Intel is extending its reach beyond PCs and servers — leveraging its advanced 18A process and high‑performance chips — into the automotive space. The smart cockpit is becoming one of the most representative frontiers.
🎙️ Wang Jingjia · Intel China
“As vehicles become electric and intelligent, our expectations have changed. A car used to be a passive means of transport. Today, it also carries our work, life, and entertainment. With the AI wave, we want to bring differentiated in‑cabin experiences through AI.”
At the recent Chain Expo, Intel and Changan Automobile jointly unveiled an AI cockpit solution powered by the Intel AI Box Ultra. The platform combines robust on‑device AI with a mature Android ecosystem, delivering safer, smarter, and more personalised cockpit experiences.
A “needed” platform · Core Ultra
The Intel Core Ultra platform — a heterogeneous SoC integrating CPU, GPU, and NPU — is central to this achievement. It delivers up to 16 x86 cores for smooth entertainment, productivity, and agent orchestration, while the combined 180 TOPS of AI compute handles diverse workloads. Typical power consumption is just 25W, and the NPU provides low‑power AI for always‑on scenarios.
🧑💼 Cao Bin · Changan Technology & CTO of Wutong Technology
“Over the past few years, automotive intelligence has centred on ADAS and the smart cockpit. As ADAS matures, the cockpit comes to the forefront — the car becomes a ‘third space’ for mobility, work, entertainment, and human‑machine interaction.”
“The future cockpit must understand user needs, assist with route planning, offer companionship, and support office scenarios. The core enabler is AI.”
However, cockpit chip AI compute is limited; many functions rely on cloud LLMs and APIs, which are constrained by network quality, service stability, and data privacy — especially as regulations tighten around in‑cabin voice and camera data.
Cao Bin believes on‑device LLMs are the inevitable direction. Changan aims to run 4B–7B multimodal models locally, building a hybrid cloud‑edge AI framework. The traditional cockpit architecture can no longer keep up with growing compute demands.
That’s why Changan proposed the AI Box — a dedicated compute unit beside the cockpit to supply scalable AI power. Intel’s Core Ultra platform became the preferred choice.
AI Box · three pillars
- On‑device LLM deployment — local execution of large models
- Massive ecosystem — full Android and PC application support
- Premium audio‑visual experience — immersive in‑car media
💬 Wang Jingjia · “The ideal SoC for automotive AI Box”
“This SoC, born in the PC world, has expanded to edge, mobile, and automotive — supporting both Android and PC ecosystems. It unlocks limitless application possibilities. We believe it’s the most suitable SoC for automotive AI Box.”
Multi‑agent orchestration & model compatibility
The AI Box is a powerful multi‑agent orchestration engine with broad model support: from 35B MoE models down to 4B–9B Omni multimodal and VLM models. It deeply integrates ASR, KMS, TTS, and voiceprint recognition, while fully leveraging NPU for efficient computer vision — delivering high AI throughput even in low‑power scenarios.
🚗 Use case: smart travel & family Vlog
Using in‑vehicle cameras and AI‑generated content (AIGC), the system automatically blends scenery with passengers’ expressions to create story‑rich Vlogs. The driver is no longer a “hidden” figure — the car becomes a creative studio that captures and generates memories.
Thanks to the PC‑grade chip, AAA games can run natively in the car via Bluetooth peripherals. Desktop productivity tools also run smoothly, breaking the boundary between in‑car entertainment and business. The system deeply integrates the Android ecosystem — apps like Jianying (CapCut) can run locally, enabling instant clip‑capture, editing, and distribution.
Moreover, the platform acts as a future‑ready Agent PC / AI PC. With a “cloud‑edge” and “Agent + Skill” model, users can build atomic agent capabilities at low cost — turning the system into a proactive AI super‑assistant with multi‑modal perception.
The platform also doubles as an AI NAS (intelligent network storage), supporting local storage of personal videos, texts, and photos.
🔒 Privacy & continuous evolution
Wang Jingjia highlighted two core benefits: extreme privacy — all core data stays local, eliminating leakage risks; and on‑device agent evolution — the more local data accumulated, the smarter the AI becomes, continuously learning and adapting to the user.
Five strengths: 强 (powerful) · 快 (fast) · 安 (safe) · 通 (connected) · 省 (cost‑efficient)
- Powerful — AI PC‑grade platform, strong model support
- Fast — robust on‑device compute performance
- Safe — all data stored locally, privacy guaranteed
- Connected — unifies PC, Android, and agent ecosystems
- Cost‑efficient — annual chip iteration enables “board‑level upgrades” at lower cost, keeping pace with AI evolution