Last night, I visited the IM LS8 and Apple showroom for observations.

In the spring of 2026, conversations in the auto showroom are experiencing subtle changes. At the launch event for the SAIC Zhiji LS8, sales staff no longer merely emphasize the horsepower of the engine or the tuning of the chassis; instead, they shine the spotlight on an inconspicuous parameter - “This vehicle is equipped with 64GB of RAM, using original Samsung/SK Hynix chips.” This emphasis on detail is not accidental; it reflects a profound transformation in the automotive industry from being “mechanically defined” to being “computationally defined.”
Meanwhile, in front of the display stand at an Apple retail store, consumers are hesitating over the storage options for the iPhone and MacBook: jumping from 512GB to 1TB instantly raises the price by thousands, while the 2TB version comes with a staggering premium. This “storage anxiety” in the consumer electronics sector is now being directly applied to smart vehicles.
The underlying logic is common: whether it’s a smartphone or a car, they are no longer merely communication tools or means of transportation but mobile intelligent computing terminals. When L4-level autonomous driving requires real-time processing of massive point cloud data from LiDAR, and when edge AI large models need to run complex inference tasks locally, storage chips - once regarded as “supporting characters” - have become the “throat” that determines the upper limit of intelligent experience.
To understand this transformation, it is essential to clarify two core concepts: RAM (DRAM) and storage capacity (NAND Flash). If we liken smart vehicles to a person, then RAM represents “short-term memory,” determining how many tasks the brain can handle simultaneously without lag; while storage capacity represents “long-term memory,” determining how much maps, music, and driving data can be retained.
In the field of RAM, the technological competition has entered a fever pitch. As the global storage titans, Samsung and SK Hynix have already showcased LPDDR6 memory chips based on 1c nanometer technology by 2026. The single-chip capacity of these chips has surpassed 16GB, with a transmission rate reaching an astonishing 14.4Gbps. This means that smart vehicles equipped with such chips have data throughput speeds sufficient to support real-time 3D environment modeling and complex AI decision-making. In contrast, while the current mainstream LPDDR5X can also reach a rate of 10.7Gbps, when faced with the future L4-level autonomous driving’s data deluge of several GB per second, the advantages of LPDDR6 in high bandwidth and low power consumption will become indispensable.
Similarly, the gap in storage capacity is just as evident. The consumer electronics sector has already entered the 2TB era, but in the automotive-grade market, the pace of large-capacity storage proliferation is relatively slow due to the stringent requirements for reliability and longevity. Currently, Samsung and SK Hynix can mass-produce 1TB and even 2TB automotive-grade UFS 4.0 and SSD products, sufficient to handle the massive log data generated by high-level intelligent driving. However, for most vehicle models, 256GB to 512GB remains the mainstream configuration. This capacity difference directly determines whether vehicles can support future OTA upgrades and a richer in-car entertainment ecosystem.
Looking back domestically, China’s storage industry has made remarkable progress in recent years, but in the high-end automotive-grade sector, there still exists an “intergenerational gap” compared to international giants.
In the field of RAM (DRAM), domestic manufacturers such as Changxin Memory (CXMT) and Jiuan Technology are striving to catch up. Currently, domestic automotive-grade LPDDR4X memory has achieved mass production of 8GB, capable of meeting the basic needs of most smart cockpits. However, faced with the 16GB LPDDR6 already laid out by Samsung and SK Hynix, domestic chips still have a technological gap of about two generations in single-chip capacity and transmission rates. This gap is not only reflected in the process technology (international giants have reached 1c nanometers, while domestic products remain at 1x/1y nanometers), but also in yield control and power consumption optimization.
In terms of storage capacity (NAND Flash), domestic performance is even more impressive. Companies like Jiuan Technology and Biwei Storage have gained the strength to compete with international giants in the eMMC and UFS fields. Currently, domestic 512GB UFS 3.1/4.1 products have passed validation from multiple automotive manufacturers and have started small-scale supply. In the automotive-grade SSD field, Biwei Storage has launched a 1TB product that directly competes with Samsung’s high-end offerings. This indicates that in the “data storage” segment, domestic chips have already developed strong substitution capabilities, effectively alleviating the “bottleneck” risks in the supply chain.
However, in the specialized field of code storage (NOR Flash), domestic manufacturers like Gigadevice are even leading the world, with their products having entered the supply chains of international car companies such as Tesla. This proves that Chinese storage enterprises have the ability to achieve a leapfrog advantage in specific lanes.
Despite the accelerating pace of domestic substitution, the global storage market in 2026 faces severe structural challenges. The construction of AI data centers is crazily “siphoning” global storage capacity. In order to produce more profitable HBM (High Bandwidth Memory) and LPDDR5/6, giants like Samsung and SK Hynix are cutting back on the capacity of mature processes like DDR4.
This capacity shift has led to a tight supply of automotive-grade memory. For the automotive industry, while products like DDR4 from mature processes may not perform as well as the latest generation, they are stable, reliable, and cost-controlled, widely used in critical components such as body control and dashboards. Now, as international giants reduce output, these “old products” have become scarce resources, with prices soaring.
At the same time, the demand for memory capacity in L4-level autonomous driving is growing exponentially. Micron Technology predicts that the memory demand for future L4 vehicles will exceed 300GB, which is 20 times that of current mainstream models. This explosive growth in demand, combined with AI’s occupation of capacity, has plunged the automotive storage market into a super cycle of “rising quantity and price.”
From the emphasis on 64GB of memory in the Zhiji LS8 to the exorbitant premium on Apple’s 2TB storage, what we see is not just technological advancement but also a reflection of the intelligent terminal’s thirst for data processing capabilities. In this competition for “brain capacity,” Samsung and SK Hynix, with their absolute advantages in advanced processes, still hold the industry’s pricing power and discourse power.
But for the Chinese industry, this is both a challenge and an opportunity. Although there is still a gap in the high-end DRAM field, in NAND Flash and NOR Flash areas, domestic chips have already opened a breach. In the future, as companies like Changxin Memory continue to make breakthroughs in the DRAM field, and as module manufacturers like Jiuan Technology and Biwei Storage delve deeper into automotive-grade products, China’s automotive industry is expected to gradually reduce its dependence on overseas storage giants. After all, in the second half of intelligent development, only by mastering the core lifeblood of storage can we truly achieve the leap from being a “big automotive country” to a “strong automotive country.”