Wall Street debates the storage chip "super bull market": Riding on AI "crossing the cycle", or the last madness before the recession cycle?

In 2026, the global storage chip industry is experiencing an epic revaluation of its value. Samsung Electronics' market value surpassed 1 trillion US dollars, SK Hynix's stock price soared nearly 200% within the year, Micron Technology, Inc. continuously refreshed its historical highs, and SanDisk even set a new semiconductor industry record with a cumulative increase of over 300% in the past 12 months. This AI-driven storage feast has propelled the South Korean KOSPI index to break through the 8000-point mark in less than six months, rising by more than 85% and outperforming major global indices. However, doubts have never disappeared behind the celebrations. From the panic selling triggered by Alphabet Inc. Class C's TurboQuant algorithm to the unprecedented concentration risk in the South Korean stock market, and the grim warnings from industry veterans that "a leopard cannot change its spots," the market is embroiled in a fierce debate over "structural change" versus "cyclical recurrence." Stock prices of storage companies surge in the AI boom Samsung Electronics: A milestone of 1 trillion market value On May 6, 2026, Samsung Electronics' stock price surged by nearly 16% intraday, reaching 27,000 Korean won per share, a historic high. Its market value surpassed 1 trillion US dollars, becoming the second Asian technology company to achieve this milestone after Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR. And this was just the beginning. On May 21, Samsung Electronics' stock price closed at 299,500 Korean won per share, setting another historical high. The market value surpassed 1 trillion US dollars, making it the second Asian technology company after Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR to achieve a market value of over 1 trillion US dollars. On the same day, with the news of the union strike being suspended, Samsung Electronics' stock price broke through the 300,000 Korean won mark for the first time during intraday trading, reaching a high of 300,500 Korean won, another milestone. As of May 22, Samsung Electronics' market value has exceeded 170 trillion Korean won, with a year-to-date stock price increase of 128%. The core driver of Samsung's stock price surge comes from the explosive growth of its semiconductor business. The financial report for the first quarter of 2026 showed that due to the soaring prices of storage chips and the surge in demand for AI computing power, Samsung Electronics' operating profit soared to 57.2 trillion Korean won, with revenue reaching 133.9 trillion Korean won. The operating profit for the first quarter exceeded its full year profit forecast of 43.6 trillion Korean won for 2025. Several international investment banks have raised their target prices: Nomura Securities has given the most optimistic target price of 590,000 Korean won, nearly double the current price; Korea Investment Securities raised its target price to 570,000 Korean won, citing structural supply shortages in storage chips; Shinhan Securities raised it to 550,000 Korean won, NH Investment Securities increased it to 490,000 Korean won, and Future Asset Securities and JPMorgan both raised it to 480,000 Korean won. SK Hynix: Consolidating its dominance in HBM SK Hynix has seen an astonishing surge in its stock price this year. Data shows that its stock price has increased by 187% year-to-date, currently trading at 194.1 Korean won. As a major supplier of HBM chips for NVIDIA Corporation, SK Hynix holds an unshakeable position in the AI storage field. KB Securities has forecasted a year-over-year increase of 194% in DRAM prices and 244% in NAND prices in 2026. Based on this, SK Hynix's operating profit forecasts for 2026 and 2027 have been raised to 277 trillion Korean won and 428 trillion Korean won respectively, surpassing market consensus by 10.2% and 25.1%. Nomura Securities has set a target price of 4 million Korean won for SK Hynix, with over twice the potential upside from the current price. Macquarie raised its target price by 61% to 2.9 million Korean won, citing worsening memory shortages expected to continue beyond 2027, with HBM prices set to soar further. Micron (MU.US) and SanDisk (SNDK.US): Twin stars of the US storage market On the other side of the ocean, the US storage chip giants are also experiencing a wealth legend. Micron Technology, Inc. set a new historical high for the seventh consecutive trading day on May 11, 2026. The current stock price has slightly retreated to 751 US dollars, still up over 160% year-to-date, with a market value approaching 850 billion US dollars. Citigroup has raised its target price from 425 US dollars to 840 US dollars, citing expectations of a 200% year-over-year surge in DRAM prices in 2026 and further increases in HBM prices in 2027. Even more eye-catching is SanDisk. Following a 580% surge in 2025, SanDisk has surged by approximately 523% year-to-date in 2026, with a cumulative increase of over 3800% in the past 12 months. The latest financial report shows that SanDisk's quarterly revenue reached 5.95 billion US dollars, a year-over-year growth of 251%, with a staggering gross profit margin of 78.4%. Several bullish analysts have even set a target price of 3000 US dollars, which implies nearly double the current share price of 1479 US dollars. Citigroup has also significantly raised SanDisk's target price to 2025 US dollars. "Beyond the cycle" or "the cycle is just late"? The core logic of the bulls: Structural change The main argument supporting the bull market in storage stocks is clear: AI has completely rewritten the operating rules of the storage industry. This argument is based on three pillars: First, the structural shortage of HBM. Sandisk CEO Sanjay Mehrotra warned in an interview that the global shortage of storage chips could last beyond 2026, as the growth in demand for AI far exceeds the industry's expansion rate. He emphasized that although the industry is actively building new wafer fabs, the current supply-demand imbalance is a long-term structural issue rather than a traditional cyclical fluctuation, and the release of truly significant new production capacity is not expected until at least 2028. Second, the irreversible transfer of capacity to HBM. In terms of bit calculations, HBM still accounts for only a single-digit percentage of the entire DRAM market share, with a significant room for penetration rate improvement. TrendForce predicts that HBM shipments in 2026 will exceed 30 billion Gb. UBS Group AG has raised the terminal bit demand for HBM in 2026 from 315 billion Gb to 329 billion Gb (up 88% year-over-year) and to 580 billion Gb in 2027 (up 76% year-over-year). Analysts predict that by the end of 2026, up to 70% of the global output of memory chips will be consumed by data centers, and by 2027, global HBM capacity will only be able to meet approximately 60% of market demand. Third, industry restructuring. More and more manufacturers are shifting their focus to AI-specific storage products. Micron has started to reduce its consumer-level storage business and concentrate resources on the more profitable enterprise market. UBS Group AG predicts that by 2027, Samsung is expected to be on par with SK Hynix in HBM bit shipments, each occupying about 40% of the market share. This means that the industry is transitioning from "scale-driven cyclical games" to "technology-driven value competition." Warnings from short sellers and cautious voices However, not everyone is on board with this narrative. More and more market observers are issuing warnings. The comments of BlueBox Asset Management portfolio manager William de Gale are particularly representative. He pointed out that the industry often experiences "huge ups and downs," and "in the long run, this is a rather bad industry. I doubt that everytime people say the storage cycle is over and storage industry is now a long-term value creation industry, it is still the case - just before everything turns really bad." Andrew Lapping, Chief Investment Officer of Ranmore Fund Management, used an African proverb as a warning: "A leopard rarely changes its spots." He noted that investors must be cautious when investing in an industry where the "historical return on capital is average, but the expected future return is very high." These warnings are not unfounded. In the previous cycle of rising storage chip prices, Samsung Semiconductor and SK Hynix had expanded their capacity significantly, only to suffer the consequences in 2023 when demand waned and prices fell below the cost line. Will history repeat itself is the biggest unanswered question in the current market. It is worth noting that hedge fund giant Paul Tudor Jones has built a bearish position on Micron, signaling a notable preparedness for extreme downturn scenarios amidst the bullish euphoria. The market sees this as a signal that a top macro trader is preparing for extreme downturn scenarios. "The Rise of Equitable AI": concerns about excess computing power On March 24, 2026, Alphabet Inc. Class C Research Institute released an article introducing TurboQuant, a vector quantization compression algorithm that compresses the key memory usage of large models during inference to one-sixth of the original while improving performance up to 8 times without sacrificing accuracy. As soon as the news broke, the market value of global storage chip giants evaporated over 90 billion US dollars in a single day (equivalent to 620 billion RMB). SK Hynix's stock price fell by 6.23%, causing a market value loss of around 29.38 billion US dollars; Samsung Electronics fell by 4.71%, leading to a market value loss of around 38.45 billion US dollars; Micron fell by 3.4%, and SanDisk saw a substantial drop of 6.5%. The Calm After the Storm: However, the market quickly calmed down. Several foreign investment banks pointed out that TurboQuant only affects the KV cache during the inference stage, while the essential demand for HBM comes from weight storage and gradient computations during the training stage, which is unaffected; NAND Flash completely uninvolved in long-term data storage. More importantly, the logic of the "Jaevins Paradox": when technological efficiency improvements result in reduced memory demand per unit of computing power, the significant cost reduction in inference will lower the barriers to AI deployment, sparking a larger overall demand for computing power, thus expanding rather than contracting the memory market size. Deutsche Bank Aktiengesellschaft, however, reminds investors in their report to "prepare for continued disruption caused by artificial intelligence." TurboQuant is a classic example, and its release led to a significant drop in the share prices of the largest memory providers. Although whether this technology will bring about a structural change in demand is still "to be observed," it reveals a deep-seated risk: in the era of rapid AI technological advancements, any breakthrough in software efficiency can potentially impact assumptions regarding hardware demand. TurboQuant precisely hit the core cost pain point of large models - the KV cache. During inference, a 700 billion parameter model in a scenario of 512 users and 2048 Tokens requires about 512GB of memory for the KV cache, roughly four times the model itself. TurboQuant compresses the KV cache to 3 bits through two optimizations - PolarQuant polar coordinate quantization and QJL residual correction, resulting in a memory usage reduction of approximately 83%. Efficiency Improvements in Models Furthermore, cutting-edge artificial intelligence is becoming rich and affordable. Chinese labs charge only a fraction of what American labs charge for similar work. Meanwhile, competitors like NVIDIA Corporation(NVDA.US), Cohere, Reflection, and Mistral are developing cheaper, smaller, and more efficient alternatives for enterprises. Artificial Analysis, an AI benchmark testing company, conducted identical evaluations of all mainstream models and tracked the total cost. For the most powerful models from each lab: the cost of Anthropic's Claude is $4,811; OpenAI's ChatGPT is $3,357; DeepSeek is $1,071; Kimi is $948; KNOWLEDGE ATLAS's GLM is $544. The cost of Claude is nearly nine times that of the cheapest Chinese alternative under similar workloads. Moreover, these low-cost alternatives are no longer behind. DeepSeek, a Chinese AI lab that caused tech stock sell-offs in the US, released a preview of its next-generation models last month. The model performs on par with, if not almost identical to, the latest models from OpenAI, Anthropic, and Alphabet Inc. Class C in encoding, intelligence agent, and knowledge benchmarks. Over the past four months, other Chinese labs including Beijing Dark Side of the Moon, Xiaomi, and KNOWLEDGE ATLAS have also released models with similar performance levels. The Risks of a Bull Market - South Korea's Concentration Risk: A Stock Market with a "Double-Headed Monster" The South Korean KOSPI Index surpassed 4300 points on the first trading day of 2026, and in less than six months, it broke through the 8000-point mark. Compared to the beginning of the year, it has risen by over 85%, outperforming major global indices. However, the structure of this bull market is extremely distorted. With over 800 constituents, the KOSPI Index is almost completely controlled by the movements of just two stocks - Samsung Electronics and SK Hynix, which account for nearly 50% of the index weight. This means that the index is almost entirely influenced by the movements of these two stocks, with over 600 other stocks underperforming the index, highlighting a severe imbalance in the market fundamentals. Calls for Profit Taking Steve Brice, Chief Investment Officer of the Global Wealth Management Company Standard Chartered, stated in an interview on May 13 that he believed the optimism in the South Korean stock market "won't last long." He revealed, "I was in South Korea last week, and we advised clients to take profits on some investments and rotate them into diversified global investment portfolios." This risk was illustrated vividly on May 15. KOSPI plummeted by 488 points or 6.1%, closing at 7,493 points, with the two heavyweight stocks Samsung Electronics and SK Hynix falling by 8.6% and 7.7% respectively. Although the market rebounded later on, this "stock disaster" fully exposed the market's vulnerability under the overconcentration structure in South Korea. Political Risk and the GEO Group Inc. Situation should not be overlooked. Some analysts point out that against the backdrop of the tense global GEO Group Inc. situation, the structural imbalance and cyclical issues in the South Korean stock market add significant uncertainty to market trends. Bull Market Leaders Still Singing High Despite this, some top international investment banks remain extremely bullish on the prospects for the South Korean stock market. Nomura Securities raised the KOSPI index's 2026 target from the previous range of 7500 to 8000 points to 10,000 to 11,000 points on May 21, citing a supercycle in general memory and HBM, as well as additional catalysts such as South Korea being on MSCI's watchlist for developed markets. UBS Group AG also raised its earnings growth rate estimates for KOSPI in 2026, 2027, and 2028 to 258%, 46%, and 9% respectively, with the 2026 growth rate higher than the market consensus of 235% and the strongest among Asia-Pacific emerging markets. Risk Landscape: Hidden currents beyond the Bull Market narrative The Cost of Imbalance in the Consumer End The soaring prices of storage chips are triggering chain reactions downstream in the industry. In the first quarter of 2026, the Chinese smartphone market saw a 6% year-on-year decline in sales, mainly due to a shortage of memory combined with a 10% to 30% price increase in new phones, leading to consumers extending their replacement cycle to over 36 months. The cost increase for the mainstream 8GB+128GB storage configuration is as high as 290%. The PC market is also under pressure, with IDC forecasting a 9% drop in PC shipments by 2026 due to high component costs, making it challenging for many consumers and businesses to afford "AI PCs." In niche markets, the price of MLC NAND flash memory has risen by around 300%, with the cost for 16Gb MLC NAND increasing from around 3.2 US dollars in the first quarter of 2025 to around 12.9 US dollars in the second quarter of 2026, potentially impacting industries such as automotive electronics, industrial control, and medical devices. The Sword of Damocles of Expansion Cycles Micron announced in June 2025 that it would invest 200 billion US dollars in capacity expansion over the next few years, with Samsung and SK Hynix also accelerating their investments. While large-scale new capacity release is not expected until at least 2028, once operational, the supply-demand landscape could quickly reverse. Historical experience shows that the conversion from scarcity to surplus in the storage industry accompanying a period of capacity expansion often happens faster than the market expects. As an analyst put it, "If Micron and competitors' new capacity comes online by 2027 to 2028, the length of this price surge cycle will be the core variable behind the entire bull argument." Uncertainty of Technological Disruption The TurboQuant event shows that the rapid iteration of AI technology could potentially impact hardware demand assumptions at any time. Taking a more macro view, large model architectures themselves are also undergoing transformation - trillion-parameter models using dynamic parameter activation mechanisms can reduce inference costs by 73% and cache usage by 90%. While these efficiency improvements may lead to an increased total demand in the long term through the Jevins paradox effect, in the short term, they are undeniably adding to market volatility. JM Finn, a wealth management firm, maintains a cautious and pragmatic stance on this. Jon Cunliffe, Office Manager at JM Finn Investment Office, points out that there is significant room for production growth over the next three years, which would alleviate supply constraints, especially given that AI demand is growing at a more normal pace. He adds, "Today's stock prices assume that stock prices will remain high in the long run, companies maintain a high level of discipline in investments, and profit margins will be much better than in the past. We also emphasize that in recent weeks, this industry has experienced a high momentum squeeze, making it susceptible to market reorganization effects." Looking Ahead: The Feast is not over, but buckle up for safety The fundamentals of the storage market for the second quarter of 2026 remain strong. TrendForce data suggests that general DRAM contract prices are expected to rise by 58% to 63% quarter-over-quarter, while NAND flash contract prices are expected to rise by 70% to 75%. In the first ten days of May, South Korea's semiconductor exports approached historical highs at 8.5 billion US dollars, with a year-over-year increase still reaching 150%. The bullish positions of top investment banks such as Nomura Securities and UBS Group AG, Micron's CEO's belief that the supply shortage will continue until 2028, the incremental demand brought by the AI wave - all these factors support the current narrative of high valuations. However, history teaches us that the strongest consensus in the market is often