Intel's Latest Core Platform Empowers AI Box, Bringing AI Accelerator Engine to Automotive and Industrial Scenarios

Intel Announces Launch of AI Box Solution Based on the Latest Intel Core Ultra Architecture, Bringing PC-Grade Computing Power to Automotive, Industrial Automation, Rail Transit, Robotics, and Other Industrial Environments, Providing a More Efficient and Flexible Path for Industries to Access AI Large Models. Currently, Intel has partnered with Tier 1 suppliers such as ADAYO Huayang General to develop platform-based solutions and has successfully secured core intelligent projects with leading domestic autonomous vehicle brands, completing the transition from technical validation to commercial application.

The advent of AI has brought more imagination to traditional industries like automotive. What was once a single tool has become an intelligent partner capable of seeing, hearing, and understanding. However, this also sharply increases the demand for local computing power. The AI Box, with its open, flexible, and scalable features, provides a “plug-in” AI processing capability with stronger local AI handling power for rapid commercialization of large models, becoming another brain for devices.

Heterogeneous Collaboration, Unleashing Powerful AI Computing Power

The new AI Box launched by Intel is equipped with the Intel Core Ultra processor. Its core advantage comes from the unique heterogeneous design of the Core platform—integrating CPU, GPU, and NPU engines to schedule computing power for different AI tasks.

In processing complex tasks, only when CPU and GPU work in balanced collaboration can efficient data supply be achieved, maximizing the value of each computing resource. Specifically, tasks such as Android applications, natural language understanding, and task scheduling, which require heavy computation, are handled by the CPU. The NPU acts as a dedicated AI inference engine, providing stable and strong inference capabilities at low power consumption. Visual tasks like local large model inference and environment perception are efficiently accelerated by the GPU, which also offers additional 3D rendering and gaming capabilities. Multi-modal interaction tasks can be dynamically scheduled across the three engines, ensuring smooth operation without lag.

The Core Ultra platform, with its powerful hardware computing power and deep optimization, can easily handle more complex and demanding AI application scenarios, providing comprehensive support for edge intelligence: supporting various machine vision models, adapting to dense architectures suitable for edge deployment, and compatible with large language models (LLMs) with MoE (Mixture of Experts) structures. It also offers stable support for multi-modal VL models with 7B/8B parameters, and through performance tuning, meets the strict latency and throughput requirements of cockpit scenarios, ensuring smooth and reliable intelligent interactions.

The Core Ultra platform is the first to natively support full-modal Omini models, breaking the limitations of single-modal interaction, making edge AI interactions more humanized, richer, and closer to real needs. Its high-concurrency intelligent task scheduling capability makes multi-scenario and multi-task parallel processing more efficient, enabling better handling of complex AI tasks and expanding the application boundaries of edge AI.

The Core Ultra platform is an important carrier for Intel’s deployment across various industries. Whether in automotive cockpits, robotics, rail transit edge devices, or under conditions of high temperature, vibration, and electromagnetic interference, the platform maintains stable performance, meeting customers’ comprehensive requirements for computing power, energy efficiency, and durability.

Open Ecosystem, Promoting Industry Intelligence Upgrades

The advantages of Intel AI Box go beyond its heterogeneous architecture; it also features software compatibility across Linux and Android, inheriting the rich AI PC ecosystem. After installing the Intel AI Box, it can seamlessly integrate with Android ecosystems, meeting the needs of automotive cockpits and smart terminals, and quickly porting rich AI PC applications to vehicles or other scenarios. The solution natively supports mainstream Linux distributions, enabling rapid integration of the latest AI workloads on devices.

For developers, Intel provides tools like OpenVINO, significantly reducing the difficulty of developing AI capabilities for deployment and shortening product development cycles, allowing feature-rich industry AI applications to reach the market quickly.

To enable users to access the latest AI applications instantly, Intel has deeply participated in the construction of China’s large model ecosystem, collaborating with multiple AI vendors to enable Day 0 support for mainstream large models and deeply optimizing for multimodal scenarios such as text, speech, and vision. The Intel AI Box also offers enterprise-level local deployment capabilities, addressing security concerns in fields like industry and transportation that require high data security.

This “hardware + software + tools” support gives the AI Box the ability to upgrade via AI plugins, serving as a flexible entry point for integrating AI into both new and old devices—allowing older devices, even if they were shipped without intelligent systems, to be revitalized with AI Box. For new devices, the AI Box reserves ample computing redundancy, providing automakers and robotics manufacturers with better AI services.

"

Intel aims to bring mature AI capabilities from PC AI to more industries. Through an open ecosystem, powerful computing power, and stable architecture, every vehicle, industrial equipment, and space can enjoy a more natural and powerful AI experience. Intel will continue to deepen technological innovation on the Core platform, continuously improve AI Box solutions, and inject new momentum into the intelligent upgrading of thousands of industries.

— Guo Wei

Vice President of Intel’s Marketing Group

General Manager of China

Intel, Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands mentioned are the property of their respective owners.