Header photo source: Intel.
Intel is pushing deeper into the smart car space, and it sees China as a core part of that strategy. That was the message from Jack Weast, vice president and general manager of Intel’s automotive division, during the company’s first appearance at Auto Shanghai on April 23. Speaking at a media roundtable, Weast emphasized how central the Chinese market has become to Intel’s automotive ambitions.
But with Qualcomm and Nvidia dominating the in-car chip market for infotainment and driver assistance, Intel faces an uphill battle. Can it find space to grow?
Earlier this year, Intel launched its first-generation software-defined vehicle (SDV) cockpit system-on-chip (SoC), based on the same architecture as its PC-grade Core chips. Still, the cockpit chip market in China is fiercely competitive. Qualcomm remains the leader, and domestic players like SiEngine and SemiDrive have carved out strong positions. So far, Intel’s offering has made little impact.
Now, Intel is trying a different approach.
At Auto Shanghai, the company unveiled a joint solution with Black Sesame Technologies: a cockpit-driving fusion platform. The system integrates Intel’s cockpit SoC with Black Sesame’s Huashan A2000 and Wudang C1200 chips. The Huashan A2000 is designed for comprehensive assisted driving scenarios, while the Wudang C1200 targets cross-domain computing. Together, the system is intended to support driving automation from Level 2 to Level 4. A joint task force is already in place, with a reference design expected in the second quarter and mass production preparations underway.
Intel is also collaborating with artificial intelligence startup ModelBest to develop a fully on-device graphical user interface (GUI) for vehicles. This system will operate offline and support voice command recognition, contextual memory, personalized recommendations, and touch-screen interaction. The focus is on enabling natural, multi-turn dialogue—all without relying on cloud connectivity.
By combining chips with different capabilities from Black Sesame, Intel aims to give automakers more flexible options across intelligent cockpit and driver assistance functions.
Still, Weast emphasized that the Black Sesame partnership isn’t exclusive, and that Intel is open to working with other vendors.
Upgraded hardware, modular architecture
Beyond partnerships, Intel is steadily enhancing its own automotive chips.
At the event, the company introduced its next-generation cockpit SoCs, claiming significant performance improvements. Compared to the first generation, the new chips are said to offer up to ten times better generative and multimodal AI performance, and up to three times better graphics performance.
They also represent a major architectural shift. The new chips are built using a chiplet-based design—a method that assembles smaller, pre-fabricated chip modules, each dedicated to specific tasks, into a larger integrated unit.
Weast explained that this modular structure reduces production costs. Mature functions can be assigned to older, more cost-effective process nodes, while high-performance modules use the latest nodes—allowing for greater efficiency and flexibility.
The result is a more customizable product. Intel says the new SoC enables automakers to fine-tune configurations for computing, graphics, and AI, helping reduce both development time and costs. The chips are slated for integration into mass production vehicles by 2026.
Intel is also focusing on domain controllers.
Two decades ago, the company pioneered hardware virtualization in data centers to optimize computing workloads—a framework it refers to as “software-defined infrastructure.” Now, Intel wants to bring that same model to automotive.
Weast noted that reducing the number of electronic control units (ECUs) is key. Even after consolidating infotainment and driving assistance systems, a car might only go from 100 ECUs to 99. Automakers, however, aim to cut that number in half.
Intel proposes using software-defined domain controllers that can dynamically allocate resources across tasks.
For example, air conditioning and seat heating systems are rarely used at the same time. Today’s vehicles may use separate chips for each. But with Intel’s platform, both functions could run on the same chip, using resources more efficiently. This mirrors how data centers manage fluctuating workloads.
“We can virtualize hardware, manage computing power dynamically, and allocate resources to the functions that need them most,” said Cloud Li, sales director for Intel’s China automotive division. “That’s what we mean by a truly software-defined vehicle.”
Li added that achieving this vision still requires deep integration into automakers’ electronic and electrical architectures. With more Chinese carmakers developing their own operating systems and vehicle platforms, Intel’s next challenge is convincing them to relinquish full control.
Why Black Sesame, not Mobileye?
When asked why Intel chose Black Sesame for the fusion platform instead of its own autonomous driving subsidiary, Mobileye, Weast said the decision reflects Intel’s commitment to an open ecosystem. The platform is designed to allow vendors to plug in their own chips, making the choice ultimately up to automakers.
Currently, the fusion system doesn’t yet use a chiplet structure. Intel’s cockpit chip connects to the external autonomy chip via PCI Express (PCIe), which enables real-time sharing of camera feeds—a key upgrade over traditional Ethernet, which doesn’t support real-time driving data.
“Traditionally, cockpit and driving chips rely on Ethernet, which lacks the bandwidth for real-time access to driving data,” Li said. “Now, thanks to PCIe, cockpit systems can directly process camera inputs from the vehicle environment and deliver more intelligent interactions.”
Redefining the vehicle, chip by chip
So, is Intel shifting away from driver assistance systems to focus solely on cockpits?
Not exactly.
According to Weast, Intel no longer builds chips for specific automotive features like self-driving or infotainment. In a software-defined architecture, everything is treated as a workload or application—software, essentially.
The goal is to build a general-purpose, high-performance compute platform. As this model gains adoption in other industries, Intel expects automakers to follow. Over time, advances in general-purpose chips could reduce the need for narrowly specialized components.
Asked whether Intel would lower prices to gain market share, Weast argued that its chiplet architecture already offers better value. Most vendors, he said, rely on a single large chip for both high-end and entry-level models—resulting in unused capacity in cheaper cars, but no cost savings in production.
By contrast, chiplets allow Intel to create more tailored, lower-cost designs—particularly by mixing process nodes. The company says it’s the first in the industry to offer such a modular platform for automotives.
While Intel doesn’t disclose revenue from its automotive business, Weast noted that the unit operates under the Client Computing Group (CCG), which also includes Intel’s PC business. That structure allows the automotive team to leverage Intel’s broader investments in AI and computing infrastructure.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Li Anqi for 36Kr.