Gasgoo Munich- 2026 is shaping up to be a landmark year for Chinese automakers developing their own smart-driving chips.
From BYD kicking off mass production of its 4nm "Xuanji A3" to Li Auto's stunning debut of its 5nm "Mach M100"; from NIO's "Shenji" deliveries surging past 250,000 units to XPENG's "Turing" chip winning a contract with Volkswagen:
The core supply chain for China's smart vehicles is entering a new era—moving from outsourcing critical components to defining them in-house.
There is no denying that as domestic smart-driving chips flood into vehicles, they have risen from supporting roles to strategic high ground. At the same time, the reality of memory chip prices surging 180% has made supply chain security an increasingly urgent crisis.
Collective Entry: The Underlying Logic of Automaker Chip Development
The wave of automakers building chips didn't start this year, but the concentrated explosion in 2026 signals that the trend has moved from "testing the waters" to a "harvest phase."
Looking at the timeline, NIO's move was the earliest and deepest. In 2025, NIO launched its in-house Shenji NX9031, defining it as the "world's first automotive-grade 5nm smart-driving chip" with computing power exceeding 1,000 TOPS.
Since then, NIO has spun off its chip business into an independent subsidiary, Shenji Technology. It also formed a joint venture with Axera and OmniVision to launch the M97 chip (over 700 TOPS) for a broader market, actively reaching out to automakers like Leapmotor and Geely for external supply deals.
Image Source: NIO
According to NIO's first-quarter 2026 earnings call, over 250,000 units of the Shenji NX9031 have been delivered. At a forum in Chongqing in June, NIO Vice President Ma Lin stated bluntly that at its peak a few years ago, NIO spent $300 million annually purchasing Nvidia chips, whereas the in-house chip "has already saved the company a lot of money."
XPENG has also carved out a unique path. The Turing AI chip, released in 2024, is touted as the "world's first multi-end universal chip." It is not only equipped across XPENG's entire model lineup but also bridges three terminals: smart vehicles, flying cars, and humanoid robots.
Even more noteworthy, the Turing chip has secured a mass-production contract with Volkswagen. Volkswagen's software unit CARIAD also established a joint venture called CARIZON with Horizon Robotics, investing $200 million to develop the high-computing smart-driving SoC chip, the C7H. With XPENG Group now present in over 60 countries and regions, the "external output" of its Turing chip is opening a second growth curve.
Li Auto, meanwhile, made the leap from "skepticism" to "debut" at its Livis Day event on June 15. CTO Xie Yan defined the Mach M100 as the "world's most powerful AI chip." Built on a 5nm process with single-chip computing power of 1,280 TOPS, it pairs with the in-house Mach VLA intelligent driving system, officially claiming reaction speeds 40% faster than humans.
Holding up the chip at the launch event, Li Xiang said: "Take a picture of me, or else the internet will just be full of photos of me posing with spec sheets."
Behind the joke lies a strategic intent: Li Auto views the chip as the core carrier of its brand's technological image.
BYD's entry carries even greater symbolic weight. As the world's largest NEV seller, BYD released its Xuanji A3 smart-driving chip in May. Using a 4nm process, it offers over 700 TOPS per chip, with a three-chip setup exceeding 2,100 TOPS, supporting L3 and L4 autonomous driving. BYD's sheer volume means that once its in-house chips are mass-produced, the marginal effects on cost amortization and supply chain security will be significant.
So, why are automakers so determined to build their own chips?
The most direct driver is cost. Take NIO: spending $300 million annually on Nvidia chips will only rise as sales grow. While the upfront investment for in-house chips is massive—developing a high-computing chip can cost over 1 billion yuan—per-vehicle costs drop significantly once mass production begins. Crucially, in-house chips allow for architectures customized to specific algorithm needs, avoiding the wasted computing power of generic chips and achieving deep optimization through "software-hardware integration."
Supply chain security is another critical consideration. Geopolitical uncertainty has turned the theoretical risk of "chip supply cuts" into a tangible anxiety. With a single smart vehicle carrying thousands of chips, a shortage of just one can halt production. By developing chips in-house, automakers are building a safety barrier around their core operations.
A deeper logic lies in the shift in competitive paradigms. The race for smart driving has moved from "having it" to "having it good"—and the foundation of "good" is computing power and chips.
One auto executive put it bluntly: "The next battleground is chips." This sentiment is gaining traction among peers: whoever controls the chips controls the initiative to define the product.
However, the move to build chips is not without controversy.
At an industry forum on June 13, a head of automaker chip operations stated publicly: "Developing a chip, especially a high-computing one, involves an investment of over 1 billion yuan. Automakers must anchor their position based on their installed base and intelligent driving penetration rates."
Another executive was even more direct, explicitly stating, "We're not making chips anymore."
These rational voices remind the industry: chip development is not a "mandatory requirement" but an "optional choice," and the answer varies by company.
The Computing Race: Domestic Substitution of Smart-Driving Chips in Progress
Behind the wave of automaker chip development lies an industrial competition landscape being profoundly reshaped.
For a long time, the market for automotive smart-driving chips was monopolized by overseas giants like Nvidia and Mobileye. The Nvidia Orin-X became the "standard" for high-end driving, with high per-chip costs. Moreover, automakers were constrained by suppliers' product timelines regarding technical routes and feature definitions. This passive situation—where "chips are in someone else's hands"—is the fundamental reason Chinese automakers are collectively entering the chip arena.
In terms of performance specs, domestic smart-driving chips already have the confidence to compete head-on with overseas products.
NIO's Shenji NX9031 exceeds 1,000 TOPS; XPENG's Turing chip solution tops out at 3,000 TOPS for the entire vehicle; Li Auto's Mach M100 offers 1,280 TOPS per chip (2,560 TOPS dual-chip); and BYD's Xuanji A3 exceeds 2,100 TOPS with a three-chip setup.
Image Source: Li Auto
Compared to the Nvidia Orin-X's single-chip 254 TOPS, domestic chips have achieved an order-of-magnitude leap in computing parameters.
But computing numbers are just the tip of the iceberg. The real competition lies in "effective computing power": the efficiency a chip can deliver in real-world driving scenarios.
XPENG claims the effective computing power of a single Turing chip equals roughly 10 Orin-X chips. While this claim awaits verification in real-world scenarios, it sends a clear signal: domestic chips are pursuing optimized application efficiency, not merely stacking paper specifications.
An even more critical shift is the expansion of business models. NIO spun off Shenji into an independent subsidiary to seek external supply; XPENG's Turing chip won a designation from VW; and Li Auto's Mach chip is also expected to be supplied to other brands.
This means automaker chip development is moving from "internal use" to "external supply," shifting from cost centers to profit centers.
A Qualcomm executive commented at the 2026 Auto Summit: "No matter how well automakers do with their in-house chips, they only serve a limited number of companies. Every generation of Qualcomm chips aggregates the needs of dozens of automakers."
However, Chinese automakers' logic is exactly the opposite—precisely because the upfront investment is so huge, they need to amortize costs and expand scale through external supply, ultimately creating a "cost reduction via internal use + profit via external supply" flywheel effect.
From the perspective of industrial security, the mass adoption of domestic smart-driving chips also signals that China's self-reliance in automotive semiconductors is accelerating.
Although gaps remain with international top-tier levels in advanced process nodes and underlying architecture design, the first step of "domestic substitution"—the breakthrough from zero to one—has been achieved. The successful mass production of NIO's 5nm and BYD's 4nm automotive-grade chips is itself a landmark result of the synergistic development of China's semiconductor and automotive industries.
At the same time, "cockpit-driving integrated" chips are becoming a new technological direction.
The "cross-domain" nature of AI Agents is forcing a merger of smart cockpits and autonomous driving from two independent systems. A single chip handling both cockpit interaction and driving decisions offers significant advantages in cost and computing efficiency. Brands like Li Auto and Great Wall Motor's Wey have begun exploring this direction, which will become the next growth point for domestic chips.
Yet, the computing power race has its hidden concerns.
Excessively chasing TOPS numbers could lead to "computing overcapacity"—most daily driving scenarios simply don't require thousands of TOPS. Meanwhile, the power consumption, heat dissipation, and cost issues associated with high-computing chips could become a burden on the product.
Finding the optimal balance between computing power and cost is a question automakers must answer.
Rising Memory Prices: An Alarm Bell for Supply Chain Security
Just as the race for smart-driving chips heats up, another crisis in the chip sector is quietly spreading.
According to CCTV Finance, overall prices for automotive-grade memory chips have risen by about 180% in the last three months. In the first quarter of 2026, the average price of DDR5 memory surged 288%, with some spot prices increasing nearly tenfold.
The three major manufacturers—Samsung, SK Hynix, and Micron—have shifted 70% to 80% of their advanced process capacity toward HBM and high-end server DDR5 memory to meet the explosive demand of AI servers.
In this battle for capacity, the automotive industry, holding only about 3% of the global DRAM market, has been pushed to the back of the line for capacity allocation.
The transmission effect of rising prices is already visible. BYD stated bluntly that the core reason for price adjustments was the continuously rising cost of global automotive-grade storage hardware. Exeed made the 28,800-yuan smart-driving package on the ET5 high-spec version a paid option instead of a free gift. XPENG raised the price of its XNGP full-scenario advanced driving package by 20%. BYD increased the price of the LiDAR smart-driving option on several models in its Dynasty and Ocean networks from 9,900 yuan to 12,000 yuan, a hike of over 21%.
Image Source: BYD
Industry estimates suggest the rise in memory chip prices adds about 15,000 to 20,000 yuan to the cost of each new energy vehicle. For automakers whose margins are already under pressure, this cost must either be absorbed—eroding profits—or passed on to consumers. The latter is almost impossible in a market where the price war is at a white-hot stage.
NIO's first-quarter 2026 earnings call offered a clear snapshot. Founder William Li revealed that rising raw material costs had increased average per-vehicle cost pressure by over 10,000 yuan. Vice President Ma Lin admitted candidly: "Spending 300 million on Nvidia chips was a loss; developing our own saves money."
In the context of rising memory chip prices, this statement carries even more weight. When price fluctuations in the external supply chain can directly swallow per-vehicle profits, in-house development is not just a competitive necessity—it is a survival necessity.
"In the first half of 2026, automotive memory chips experienced structural tension. This is not a traditional cyclical fluctuation, but a cross-domain capacity battle caused by the explosion of the AI computing industry and the rapid rise in smart vehicle penetration, leading to a deep mismatch in the supply-demand system," said Zheng Yali, Deputy Secretary-General of the China Society of Automotive Engineers and Executive Director of the National Innovation Center of Intelligent Connected Vehicles. Unlike the acute shortage of general chips in 2021, this round of "chip shortage" is a long-term "chronic disease" for high-spec, specialized memory chips. The root cause is that the traditional passive response and multi-layer progressive supply chain management model can no longer adapt to the rapid iteration rhythm of smart vehicle demands.
She pointed out that while this situation brings short-term challenges, it also creates a rare strategic window for domestic memory chips. In the long run, the automotive industry's response strategy should shift from short-term "emergency relief" supply guarantees to a systematic reconstruction of long-term capabilities. This emphasizes deep collaboration with chip companies in demand definition, collaborative design, and supply chain mechanisms, rather than simply pursuing in-house chips.
It is worth noting that the chip challenges facing automakers are not limited to smart-driving and memory chips. Automotive-grade MCUs, power semiconductors, and sensor chips are equally the "lifelines" of smart vehicles.
From BYD developing its own IGBTs to automakers laying out silicon carbide, from NIO making smart-driving chips to XPENG exporting technology to Volkswagen, Chinese automakers are systematically reconstructing their chip supply chain landscape.
Predictions from TrendForce indicate that the trend of rising memory chip prices will be hard to ease in the short term, and automakers will have to endure it for at least another year. This means 2026 could be the year when automaker profits face the greatest pressure, but also the turning point that pushes more companies to resolve to develop their own chips.
Returning to the essence of automaker chip development: this is not merely a technological upgrade, but a redistribution of industrial power.
From "car makers" to "chip makers," Chinese automakers are breaking the boundaries of the traditional auto industry and extending their reach into the semiconductor sector. The independent operation of NIO's Shenji, the external supply of XPENG's Turing, and the mass production of BYD's Xuanji—all these cases point to a single trend: automaker chip development is moving from strategic defense to strategic offense, and from being cost-driven to value-creating.
As one industry insider put it, the competition for smart vehicles has electrification in the first half and intelligence in the second half—and the outcome of the second half is determined precisely by that fingernail-sized chip. Automakers have clearly understood the question—and they have already started answering it.
