TECH NEWS – According to Phison, the demand for NAND chips is at an all-time high, and we will feel the effects in our wallets.

The AI industry has disrupted every area of the supply chain. After affecting the chip market, the industry has set its sights on the DRAM and NAND supply chains. In Phison’s third-quarter earnings report, CEO Pua Khein-Seng revealed that the current memory cycle is unprecedented. In addition to DRAM, he said that the average selling price of NAND TLC memory has risen sharply (by 50-75% in a few months), suggesting that consumer SSDs and related NAND products will become much more expensive. The NAND industry is experiencing huge demand because storage drives using this technology are being deployed on a large scale in the era of inference. In data center environments, NAND-based drives store models and preload LLMs onto local SSDs for low-latency startup. When the model is updated, these changes are reflected in the storage devices (e.g., SSDs). Phison reports tight market supply, and NAND TLC prices are expected to remain stable.

Players in the NAND industry have experienced a decline in demand for some time; however, AI has led to a significant increase in demand for storage devices in AI environments. Pua says that NAND facilities have been underutilized for several months, meaning suppliers have been cautious about expanding capacity. However, with AI demand at an all-time high, supply is expected to increase. Pua believes that NAND manufacturers have been reluctant to expand production capacity because total profits have remained steady since the beginning of the pandemic five years ago. They have not been making substantial profits, but they will be profitable from now on.

Due to the initial increase in DRAM demand, consumers can expect RAM prices to continue rising. However, the average selling price of modules has already increased dramatically, often doubling. Similarly, the price of consumer storage devices, such as SSDs, is expected to rise dramatically in the NAND industry, especially for models with larger storage capacities.

Source: WCCFTech, Yahoo