Storage chip stocks continue to surge: AI algorithm construction shifts from "lacking GPU" to "lacking memory"

Recently, the storage chip sector of US stocks has continued to be the focus of the technology stock market. Micron Technology and SanDisk have continued their historic upward momentum this week, with the core logic behind it not being complicated: investment in artificial intelligence infrastructure is continuously expanding, driving the demand for storage chips to increase rapidly, while industry supply remains tight. With the triple factors of strong demand, tight supply, and rising prices, the market is beginning to reprice the profit elasticity of storage chip companies. Looking at the stock price performance, the rise of the leading storage chip companies has been quite sharp. Micron Technology rose more than 15% in a single day on Friday, with a cumulative increase of 36% for the week, marking its best weekly performance since 2008; SanDisk rose over 16% on Friday, with a cumulative increase of about 30% for the week. Data also shows that since SanDisk was spun off from Western Digital and listed in February last year, its stock price has seen a huge cumulative increase, indicating a significant increase in market expectations for its benefit in flash memory and AI data center storage demand. The underlying driver of this round of market trends is still AI. In the past two years, the main focus of AI hardware investment has been more on GPUs, AI accelerator cards, and advanced packaging. However, as the scale of training and inference clusters expands, the market is gradually realizing that AI systems not only need more computing power, but also higher bandwidth, larger capacity, and more stable data access capabilities. In other words, the bottleneck of AI data centers is shifting from simple "computing chips" to the entire chain of memory, storage, network, and power. The importance of high-bandwidth memory HBM, DRAM, and NAND flash memory products has therefore significantly increased. The supply side is also an important factor driving the market up. Since the end of February, spot prices of storage chips have been continuously rising, with spot prices of DRAM increasing by about 25% and NAND flash memory by about 18%. The price increases have directly improved market expectations for the revenue and profit margins of storage manufacturers. More importantly, high-end storage products such as HBM needed for AI servers will consume a large amount of advanced process capacity, with wafer area and production complexity higher than that of ordinary memory products, thereby squeezing the capacity of traditional mobile phones, PC-end DDR5, and other products. After the supply side tightens, the transmission of price increases becomes more apparent. The industry has previously experienced a deep downturn cycle. Around 2022, the storage industry suffered losses due to high inventories and weak demand, with major manufacturers such as Samsung, SK Hynix, and Micron taking measures to reduce capital expenditures and control production. After the price recovery, leading manufacturers did not quickly release capacity on a large scale, with the market generally believing that they were more inclined to maintain supply-demand balance rather than repeat the mistake of excessive production leading to price declines. This supply discipline has strengthened investors' judgment on the sustainability of this current prosperity cycle. Signals from the demand side are also getting stronger. Recent financial reports from multiple tech giants have shown increased investment in AI, with chips and related components in short supply. Reports also cited Reuters as saying that several tech companies have expressed willingness to invest in the production line of South Korean storage chip manufacturer SK Hynix and provide funds to purchase manufacturing equipment. The significance of such news is that customers are not just passively accepting price increases, but actively participating in capacity building, reflecting the urgency of AI infrastructure customers for future storage supply. In terms of industry structure, the concentration of the storage chip industry is relatively high. Data shows that Micron, Samsung, and SK Hynix together produce over 90% of the world's DRAM. High concentration means that when demand expands rapidly and supply cautiously recovers, the elasticity of prices and profit margins will be more concentrated on a few leading companies. Micron, as a major US storage chip manufacturer, naturally becomes an important target for US stock investors betting on the AI storage cycle; SanDisk, on the other hand, is receiving attention from funds due to its focus on flash memory and storage business, particularly in NAND and enterprise storage demand. Although the stock price gains have been significant, analysis from Barron's magazine pointed out that investors are not simply speculating on valuations. Based on expected earnings for the next 12 months, SanDisk is valued at around 9.5 times and Micron at around 8.6 times, still significantly lower than the average price-earnings ratio of approximately 25.9 times for the Philadelphia Semiconductor Index. This means that the logic behind market purchases of storage stocks comes from both the long-term demand imagined by AI and the fact that valuations do not seem expensive even after profit upgrades. Overall, the continued surge in storage chip stocks is essentially the result of the spread of AI capital expenditure. In the first phase, the market pursued GPUs and AI servers; in the second phase, funds began looking for complementary links benefiting from AI data center construction, with storage chips being one of the most certain directions. As long as high-intensity investment in AI data center construction continues, there will still be support for storage demand, product prices, and leading companies' profit margins. However, the storage industry itself still has cyclical characteristics, and short-term stock price increases may also face fluctuations. What will ultimately determine whether the trend can continue is the ability of AI demand to continue to materialize, and whether manufacturers can maintain a balance between capacity expansion and prices.