momentum contributed to a solid financial year overall, with Ceva reporting a 2% rise in total revenue to $109.6 million for 2025 and projecting 8-12% growth for 2026. The company's focus on the smart edge, which now generates 86% of its revenue, appears to be paying significant dividends.

Navigating a Competitive AI IP Landscape

Ceva's success comes amidst a fiercely competitive market for AI IP. The company contends with industry heavyweights like Arm, with its Ethos NPU series, and Synopsys and Cadence, which also offer scalable NPU solutions for the edge. The global NPU IP market, valued at an estimated $172.5 million in 2025, is projected to grow to over $600 million by 2035, attracting intense competition.

In this crowded field, Ceva differentiates itself by offering a holistic platform for "Physical AI." It provides not only the AI inference engines (NPUs and AI DSPs) but also the critical connectivity and sensor fusion IP needed to make devices truly intelligent and aware of their surroundings. This ability to offer an optimized, integrated package of connect, sense, and infer technologies allows customers to reduce system costs, improve power efficiency, and accelerate their time to market.

As the world moves toward an era of pervasive intelligence, the foundational IP that enables devices to understand and interact with the physical world becomes increasingly critical. With its recent string of successes and a clear strategy for the smart edge, Ceva has firmly established itself as a central architect of this unfolding technological revolution.