After DRAM and NAND flash shortages, a supply crisis is brewing in another memory segment, and AI is the key driver behind the NOR flash supply crunch as well. Demand for NOR flash in AI servers is rising sharply; the number of NOR devices per server rack has increased to more than 30 units, up from roughly three to five previously.
Taiwan’s Commercial Times reports that in Nvidia’s GB200 NVL72 system, NOR content per rack already exceeds $600 and could reach $900 within two years. As a result, this surge is intensifying competition for NOR production capacity between embedded applications and AI servers.
NOR flash offers fast random-access read speeds in demanding execute-in-place applications as well as high reliability and functionality at extreme temperatures. That makes it a staple in applications where code storage is crucial, including automotive, cloud computing, and industrial systems. NOR’s high write endurance also makes it suitable for applications such as over-the-air (OTA) programs, which undergo multiple updates.
www.eic.net.cn
According to Taiwan’s Economic Daily News, Macronix, the world’s largest NOR flash supplier, could announce a 30% price hike for its NOR flash chips in the first quarter of 2026. Though it hasn’t happened yet, the Hsinchu, Taiwan-based memory supplier has reportedly scaled down some of its NOR capacity to increase its MLC NAND flash production.
易IC库存管理软件
Partner Content
Reshaping the Eye of AI: How Will HVS Transform AI Machine Vision in 2026?
By Illumi Huang, Senior Industry Analyst, EETimes 02.09.2026
Galaxy Wins “FPGA Professional Spot Service Provider” Award
By Galaxy Semicon Group Limited 02.09.2026
How to Select High-Performance Capacitors for GaN and SiC Power Systems
By Shanghai Yongming Electronic Co.,Ltd 02.05.2026
To understand the imminent supply crisis in the NOR flash industry, it’s important to know how NOR flash devices are consumed in AI systems. And how AI applications are competing for NOR capacity with other segments such as automotive, consumer, industrial, and IoT.
NOR flash in AI
Where does NOR flash stand in the AI scheme of things? AI data center components, including network interface cards (NICs), controllers, and accelerator boards, benefit from NOR’s stability features, such as error correction code (ECC), cyclic redundancy check (CRC), continuous read support, and wrap-based data patterns that align with modern caching architectures.
NOR flash is increasingly used in AI servers and data centers as a reliable, low-latency storage solution for firmware, boot code, and system initialization. It integrates well with graphics-driven solutions, which makes it highly suitable for code storage and other AI-driven workloads.
NOR flash is critical for the safe boot and initialization processes of AI servers and high-performance computing (HPC) systems. Next, NOR flash provides fast, reliable access to firmware and underlying instructions. Then, there is the high-bandwidth memory (HBM) factor: NOR flash provides the independent power management and initialization required for each DRAM layer in HBM devices paired with AI processors.
In edge AI, NOR flash is crucial for storing operating systems and code in AI-enabled IoT, automotive, and industrial devices, where data integrity is vital. In fact, many edge SoCs are incorporating NOR for model storage because it offers deterministic read latencies. And many edge AI modules stream neural network weights directly from NOR flash for inference workloads.
The rise of AI has also exposed the limitations of 2D NOR flash, as AI applications demand ever higher densities while still requiring NOR’s high reliability. Traditional 2D NOR flash, while effective in many AI applications, is approaching physical and performance limits because its planar architecture constrains scalability toward higher-density, lower-latency memory solutions.
Enter 3D NOR flash.
3D NOR flash stacks memory cells vertically to overcome the scalability issues inherent in 2D NOR’s planar architecture. Vertical stacking, which enables higher storage densities within the same footprint, reduces the need for multiple storage devices such as eMMC and NAND. 3D NOR flash also offers shorter latency, improving boot performance, which is critical for applications that need near-instantaneous access to stored data.
3D NOR flash offers an eight-fold density increase, achieving 512 Mb on a single die compared to 2D NOR’s maximum density capability of 512 Mb. It also employs the 200 MHz double transfer rate (DTR) to facilitate high-speed data transfer for AI applications that require rapid access to large amounts of data.
But while 3D NOR flash promises a new era of high-capacity embedded storage, it’s still years away from commercial availability on a wide scale. Especially when the large memory makers aren’t investing in NOR flash technology anymore. Furthermore, leading supplier Macronix plans to start sampling 3D NOR flash devices in the second half of 2026 and begin full-scale production in 2027.
Till then, the industry will have to deal with 2D NOR flash and its supply crunch. It’s a smaller memory segment compared to DRAM and NAND, but its surging demand in AI designs and a larger reshuffle in manufacturing capacity due to increasing demand for HBM make NOR flash also a memory technology to watch.
See also:
The Shift from 2D to 3D NOR Flash
As systems become more complex, the need for memory components that can keep up with these demands is critical. Enter 3D NOR flash memory.
DRAM Cannot Keep Up With AI Demand
Undersupply will keep DRAM prices high into 2026 due to AI, while emerging memories seek out niche opportunities.
RELATED TOPICS: 3D ARCHITECTURES, AI, DATA CENTERS, DRAM, EDGE AI, HIGH BANDWIDTH MEMORY, MEMORY, NOR FLASH, SERVERS, STORAGE
Share this:
Majeed Ahmad, Editor in Chief of EDN and Planet Analog, has covered electronics design industry for longer than two decades. During this period, he has worked in various editorial positions, including assignments for EE Times Asia and Electronic Products. He holds a Masters’ degree in telecommunication engineering from Eindhoven University of Technology. Follow Majeed on LinkedIn
