AI Edge Computing Drives DRAM Shortage as Global Supply Imbalance Hurts Industry

The wave of artificial intelligence is extending from large-scale data centers to terminal edge computing. With the rapid surge in demand for real-time data processing in applications such as smartphones, Internet of Things (IoT) devices, and self-driving cars, AI models no longer rely solely on the massive computing power of the cloud, but must instead complete inference and decision-making directly on the devices themselves. This "edge AI" trend makes high-speed, low-latency memory a critical infrastructure—especially high-capacity Dynamic Random-Access Memory (DRAM) capable of real-time read and write operations, the supply status of which directly impacts the deployment speed and cost of AI applications.

Over the past six months, clear signs of supply and demand imbalances have emerged in the global DRAM market. On the demand side, benefiting from the continuous expansion of parameter scales in AI training and inference models, memory demand for data centers and high-performance computing (HPC) equipment has grown by more than thirty percent compared to the same period last year. Meanwhile, terminal devices have also begun adopting high-bandwidth DRAM in large quantities to support local inference. On the supply side, limited wafer fab capacity expansion, fluctuations in raw material prices, and delivery delays for certain process equipment have caused the rollout speed of new capacity to lag behind demand growth. As a result, DRAM prices continue to rise in international markets, inventory turnover rates have declined, and pressure is being felt across all links of the supply chain.

Addressing this situation, Pao Chien-cheng, founder and CEO of flash memory controller IC maker Phison Electronics, pointed out in a media interview today that the shortage of DRAM is precisely the biggest challenge to the implementation of AI. "Judging from industry feedback, almost every company is struggling with insufficient DRAM supplies," Pao explained. Although Phison primarily focuses on NAND controllers, it faces the same constraints within the overall memory demand chain for AI applications. To ensure product delivery schedules, industry players have begun negotiating with memory foundries to advance production scheduling and are exploring the use of higher-performance memory architectures to lower unit costs. However, it remains difficult to fundamentally resolve the supply and demand gap in the short term.

Among the many trapped enterprises, media observation indicates that Acer appears to be the sole exception. Pao mentioned that, based on public information, Acer's recent financial reports and supply chain disclosures do not show increased costs or delivery delays caused by the DRAM shortage. Analysts speculate that Acer's product mix leans more toward laptops and commercial desktop computers, for which memory demand is relatively stable and can be adjusted through longer inventory buffers. Additionally, Acer may have locked in sufficient DRAM inventories ahead of time through its procurement strategy to mitigate the impact of market volatility. This differentiated supply chain management approach has allowed Acer to appear more composed amid the current tight supply and demand situation.

For Taiwan's semiconductor industry as a whole, the DRAM supply-demand imbalance not only affects memory manufacturers, but also ripples through upstream wafer foundries, packaging and testing houses, and downstream system integrators. If supply tightness persists, the revenues and gross margins of related companies could be squeezed, while simultaneously driving up the overall cost of AI-related hardware and dampening demand growth in terminal markets. Industry consensus holds that in the short term, pressure can only be alleviated by increasing capacity utilization rates, accelerating the commissioning of new plants, and diversifying memory technologies (such as HBM and LPDDR). In the long run, policy-level encouragement for investment and strengthened supply chain resilience are needed to prevent similar supply-demand imbalances from recurring. When evaluating relevant enterprises, investors should pay special attention to their memory procurement and inventory management strategies, as well as their capacity to rapidly adjust production during supply shortages.

In summary, the development of AI toward edge computing has pushed the supply-demand balance of DRAM into unprecedented challenges. From the perspective of Phison's CEO, the industry generally feels the pressure of memory shortages, while Acer's relative stability demonstrates that differentiated supply chain management strategies can become the key for enterprises to resist risks. Moving forward, as AI applications continue to spread, the capacity deployment and supply chain resilience of the memory industry will directly impact the competitiveness and growth momentum of Taiwan's technology sector.