Ceva’s NeuPro positioned as foundational IP for on‑device AI, licensing fueling 2026 royalties

Ceva positions NeuPro as foundational IP for real‑world AI

NeuPro licences drive on‑device AI expansion

Ceva is rapidly scaling its NeuPro neural processing unit (NPU) business, saying 2025 is a breakthrough year after signing 10 NPU agreements and reporting that AI contributes more than 20% of its annual licensing revenue. The Rockville, Maryland‑based semiconductor IP firm says demand for power‑efficient, optimized architectures for edge and hybrid AI models is driving broader adoption of on‑device inference across consumer, automotive, industrial and PC markets, positioning the company as an enabler of what it terms "Physical AI" — where connectivity, sensing and inference converge inside devices.

A strategic licensing agreement with one of the world’s leading PC original equipment manufacturers validates Ceva’s scalable NPU approach, the company says, with the OEM selecting NeuPro as foundational IP for its next‑generation on‑device AI compute architecture. Ceva also signs a comprehensive NeuPro portfolio licence with Microchip Technology and wins with ALi Corp and Nextchip, expanding its footprint into consumer electronics, video platforms and automotive ADAS applications. The deals cover a broad performance range from ultra‑low‑power to high‑performance inference for audio, vision, video and multi‑sensor workloads.

Ceva expects the new licences to convert into silicon and royalties as customers move from AI evaluation to high‑volume deployment. Several agreements involve existing high‑volume connectivity customers increasing content per device, which Ceva says boosts long‑term royalty potential. The company reports that six customers are expected to have silicon back by the end of 2026, creating a runway for royalty ramps beginning in 2026 and strengthening visibility for longer‑term revenue and market‑share gains.

Automotive DSP deals complement NPU momentum

Alongside NeuPro traction, Ceva signs multiple AI digital signal processor licences across automotive applications in 2025. The company positions these DSP deployments as complementary building blocks for in‑vehicle inference and sensor fusion, with expected royalty contributions as programs move into production.

Broader market implications and timing

Ceva frames the 2025 wins as a shift in the industry as AI extends beyond the cloud into devices, arguing that power‑efficient on‑device inference is becoming a design priority across PC, automotive, consumer and infrastructure platforms. The firm anticipates measurable royalty ramps beginning in 2026 and positions its IP portfolio to capture increased compute content per device as manufacturers embed more on‑device AI.