Gasgoo Munich- "The battle for platforms has begun, but platform dominance remains undecided," Wang Chao, a professor at Beijing University of Technology, told Gasgoo and other outlets at the 2026 China Automotive Supply Chain Conference.
China now accounts for roughly 36% of the global cockpit market. It holds clear advantages in displays, head-up units, voice interaction, and domain controller integration, topping multiple domestic segments—a formidable foundation.
The colder reality: overseas giants still control over 70% of China's cockpit domain controller chip market, while base-layer software—real-time operating systems and virtualization—remains heavily reliant on foreign suppliers.
Wang sees 2026 and 2027 as the critical window to shore up the base layer. "A few years ago, conditions weren't ripe; a few years from now, the seats will be taken."
Why are these two years so pivotal? Where exactly are the weak links, and how can they be fixed?
Image source: Huaban
Why 2026–2027?
Why these two years in particular? Wang believes three forces are converging at exactly this moment.
Consider architecture first. Standard front-install deliveries of cockpit domain controllers hit 9.85 million units in 2025—a 42.9% fitment rate, according to the Gasgoo Auto Research Institute. That same year, single-chip cockpit-driving integration platforms entered mass production. The domain controller is evolving from a box that merely links screens into a computing node fusing cockpit and driving functions.
Wang stresses a temporal logic: platforms specified today must support the next two or three generations of vehicles. "Once the interfaces are chosen, the room for reuse is locked in." Every architectural decision made today draws the boundary for the platform's future ecosystem.
Driving this is a shift in the logic of market growth.
The global smart cockpit solution market for passenger vehicles reached 366.8 billion yuan in 2024 and is projected to hit 762.8 billion yuan by 2029, a compound annual growth rate of 15.8%, according to CIC. China is outpacing the world, with a CAGR of 18.4% over the same period.
The pie keeps growing, but regional roles are hardening: Asia-Pacific sets the scale, while Europe and the US define the high end. China dictates the mass-production cadence and the cost curve; Japan and South Korea bring deep expertise in instrument clusters and HUDs; Europe and the US control the premium entry points and the critical nodes of regulatory evolution.
The shift is most visible on screens. Global automotive display panel shipments reached 232 million units in 2024, but growth slowed to 5.1% in the first half of 2025. As the expansion in screen counts eases, the center of gravity shifts from hardware to software-defined platforms—a key reason architectural upgrades are accelerating now.
AI is the second force. Mass-production vehicles featuring cockpit-driving integration reached roughly 1.67 million units in 2025, a 43% annual jump, while the front-install AI cockpit rate hit 12%. Dubbed a breakout year for automotive AI agents, 2025 marks the shift from conversational AI to executive AI.
When agents move beyond chatting and answering questions—when they can call tools, orchestrate vehicle controls, and string together services—they fundamentally rewrite the demands on base-layer computing, operating systems, and data architecture. This leap in AI capability is forcing a rebuild of the interaction base, and the window for that rebuild is right now.
The hardest constraint is regulatory. The logic of oversight has shifted from encouraging feature adoption to banning non-compliant sales. Starting July 1, 2026, two mandatory national standards—on vehicle cybersecurity and software updates—will apply to all newly approved models. From January 1, 2027, GB 47955–2026 will impose strict human-machine interaction safety rules for combined driving assistance, mandating hands-off and eyes-off detection, tiered warnings, and ultimately system disablement—hardwired into product definitions.
Between January 2026 and July 2027, a dense wave of regulations will land—covering cybersecurity, software updates, cross-border data, and cockpit intelligence grading. Compliance is no longer the exclusive domain of premium models: since 2026, L2 combined driving assistance penetration has hit 70.5%, and navigate-on-autopilot (NOA) has reached 34.2%. Compliance is now a mass-production prerequisite for mainstream vehicles.
Overseas, the regulatory timeline is just as stark. The EU's Advanced Driver Distraction Warning (ADDW) requirement applies to new vehicles in scope from July 7, 2026, while the Data Act's design obligations kick in for newly placed connected products on September 12. With China exporting 7.1 million vehicles in 2025, domestic cockpit suppliers must clear these hurdles before they ship out.
Maturing architectures, an AI leap, and a regulatory countdown—three forces converging in 2026 and 2027. A few years earlier, domain controller penetration was too low, AI hadn't moved beyond chat, and mandatory standards were not yet in force. A few years later, this round of platform specifications will be set in stone, and the cost of changing interfaces and ecosystems will be prohibitive.
That is the weight of this window. It belongs to no single company; it is the industry's chance to reposition itself within the global value chain.
Strong Endpoints, Shaky Foundations
To grasp what must be fixed in these two years, the line between strengths and weak spots must first be drawn.
The strongest hand lies in endpoints and integration. The China Automotive Parts Industry Development Report (2026), for which Wang serves as executive editor, reveals that Chinese panel makers captured 56.6% of the global automotive display market in 2025—up from 28.1% in 2019, doubling in just six years. In HUDs, leading Chinese firms hold the top domestic spot with a 25.6% share. In voice interaction, domestic leaders command 41.9%.
Integration is equally formidable. Mainland Chinese companies hold seven of the top ten spots in domestic cockpit domain controller assemblies, claiming a combined majority; the leader sits first with 16.1%. This is the "strong endpoints, strong integration" dynamic the report highlights.
Moving up the stack, domestic platforms are gaining speed. A cadre of Chinese in-vehicle operating systems and cockpit solutions is transitioning from single-model deliveries to reusable platforms; one local cockpit solution has already reached a deployment base of 9.4 million units.
The true vulnerability lies at the base layer. In China's cockpit domain controller chip market, Qualcomm still commanded a 71.4% share in the first half of 2026. Real-time operating systems and virtualization layers remain QNX-dominated, with cumulative global deployments exceeding 275 million vehicles (measured across multiple in-vehicle systems, not just cockpit OS share). Automotive-grade EDA/IP and certification toolchains represent the deepest waters for domestic substitution.
Image source: Qualcomm
Above the base layer, the battle for four control points is just as fierce. Beyond sheer computing power, interface control is being contested by mobile ecosystems and international Tier 1s; next-generation phone connectivity can now span all driver displays, including instrument clusters. Update control is locked by chip platforms and mobile ecosystems—some international systems already span 16 brands and over 100 models, and are beginning to push large language models directly to vehicles. Data and security control is the most concentrated, with security foundation and AI middleware firms deeply entrenched after years of work.
Wang distills this into a rule: the more implicit the control point, and the closer it sits to the base layer, the higher the concentration—and the harder it is to displace.
Most troubling, the substitution curve at the base layer is flattening. Gasgoo data shows Qualcomm's share of China's cockpit domain controller chip market slipped from 72.7% in 2025 to 72.1% in the first quarter of 2026 and 71.4% in the first half—barely a 1.3-percentage-point drop over 18 months. The combined share of three domestic players—Huawei, SemiDrive, and AutoChips—edged from 14.8% to 14.7%, essentially flat.
Wang sees two thresholds here. The first—"can it be built, and can it go into production?"—has been crossed. The next is how to keep winning orders in a market defined by crowded entry, falling prices, and faster model cycles. "This is no longer just a technology problem," he notes. "It's a commercial problem, a continuous delivery problem, a customer relationship problem."
The numbers also conceal a seductive illusion: domestic assembly rates do not equal platform capability.
Wang offers an example: a cockpit domain controller delivered by a Chinese Tier 1 might register as 100% domestic by assembly rate, yet its chips, real-time operating system, and functional safety assessments could all be foreign. Measured by platform capability, the true depth of substitution might be less than 30%.
The blue book offers another yardstick, he notes: can the same software stack pass certification and sustain long-term maintenance across at least two chip platforms and two regional markets? "By this metric, fewer than five domestic companies currently measure up."
The base-layer challenge is about ecosystems, not just silicon. A single automotive-grade chip carries with it toolchains, virtualization, driver adaptations, and an entire certification regime. It must clear four gates—performance viability, ecosystem usability, certification credibility, and long-term maintainability—each measured in years.
As Wang succinctly puts it: "Scale can compress a software adaptation cycle from 12 months to six, but it cannot shrink an 18-month automotive chip validation cycle to nine."
Meanwhile, base software fragmentation is the biggest efficiency black hole. In functional safety, Wang stresses that "what is being fused is the computing platform, not the safety liability." Data compliance is shifting from policing outcomes to governing at the source. Cross-border data rules, for instance, dictate that exporting sensitive personal information of more than 10,000 individuals in a year generally triggers a security assessment—and the collection end must limit scope at the source.
And if these gaps aren't closed? Wang is unequivocal: "We'll be stuck with 'installations without platform power.' No matter how large the market share, profits and influence will flow to those who control the base layer."
Shoring Up the Foundation: Where to Start?
The window is tight, and it demands concrete action. For suppliers, automakers, and regulators, Wang prescribes the single most important move each must make right now.
For suppliers, the keyword is "dual-chip platforms." Wang argues that companies need not choose between short-term supply stability and long-term autonomy. Instead, they should build a "dual base": the same software stack running on both an international mainstream chip and a domestic alternative, compressing migration work to below a quarter of the initial development effort—a target to be calibrated per project. This secures supply in the near term while keeping a long-term fallback.
A more realistic entry point for domestic platforms lies not in flagship launches, but in the replacement cycles of mid-range, high-volume models—build scale first, then move upmarket. Functional safety and cybersecurity evidence must be ready before the platform is even specified, not patched on after an order is won.
Automakers, for their part, must "lock down their next-generation cockpit platforms this year." They should consolidate to two hardware sets and one software stack, write compliance into supplier qualification criteria, and avoid scrambling to pass tests right before mass production. Software architecture, user touchpoints, data loops, and service operations must stay in-house; everything else should be opened up boldly.
He captures the division of labor with a vivid analogy: "The upper layer can have personality; the middle layer must speak a common language." Interfaces must be standardized, and test results mutually recognized. Otherwise, applications will require one-by-one adaptation, validations will be rerun endlessly, and the root cause of fragmentation will never be cured.
In this platform rebuild, suppliers need not fear being sidelined by automakers' in-house efforts. Wang argues that what automakers truly want to control are the software stack, the HMI framework, and user touchpoints. Chips set the cost ceiling, base software dictates reuse efficiency, and engineering validation defines the barrier to entry—precisely the areas automakers are least equipped to handle alone, and exactly where parts suppliers find their footing.
For regulators, the lever is "building public infrastructure." Common base software and toolchains, shared validation and certification facilities, interface and data format standards—these are tasks no single company can shoulder, nor would it make economic sense to try. Policy focus should shift from subsidizing the localization of individual chips or projects to supporting the construction of platform capabilities.
Wang also stresses the need to first baseline the state of automotive-grade EDA and IP—"there aren't even public statistics right now." This aligns with the 15th Five-Year Plan for Smart Connected New Energy Vehicle Industry Development recently issued by nine ministries, which lists operating systems and automotive chips as critical weak links to overcome.
Shoring up the foundation also means building global delivery capabilities. "Overseas, you don't compete on price; you compete on systems." Wang notes that winning premium projects in Europe and the US requires a complete evidence chain for software process capability, functional safety, and cybersecurity, plus a commitment to maintain the product for over a decade. Preparing these materials typically takes 1.5 to 2.5 years—"if you start after winning the contract, it's too late." The real threshold is providing overseas clients with third-party verifiable evidence, not a mere letter of commitment.
There is more than one path overseas. Wang's research identifies three: some push products directly into global platforms, securing international automaker approvals along the way; some follow automakers abroad, build local plants, and then chase global contracts; others use joint ventures as conduits to export their system capabilities. The paths diverge, but the prerequisite is singular: having deliverable system evidence in hand before the contract is won.
This systemic capability is already paying off: leading cockpit suppliers report a 27.3% gross margin overseas, compared with 18.4% at home.
The reality of this foundation-building effort cannot be ignored. The auto industry's average sales margin was just 3.2% in the first quarter of 2026. In the first half of the year, contract prices for several storage chips common in cockpits more than doubled—and domain controllers are the most storage-intensive components in the vehicle.
Under such financial strain, Wang recommends a "7-2-1" resource allocation: 70% for mature projects with clear customers and production timelines, 20% for growth hotspots, and no more than 10% for frontier exploration. He also warns that alternative supply sources must be pre-validated. When a batch of low-value components ran short last fourth quarter, it still forced automakers to cut production—a cautionary tale.
As the industry shifts from competing on features to competing on platforms, the answer distills into three keys: the power to define, the rate of reuse, and the evidence chain. Automakers, parts suppliers, chipmakers, software vendors, and standards and testing bodies—each must hold their layer and collaborate to close the gaps.
Smart cockpits offer the most realistic entry point for China's auto industry to break into the mid-to-high end of the global value chain. The window of the next two years will determine whether domestic players remain stuck at "share substitution"—or truly graduate to "platform dominance."
