AI Boom Drives NOR Flash Demand, Creating Shortages and Doubling Prices

According to the latest memory industry report by research firm TrendForce, NOR flash memory is facing the most drastic supply and demand structural changes since the 2000s. Over the past decade, NOR flash was primarily used in embedded systems such as mobile phones, automotive electronics, and industrial controls, with relatively stable demand and a focus on lower-capacity products. However, with the rapid popularization of artificial intelligence (AI) technology—particularly the deployment of large language models and deep learning inference on edge devices—demand for high-capacity, low-power NOR flash featuring fast random access has surged. During the training and inference processes, AI algorithms require large volumes of model parameters to be loaded in real time. Traditional DRAM and NAND flash have found it increasingly difficult to simultaneously meet the requirements of low latency and high reliability, prompting wafer foundries to shift more production capacity toward high-capacity NOR flash that meets AI demands.

In terms of capacity allocation, major global NOR flash manufacturers, including Micron, STMicroelectronics, and Winbond, have expanded their investments in advanced manufacturing processes over the past two years. Nevertheless, starting in 2024, the sudden surge in AI-related applications has pushed capacity utilization rates close to full capacity, with significant bottlenecks appearing in production line scheduling, particularly in the first quarter of 2025. TrendForce predicts that without further capacity expansion, the overall supply and demand imbalance will persist through the second half of 2026. This imbalance is reflected not only in shipment shortages but also in continuously rising prices. Contract prices have remained at multi-year highs, and high-capacity products (such as those 256Mbit and above) are even likely to see their prices double.

The direct impact of the price increases is cost pressure on downstream industries. The automotive electronics and industrial automation sectors are already facing tight supplies of raw materials and semiconductors, and rising NOR flash prices will force complete equipment manufacturers to re-evaluate their product designs and procurement strategies. Some automakers have begun exploring alternative solutions, such as migrating certain functions to microcontrollers' built-in ROM or utilizing emerging magnetoresistive random-access memory (MRAM) to diversify risks. On the other hand, upstream suppliers in the AI hardware supply chain may achieve higher gross margins due to high price levels, further increasing R&D investments, improving process yields, and enhancing capacity density, thereby creating a positive feedback loop between supply-demand dynamics and technology.

From a macroeconomic perspective, the supply and demand imbalance of NOR flash may also cause fluctuations in the export structure of Taiwan's semiconductor industry. Although Taiwan's share of the global memory market is primarily driven by DRAM and NAND, the revenue proportion of NOR flash is expected to rise as AI demand spreads. If the capacity gap persists, Taiwanese manufacturers may gain stronger bargaining power in international markets, while also facing order uncertainties caused by price volatility. At the governmental level, authorities may strengthen subsidies for advanced process equipment and encourage local enterprises to enhance capacity flexibility to avoid supply chain risks arising from over-reliance on a small number of suppliers.

In summary, the rapid penetration of AI technology is redefining the market role of NOR flash, transforming it from a traditional low-capacity embedded memory into a critical component supporting high-capacity model computing. The supply-demand imbalance and price increases will remain difficult to ease in the short term. However, in the long run, if the industry can simultaneously upgrade manufacturing technology and scale up production capacity, it will help stabilize supplies, reduce costs, and promote the implementation of AI applications across more terminal devices. Future development still depends on the investment efforts of wafer foundries, the continued growth of downstream demand, and the degree of policy support for the resilience of the semiconductor supply chain.