Automotive-grade chip prices are rising, and new energy vehicles are starting to face the pain brought by AI
“It was 9,900 last month, and now it's 12,000. The raw material prices have gone up,” explained a salesperson at a BYD dealership to a customer looking at cars, regarding why the price of the Celestial Eye optional package had increased by 20% in just one month. Meanwhile, the entire Xiaomi SU7 series saw a price hike of 4,000 yuan, the Wenjie M9 went up by 10,000 yuan, and the Changan Qiyuan Q07 Tianxuan Intelligent Laser Edition increased by 3,000 yuan.
While traditional fuel - powered cars are collectively announcing price cuts, new energy vehicles are collectively entering a price - increase mode. Li Bin of NIO revealed that due to the price increases of various raw materials and components such as nickel, cobalt, lithium carbonate, and chips, the production cost per vehicle has increased by over 10,000 yuan year - on - year. A highly intelligent new energy vehicle is equipped with over 1,000 chips, almost twice as many as a traditional fuel - powered car. The more intelligent features a vehicle has, the more it is affected. High - end intelligent driving models have become the hardest - hit area, and some car manufacturers have started to slow down the pace of technological upgrades.
From the end of 2020 to 2022, the price of MCUs often increased by 8 to 10 times. An automotive ESP chip originally priced at 13 yuan was once speculated up to 6,000 yuan. However, this time the chip shortage is different. The important categories of chips in short supply in the automotive industry this time are power devices and memory chips. The ones snatching these two types of raw materials are not competitors in the same industry, but the booming AI data centers.
A Wave of Price - Increase Notices Is Coming
On May 26th, Infineon notified its customers that the prices of some products would be raised starting from July 1st. This is its second price increase this year following the one in April. Two days later, STMicroelectronics followed suit and announced price adjustments starting from June 28th, also for the second time this year. On June 2nd, just a few days after sending a price - adjustment notice to automotive customers, STMicroelectronics announced another increase in its revenue target for the data center business in 2026, doubling it directly from 500 million US dollars to 1 billion US dollars. This is its second upward adjustment of the AI - related business forecast within three months. The official statement is very straightforward: Due to the continuous strong demand for AI infrastructure and the progress of production capacity ramping up, the company has decided to significantly increase the revenue target for the data center. If the growth trend continues, the relevant revenue is expected to double again in 2027. Texas Instruments is also proving the money - making ability of data centers with its financial reports. Texas Instruments' data center business has had eight consecutive quarters of sequential growth, and its revenue in the first quarter of this year increased by 90% year - on - year. The company expects that the AI data center will soon account for 20% of its revenue. While raising prices for the automotive industry, it is increasing investment in AI. Reading these two announcements a few days apart, the meaning is clear: Production capacity follows the money, and the automotive industry is not where the most money is.
The situation in the memory industry is even more exaggerated. With the explosion of large AI models, the demand for AI servers has soared. NVIDIA GPUs are in short supply, which has led to a simultaneous surge in orders for HBM high - bandwidth memory and LPDDR5 server memory. Facing higher - priced and more stable AI orders, Samsung, SK Hynix, and Micron have rationally made the same choice: to allocate their best production capacity to AI. In the past year, the prices of automotive - grade DRAM and NAND flash memory have increased by over 100%, and the prices of some specifications have doubled. In the past three months, the prices of automotive - grade memory have soared by about 180%. This widespread price increase of automotive - grade memory is taking the blame for AI. Industry insiders said that in the past, price increases in the upstream of the automotive industry chain were mostly due to the supply - demand cycle fluctuations within the automotive industry itself. This round of price increase of automotive - grade memory has a distinct cross - industry nature: The implementation of large models has led to an exponential growth in the demand for high - performance memory from AI servers and computing power centers. A large amount of production capacity has flowed to the computing power track, directly squeezing the supply of automotive chips. An automotive - grade memory manufacturer described to "Semiconductor Industry Vertical and Horizontal" the transmission path of this structural shortage: The giants are flocking to AI, the mature product lines are losing resources, and the shortage is spreading to the terminals of various industries.
To make matters worse, the storage demand of cars themselves has also exploded at the same time. The application of large AI models in cars is rewriting the electronic and electrical architecture of cars: from distributed ECUs to domain controllers, and then to central computing platforms, the integration of cockpit and driving and the sharing of computing power have become trends. Storage has shifted from independent small - capacity eMMCs of a few gigabytes in each domain to a high - performance and large - capacity storage pool. UFS or BGA SSDs of 128GB or more have become the standard. The bandwidth has jumped from 400MB/s of eMMC to 5800MB/s of UFS 4.0, and even 8000MB/s of PCIe SSD. The status of storage in the chips of a whole vehicle has changed from "inconspicuous" to an important component second only to the computing power SoC. The production capacity of automotive - grade DRAM is being squeezed, and the price is rising without any room for negotiation.
The fact that the supply is taken away by AI, the price is pushed up by AI, and even the incremental resources are snatched by AI is the characteristic that distinguishes this round of price increase from the previous traditional chip - shortage cycles.
Differential Shortage of Automotive - Grade Chips
The chip shortage does not cover all automotive chips equally.
Two domestic intelligent automotive chip companies told "Semiconductor Industry Vertical and Horizontal" that their production capacity has not been affected. A cockpit chip company said, "The requirements of each car manufacturer are personalized, and our products can be adjusted according to customer needs." Intelligent driving and cockpit SoCs are logic chips produced on the logic process lines of foundries. The storage and power chips taken away by AI are mostly from IDM manufacturers. Facing the strong demand and extremely low price sensitivity of AI data centers, the priority of automotive customers has been lowered.
Therefore, the accurate name for this round of crisis is not "automotive chip shortage" but "automotive storage and power device shortage". This shortage reflects the imbalance in the bargaining power of tier1/automobile manufacturers in some parts of the supply chain.
In the field of general - purpose chips, the weakness of tier1/automobile manufacturers mainly comes from two aspects.
The first is purchasing power. When data centers become the core customers, tier1/automobile manufacturers do not rank among the top in the customer list. When the price - increase wave comes, the original manufacturers give priority to protecting large customers, and the automotive industry is often the last to receive the notice.
The second is rhythm. It takes two to three years from the planning to the production of expanded capacity, which requires stable and predictable long - term demand. However, the demand for cars follows the vehicle model cycle, which fluctuates. In addition to the rhythm of capacity expansion, automotive - grade chips also need to undergo safety certification. The requirements for temperature, lifespan, and reliability of products are much higher than those of consumer - grade products, but the profit margin of automotive - grade products may not be higher than that of consumer electronics. There is no motivation for upstream chip manufacturers to specifically reserve production capacity for customers with unstable demand.
AI has not downgraded the automotive industry. It just makes tier1 and automobile manufacturers realize that the right to speak does not depend on who you are, but on your ranking in the other party's customer list.
Self - Rescue and Self - Research of Car Manufacturers
To ensure stable delivery and prices, supply chain management has become the top priority for car manufacturers.
On the one hand, car manufacturers lock in the price and delivery time by signing 3 - 5 - year long - term agreements with suppliers and establishing strategic inventories. At the same time, they have started to develop their own chips or jointly develop chips. BYD announced the mass production of 4nm intelligent driving chips, with the computing power of 3 chips exceeding 2100 TOPS. GAC has jointly developed 51 industry - leading chip products with several chip companies, filling many industry gaps.
The chip shortage has split car manufacturers into two. The door of self - research is only open to half of them. Taking mobile phone chips as an example, the proportion of self - developed SoCs in mobile phone shipments is about 30%, and the remaining 70% belongs to third - party suppliers. Qualcomm and MediaTek, the two oligarchs, account for 60%. After years of self - research on chips, Qualcomm and MediaTek still account for 70% - 80% of the chips in Samsung's own mobile phones. Self - developed chips are not always successful. A research report by Bernstein pointed out that when the annual production volume is less than about 1.5 million units, self - developed chips are not economically viable. Moreover, as the technology stack evolves rapidly, the difficulties of self - research are underestimated. Self - research is a privilege of leading car manufacturers, not a way out for the entire industry.
On the other hand, domestic chip manufacturers have seen this window period and are accelerating to fill the gap. At present, there are still gaps between domestic NAND and DRAM in automotive applications in terms of yield rate, wide - temperature consistency, and lifespan reliability compared with overseas original manufacturers. Car manufacturers still have a path - dependence on overseas original manufacturers. The domestic production capacity is insufficient, the penetration rate is low, the automotive - grade certification cycle is long, and customer trust is still being slowly built. However, in the current situation of shortage, domestic solutions have naturally come to the forefront of car manufacturers. BAW Storage said that for domestic chips, this shortage is a historical opportunity, and domestic car manufacturers are willing to pay a reasonable premium for "stable supply + high reliability + customization". This means that domestic chips no longer need to compete on price but on certainty.
Facing the self - research trend of car manufacturers, there has also been a structural change in the domestic automotive chip industry. For example, after BYD released its self - developed chips, its supplier DiPing turned to the business of "selling IP": authorizing the BPU architecture IP to car manufacturers, charging a one - time authorization fee, and then collecting royalties based on the shipment volume after mass production.
This round of price increase will accelerate the differentiation of the automotive industry. Car manufacturers with strong supply chain management capabilities will stabilize the fluctuations through long - term agreements, inventories, and self - research capabilities. Small and medium - sized players will face the double squeeze of delayed delivery and loss of market share. This round of chip shortage is a stress test of supply chain capabilities, and the gap between large and small car manufacturers is widening at an accelerating pace.
Conclusion
Once upon a time, the anxiety of car manufacturers was "unable to buy chips". In 2026, the anxiety has become "able to buy, but the supply may be snatched away at any time". The former anxiety led to hoarding and speculation, while the latter is giving rise to a deeper - level industrial evolution.
AI has taken away the production capacity and pushed up the price, causing car manufacturers to raise prices and putting pressure on the supply chain. However, in the long run, the occupation of production capacity by AI is structural and long - term and will not subside. Both the upstream and downstream of the automotive industry chain must face such pressure, so they must actively solve the problem.
The self - rescue and self - research of car manufacturers are not about "eliminating external procurement" or simply "replacing overseas products with domestic ones". Instead, they aim to build a stable and diversified supply system to make the industry more mature. From a longer - term perspective, the pain that AI is causing the automotive industry now will turn into sweetness in the future.
This article is from the WeChat official account "Semiconductor Industry Vertical and Horizontal" (ID: ICViews), author: Liu Qian, published by 36Kr with authorization.