--- title: "The summer pullback in AI semiconductors presents a good opportunity for low-positioning! $1.5 trillion in cloud capital expenditures provides a solid foundation, and the narrative of the \"storage supercycle\" is impeccable" type: "News" locale: "en" url: "https://longbridge.com/en/news/291873776.md" description: "Bank of America research report points out that global capital expenditure on cloud and AI infrastructure is expected to reach $1.5 trillion by 2027. The current summer pullback in AI semiconductors and storage chips is seen as a healthy reset rather than a structural change in demand. Institutions like Nomura rebut the \"semiconductor peak theory,\" believing that the AI computing power race is driving storage chips to become a scarce strategic asset, with a strong narrative of a \"storage supercycle,\" and core beneficiaries like Samsung and SK Hynix are viewed positively in the long term" datetime: "2026-07-07T03:56:04.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/291873776.md) - [en](https://longbridge.com/en/news/291873776.md) - [zh-HK](https://longbridge.com/zh-HK/news/291873776.md) --- # The summer pullback in AI semiconductors presents a good opportunity for low-positioning! $1.5 trillion in cloud capital expenditures provides a solid foundation, and the narrative of the "storage supercycle" is impeccable According to Zhitong Finance APP, even though the stock prices of leading storage chip manufacturers and the AI semiconductor sector have recently entered a downward adjustment trajectory, **Wall Street financial giants remain optimistic about the "storage supercycle" and the long-term bull market trajectory of AI semiconductor-related stocks under the unprecedented AI infrastructure frenzy. Renowned Wall Street investment firm Nomura has released a research report rebutting the "semiconductor peak theory," while Bank of America (BofA) released a report this week indicating that by 2027, global capital expenditure on cloud computing and AI-related infrastructure will reach $1.5 trillion**, pointing out that the current summer correction of AI semiconductor stocks, including storage chip stocks, is a healthy reset trajectory rather than any structural change in AI computing power demand. GPUs are responsible for generating intelligence, HBM/DRAM is responsible for high-speed data feeding, enterprise-level NAND/eSSD is responsible for hot data and caching, while HDD is responsible for the long-term retention of massive cold/warm data. Therefore, Wall Street financial giants like Goldman Sachs believe that the AI computing arms race led by cloud computing giants is transforming storage chips from cyclical products into scarce strategic assets. The price increase of DRAM/NAND in 2026 is not the end, but possibly the initial stage of a supercycle. Whether it is Google's incredibly large TPU AI computing cluster or the massive NVIDIA AI GPU computing cluster, both rely on the fully integrated HBM storage systems equipped with AI chips. Additionally, current tech giants are accelerating the construction or expansion of AI data centers, which necessitates large-scale purchases of server-level DDR5 storage and enterprise-level high-performance SSDs/HDDs. **Samsung Electronics, SK Hynix, and Micron Technology are simultaneously positioned in these three core storage areas: HBM, server high-performance DRAM (including DDR5/LPDDR5X), and high-end data center-level SSDs, making them the most direct beneficiaries in the "AI memory + storage stack," effectively reaping the "super dividends" of the AI infrastructure wave.** The South Korean storage chip giant Samsung Electronics has just disclosed its unparalleled preliminary Q2 performance, which is almost the most intuitive profit sample of this storage chip supercycle. **From April to June this year, operating profit soared approximately 19 times year-on-year, expected to reach 89.4 trillion won (approximately $58.4 billion), once again setting a quarterly historical record, with a strong sequential growth of 56% compared to the previous quarter.** Analysts' average forecast was about 84.2 trillion won. Revenue expectations during the same period reached 171 trillion won, exceeding the market estimate of 169.2 trillion won and representing an approximate 129% increase year-on-year. The company plans to announce its complete financial report on July 30, at which time it will disclose net profit and classified data for each business segment. **Samsung Electronics' quarterly operating profit has surpassed NVIDIA's operating profit of $53.536 billion (approximately 82 trillion won) from the previous quarter, making it the company with the highest quarterly operating profit in the world.** On Wall Street, analysts are collectively optimistic about the continued record-breaking surge in stock prices of these three major storage chip manufacturers—SK Hynix, Samsung Electronics, and Micron—based on the backdrop of the booming AI infrastructure frenzy driven by the ongoing construction of global AI data centers. Due to the explosive demand for AI computing power infrastructure, the demand for HBM, high-capacity DRAM, and enterprise-level NAND storage chips continues to show explosive growth, with AI servers, high-performance computing, and cloud computing infrastructure construction continuously driving up storage demand However, the speed of new capacity release still struggles to keep up with demand growth. According to Goldman Sachs, the AI bull market is far from over, **but has transitioned from the "AI chip purchasing frenzy" to the second phase of "large-scale construction of AI factories"—meaning that the next round of excess alpha returns will no longer solely belong to the strongest leaders in the AI GPU/AI ASIC fields, but will systematically spread to the entire stack of AI computing infrastructure, including high-performance CPUs for data centers, DRAM/NAND/HBM storage, AI PCBs, liquid cooling systems, data center optical interconnect systems, ABF substrates/glass substrates, MLCCs, electronic fabrics, and extensive wafer foundries.** ![image.png](https://imageproxy.pbkrs.com/https://img.zhitongcaijing.com/image/20260707/1783395162298616.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg) Goldman Sachs believes that **the global bull market surrounding the AI computing chain is far from over, with the market's main narrative having upgraded from the long-standing "programming/code-driven software light asset valuation expansion" since 2008 to "re-pricing of AI computing infrastructure around a series of physical assets."** The latest calculations from Wall Street financial giant Goldman Sachs indicate that the total investment related to AI infrastructure by hyperscale cloud computing companies could exceed $6 trillion by 2030; the global AI capital expenditure benchmark model is expected to grow from $765 billion annually in 2026 to $1.65 trillion annually by 2031, with cumulative capital expenditure from 2026 to 2031 estimated at approximately $7.6 trillion. The power demand for domestic data centers in the United States is expected to rise from 31 GW in 2025 to 66 GW by 2027. ## Bank of America Supports AI Semiconductors: $1.5 Trillion Cloud Capital Expenditure Strengthens "AI Super Bull Market" Narrative Logic The analyst team at Bank of America, led by veteran strategist Vivek Arya, stated: “The stock market frenzy driven by AI semiconductors is not over yet. After a record surge of 88% in the second quarter, **the Philadelphia Semiconductor Index (SOX) corrected by 11% in the third quarter, consistent with its historically weakest seasonal period. We believe that the current correction is a healthy reset rather than any structural change in AI demand. This correction is expected to be a summer reset, with a rebound likely in the fall;** short-term leadership may lean towards low-beta stocks such as NVIDIA (NVDA.US), Texas Instruments (TXN.US), Analog Devices (ADI.US), and the two leading chip design EDA companies—Cadence Design Systems (CDNS.US) and Synopsys (SNPS.US). However, historical experience shows that after a consolidation period, as investors regain strong confidence in the next round of profit and capital expenditure growth cycles, new momentum often emerges.” Analysts also pointed out that ongoing consolidation announcements, such as Texas Instruments (TXN.US) acquiring Silicon Labs (SLAB.US) and ON Semiconductor (ON.US) actively seeking to acquire Synaptics (SYNA.US), may become another enduring theme in the fragmented analog chip industry The analyst team led by Arya at Bank of America has repeatedly emphasized in their research report that the $1.5 trillion cloud computing and AI computing infrastructure capital expenditure maintains the integrity of the super demand cycle for AI computing. Arya and his analyst team stated: “We expect that by 2027, global capital expenditures on cloud and artificial intelligence computing infrastructure will approach $1.5 trillion, which means a potential year-on-year growth of 40% to 50%, strongly supported by the continuous growth of Token scale, a surge in enterprise AI agent adoption, and constrained infrastructure supply. Importantly, the focus of hyperscale cloud computing vendors remains on maximizing utilization and AI-driven performance growth trajectories, rather than optimizing depreciation.” The analysts added that **with increased visibility on cloud computing capital expenditures for 2027 in the second half of 2026, they expect the following areas to regain leadership and momentum**: Micron Technology (MU.US) in the storage chip sector; Advanced Micro Devices (AMD.US) in the CPU computing sector during the July Wall Street Analyst Activity Conference, and Intel (INTC.US), the x86 server CPU leader, at the end of 2026 analyst activities; Applied Materials (AMAT.US), Lam Research (LRCX.US), KLA Corporation (KLAC.US), and Teradyne (TER.US) in the semiconductor equipment sector; MACOM Technology Solutions (MTSI.US) in the high-speed optical interconnect sector for data centers; and Credo Technology (CRDO.US) and Marvell Technology (MRVL.US) in the high-performance network infrastructure sector for AI data centers. Arya and his team noted that AI data center storage chips/storage components currently account for about 35% to 40% of cloud AI capital expenditures, which is two to three times historical levels, but the trading levels and valuation systems of storage chip-related stocks remain below the expected growth levels. The analysts stated: “Investors remain skeptical about pricing durability, expectations for increased supply scale, and customer concentration. **We believe the market underestimates the strong trend towards longer-term product supply agreements (i.e., LTA) and more predictable pricing. As storage chips evolve from cyclical commodities to strategic assets empowering the AI era, valuation multiples should expand significantly.”** **Bank of America analysts reiterated their "Buy" rating on Micron Technology (MU.US), the leader in U.S. storage chips, as their top pick, with a target price set at $1550.** ![image.png](https://imageproxy.pbkrs.com/https://img.zhitongcaijing.com/image/20260707/1783396076568893.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg) Bank of America analysts stated that China's AI models bring about a trend of competition in large AI models between China and the United States, rather than capital expenditure risks. Arya and his team pointed out that China's open-source weighted models, such as Z.AI (formerly known as Zhipu AI)'s GLM, Kimi from Moonlight, DeepSeek's large model, and Alibaba's (BABA.UYS) open-source large model Qwen, have rapidly narrowed the gap with the leading frontier AI laboratories in the United States, while offering significantly lower inference costs. The analysts added that the thorough rise of capable and low-cost models has raised reasonable doubts about the overall economics of future global industry expansion and AI software profitability. The analysts stated, "However, we believe this is a significant positive for the expansion expectations of adoption. **The trajectory of lower-cost AI agent token expenditures will expand actual usage, broaden AI application deployment, and ultimately increase the increasingly strong demand for CPU/GPU computing, HBM/DRAM/NAND storage chips, network infrastructure, data center optical interconnect systems, and power infrastructure. In our view, the greater risk lies in the economic framework of the models, rather than the demand for AI semiconductors.**" The analysts added that investors are still overly focused on the leading positions in large model benchmarking and daily headlines about large language models (LLMs). The ultimate goal of AI is not the best model, but the best labor productivity growth outcomes. As global industries shift from models to AI agents, agent-based workflows, and enterprise automation, value creation should increasingly scale towards AI applications and the computational infrastructure that supports them. Arya and his team stated, "Similar to how the internet era commoditized information but created immense software value, open-source large AI models can accelerate the adoption and penetration of cutting-edge AI applications like AI agents, while frontier laboratories continue to push the boundaries of artificial intelligence technology." ## From Cyclical Products to AI Strategic Assets!   South Korea's Expansion Plans Do Not Deter the Super Bull Market in Storage Chips and the "Storage Super Cycle" Nomura is currently the most aggressive bullish large financial institution on the storage chip sector. In a recent research report, Nomura raised the target price for Samsung Electronics from 340,000 won to 590,000 won, and the target price for SK Hynix from 2,340,000 won to 4,000,000 won, corresponding to potential upside of approximately 118% and 120%, respectively. Nomura's core bullish logic is that **AI has transformed storage from traditional PC/mobile cyclical products into long-term growth assets for data centers: AI inference for agents requires massive key-value caches (KV Cache), and HBM supply significantly lags behind demand**; it is expected that global data center capital expenditures will increase from last year's $1.16 trillion to $6.13 trillion by 2030, with memory's share of data center investment expected to rise from the current 9% to 23%. Therefore, Samsung and SK Hynix's approximately 6 times forward P/E ratio from 12 months ago is clearly undervalued, with room for revaluation towards TSMC's approximately 20 times valuation system Currently, the most optimistic target stock price for Micron on Wall Street comes from Gil Luria, a senior analyst at the well-known asset management firm DA Davidson, who has set a price target of $2,000 per share, maintaining a "Buy" rating. **The target price has been significantly raised from $500 to $2,000. Based on Micron's current stock price of approximately $975.56, this target price corresponds to about 105% potential upside; if the stock price reaches $2,000, it would correspond to a potential market value of approximately $2.29 trillion, which is close to the $2.3 trillion level.** Gil Luria's core bullish logic can essentially be summarized as follows: the expansion of AI computing infrastructure is extending the current storage upcycle. **Micron is no longer just a traditional cyclical DRAM memory company, but has become a bottleneck asset in AI computing with stronger profit visibility and higher valuation multiples under the restructuring of HBM, server DRAM/NAND, and long-term supply agreements. He previously emphasized that this storage cycle is different from the old paradigm of "expansion—oversupply—price collapse," as AI computing construction may indefinitely extend demand prosperity; at the same time, Micron has signed multi-year HBM/storage sales agreements, enhancing the clarity of future revenue and EPS visibility.** The biggest divergence in the storage chip sector currently is not whether AI demand will disappear, but whether the pace of price increases is nearing its peak and whether the market is willing to give storage companies higher valuation multiples. Bank of America’s main line is very clear: **After the Philadelphia Semiconductor Index surged 88% in the second quarter, it corrected about 11% in the third quarter, resembling a seasonal "healthy reset," rather than a structural deterioration in AI demand; the institution expects global capital expenditures for cloud and AI infrastructure to approach $1.5 trillion by 2027, a year-on-year increase of 40%-50%, supported by token growth, agent adoption, and limited infrastructure supply.** More critically, storage chips/components currently account for about 35%-40% of cloud AI capital expenditures, reaching 2-3 times historical levels, yet the overall valuation of storage chip stocks remains below what it should be. Therefore, Bank of America maintains a "Buy" rating on Micron with a target price of $1,550, with the core judgment being that storage is transitioning from a cyclical commodity to AI strategic infrastructure. Morgan Stanley's caution does not equate to a bearish outlook on the storage cycle, **but rather serves as a reminder to the market that the "rate of price change" may have peaked, rather than the "absolute level of prices."** This distinction is very important. TrendForce recently predicted that DRAM contract prices will rise by 13%-18% in Q3 2026 from a strong high base in Q2, while NAND Flash is expected to rise by 10%-15%, which is a significant slowdown compared to previous increases of about 60%; in other words, the slowing marginal increase will suppress the short-term valuation expansion of high-beta storage stocks, but AI inference, hyperscale cloud data centers, and enterprise storage are still maintaining supply-demand tension. **The judgment that "Korean giants building factories equals the peak of the storage cycle" also needs to be unpacked. Large-scale capacity expansion is certainly a medium- to long-term risk, especially as the storage industry has historically fallen into downturns due to excessive capital expenditures after high prosperity; however, the current new clusters in Korea are more about capacity planning for the 2030s, rather than an immediate supply shock in 2026-2027.** \*\* SK Hynix's investment plan in South Korea includes the Yongin semiconductor cluster, expansion of Cheongju NAND and HBM packaging, as well as the southwestern semiconductor cluster which is still in the planning stage. The first wafer fab in Yongin is expected to be operational by 2027, with a total of four fabs completed by 2033, while the southwestern cluster is more of a long-term plan. For the current investment cycle, what truly determines stock prices are the capital expenditure guidance from ultra-large cloud vendors over the next four to six quarters, the execution of long-term agreements for HBM/server DRAM, enterprise-level SSD demand, and the continuity of NAND prices, rather than a nominal capacity blueprint spanning around ten years. ![image.png](https://imageproxy.pbkrs.com/https://img.zhitongcaijing.com/image/20260707/1783394873799101.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg) This week, the well-known Wall Street investment firm Nomura released a new research report, **refuting the "semiconductor peak theory." The key to Nomura's rebuttal of the "semiconductor peak theory" is not simply stating that AI chips will continue to rise, but pointing out that the demand for AI cloud infrastructure is spreading from a shortage of single-point GPUs to a systemic mismatch of components. According to Nomura's research framework, AI server revenue is expected to grow by 78% and 76% in 2026 and 2027 respectively, with global data center projects increasing from 240 to 280, including about 50 gigawatt-level projects. The new computing power deployment expected by 2027 is projected to reach 32GW, with visibility of 23GW already in 2028**; however, the real bottleneck is shifting from NVIDIA AI GPU and Google TPU capacity, TSMC's CoWoS advanced packaging to memory chips, wafer-level substrates, AI PCBs, copper-clad laminates (CCL), electronic fabrics, MLCCs, glass substrates/ABF substrates, IC substrates, high-end capacitors, power management chips, and optical high-speed interconnect components for data centers ### Related Stocks - [SSNGY.US](https://longbridge.com/en/quote/SSNGY.US.md) - [SMSN.UK](https://longbridge.com/en/quote/SMSN.UK.md) - [159516.CN](https://longbridge.com/en/quote/159516.CN.md) - [512760.CN](https://longbridge.com/en/quote/512760.CN.md) - [07747.HK](https://longbridge.com/en/quote/07747.HK.md) - [588170.CN](https://longbridge.com/en/quote/588170.CN.md) - [07709.HK](https://longbridge.com/en/quote/07709.HK.md) - [SMH.US](https://longbridge.com/en/quote/SMH.US.md) - [09347.HK](https://longbridge.com/en/quote/09347.HK.md) - [XSD.US](https://longbridge.com/en/quote/XSD.US.md) - [03121.HK](https://longbridge.com/en/quote/03121.HK.md) - [09747.HK](https://longbridge.com/en/quote/09747.HK.md) - [07347.HK](https://longbridge.com/en/quote/07347.HK.md) - [PSI.US](https://longbridge.com/en/quote/PSI.US.md) - [SOXX.US](https://longbridge.com/en/quote/SOXX.US.md) - [FTXL.US](https://longbridge.com/en/quote/FTXL.US.md) - [USD.US](https://longbridge.com/en/quote/USD.US.md) - [SSG.US](https://longbridge.com/en/quote/SSG.US.md) - [SOXQ.US](https://longbridge.com/en/quote/SOXQ.US.md) - [DRAM.US](https://longbridge.com/en/quote/DRAM.US.md) - [MU.US](https://longbridge.com/en/quote/MU.US.md) - [MULL.US](https://longbridge.com/en/quote/MULL.US.md) - [MUU.US](https://longbridge.com/en/quote/MUU.US.md) - [SNDK.US](https://longbridge.com/en/quote/SNDK.US.md) - [SNXX.US](https://longbridge.com/en/quote/SNXX.US.md) - [SOXL.US](https://longbridge.com/en/quote/SOXL.US.md) - [RAM.US](https://longbridge.com/en/quote/RAM.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [NVDL.US](https://longbridge.com/en/quote/NVDL.US.md) - [INTC.US](https://longbridge.com/en/quote/INTC.US.md) - [INTW.US](https://longbridge.com/en/quote/INTW.US.md) - [AMD.US](https://longbridge.com/en/quote/AMD.US.md) - [AMDL.US](https://longbridge.com/en/quote/AMDL.US.md) - [AMAT.US](https://longbridge.com/en/quote/AMAT.US.md) - [KLAC.US](https://longbridge.com/en/quote/KLAC.US.md) - [TER.US](https://longbridge.com/en/quote/TER.US.md) - [MRVL.US](https://longbridge.com/en/quote/MRVL.US.md) - [TSM.US](https://longbridge.com/en/quote/TSM.US.md) - [SOXS.US](https://longbridge.com/en/quote/SOXS.US.md) ## Related News & Research - [Samsung Non-Chip Workers Plan Protest Rally Over Bonus Disparity as AI Memory Boom Widens Internal Pay Gap](https://longbridge.com/en/news/291911119.md) - [What Is Going on With Applied Materials Stock on Tuesday?](https://longbridge.com/en/news/291959558.md) - ['Prepare for What's Coming,' Says Investor About Sandisk Stock](https://longbridge.com/en/news/291736056.md) - [Fortitude Advisory Group L.L.C. Has $1.15 Million Stake in Marvell Technology, Inc. $MRVL](https://longbridge.com/en/news/291771288.md) - [Performance Comparison: Micron Technology And Competitors In Semiconductors & Semiconductor Equipment Industry](https://longbridge.com/en/news/291784776.md)