BOS Semiconductors is heading to CES 2026 with a clear message for automakers: scaling in-vehicle AI doesn’t have to mean ripping out existing platforms.
The South Korea–based fabless semiconductor company announced it will exhibit at CES 2026 in Las Vegas from January 6–9, where it plans to showcase an AI Box demo for next-generation mobility, built around its Eagle-N AI accelerator.
The AI Box is designed as an external AI computing module that allows automotive OEMs to add high-performance AI capabilities to vehicles without replacing existing in-vehicle infotainment (IVI) or electronic systems. BOS positions the product as a faster, lower-disruption path to deploying advanced AI features across both new vehicles and updated “facelift” models.
Targeting Software-Defined Vehicles and Physical AI
At CES, BOS will highlight how its approach aligns with broader automotive trends, including autonomous driving, Software Defined Vehicles (SDVs), and the rise of Physical AI—AI systems that operate directly in real-world environments rather than relying heavily on cloud infrastructure.
According to the company, the AI Box supports a wide range of AI models based on CNN and Transformer architectures, enabling real-time perception and decision-making inside the vehicle. By handling AI-intensive workloads externally, the module allows existing vehicle systems to maintain their original roles while extending AI functionality independently.
On-Device AI Over Cloud Dependence
A key part of BOS’ pitch is its on-device AI architecture. Processing data such as voice and video directly inside the vehicle can reduce cloud dependency, improve reliability in low-connectivity environments, and strengthen data privacy, the company says [BOS Semiconductors_PR] BOS Sem….
BOS also claims that keeping inference on-device can lower long-term total cost of ownership by reducing cloud traffic, storage, and inference costs, while allowing AI features to scale without expanding backend server infrastructure.
Live Demos with VLMs and LLMs
At its CES booth (Venetian Expo, Hall A, Booth #50017), BOS will demonstrate AI models that combine Vision-Language Models (VLMs) and Large Language Models (LLMs) running on the AI Box architecture. The demo will also be shown at the Tenstorrent Demo Room during the event.
“Through our AI Box demo with Eagle-N at CES 2026, we will present a practical approach for effectively scaling AI capabilities even on existing mobility systems,” said Jason (Jeongseok) Chae, Vice President and Head of Strategic Marketing & Sales at BOS Semiconductors. He added that the company aims to expand beyond traditional automotive chips and position itself as a core player in the Physical AI era.
Beyond Automotive
While automobiles remain BOS’ primary market, the company says its chiplet-based architecture is designed to scale across broader Physical AI applications, including robotics and intelligent machines, using the same underlying product platform.
As automakers search for faster ways to deploy AI features without lengthy platform overhauls, BOS’ AI Box could appeal to manufacturers looking for incremental—but scalable—paths into AI-driven mobility.
