TL;DR

• On July 2, after a collective pullback in memory stocks, the market began to question whether the AI-driven memory price increase cycle is nearing its peak.
• Bank of America believes that Meta's plan to lease computing power, Changxin's entry into iPhone supply, and South Korea's 800 trillion won plan have not rewritten short-term supply and demand.
• Korean exports, price forecasts, and cloud vendor capital expenditures remain strong, but there is still a divergence regarding whether AI demand will materialize after 2027.

On July 2, AI and semiconductor stocks faced a concentrated sell-off, and the memory sector was swept up in it. Stocks such as SK Hynix, Samsung, Micron, Kioxia, and Western Digital saw noticeable pullbacks that day, with market concerns triggered not by a single earnings report figure but by three concurrently emerging bearish clues: Meta was reported to plan to lease out AI computing power, Changxin Memory (CXMT) might enter Apple's iPhone supply chain, and South Korea announced a semiconductor and storage investment plan of about 800 trillion won. The latest memory industry report from Bank of America indicates that these messages are worth tracking, but they are currently not enough to prove that the AI memory cycle has reversed.

The main theme of this round of memory market is not traditional PC or mobile restocking, but rather the sustained pull of AI data centers driving demand for high-end memory such as HBM, LPDDR5, and enterprise SSDs. For the market, the issue behind the stock price pullback is straightforward: Is AI demand beginning to be falsified, and will supply suddenly amplify? Bank of America's answer tends to be cautiously optimistic. Short-term price and export data still support an upward trend in the memory market, but investors are no longer just focused on price increases; they have begun to question how long these price increases can be sustained.

Meta's Computing Power Leasing Does Not Mean AI Orders Have Already Been Cut

The most direct concern from the market comes from Meta. According to Bloomberg Law, Meta is planning a cloud infrastructure business to sell AI computing power and model access to external customers. Media such as Tom's Hardware later interpreted this as a source of "excess capacity" concerns. If Meta indeed reduces long-term chip and component orders due to excess AI servers, demand for HBM, LPDDR5, and enterprise SSDs would be impacted.

However, feedback from the chip supply chain obtained by Bank of America does not support this inference. The report states that Meta's AI data centers are still actively adopting advanced memory, and long-term chip and component orders are strengthening, with no evidence of "server surplus leading to order cuts." At least from the current supply chain perspective, Meta appears to be continuing to expand its AI infrastructure rather than preemptively contracting.

This is also why memory stocks are so sensitive to rumors about individual customers. AI servers consume memory far more than traditional servers; HBM is used for GPU acceleration, while LPDDR5 and enterprise SSDs also meet higher bandwidth, lower power consumption, and higher storage performance demands. Once large cloud vendors cut capital expenditures, high-end memory prices and order expectations will be quickly under pressure. Conversely, as long as hyperscale cloud vendors continue to increase investment, short-term supply and demand tension will be hard to alleviate rapidly.

Changxin's Entry into iPhone: More Like a Negotiation Chip for Apple in the Short Term

The second concern is that Changxin Memory may enter Apple's iPhone supply chain. If Apple widely adopts Changxin DRAM, the pricing power of Korean and American memory giants in mobile DRAM may be weakened, and it will also reinforce expectations for accelerated domestic substitution in China.

However, this impact has clear limitations in the short term. Public reports mainly cite Bank of America's view that if Changxin wants to enter the iPhone supply chain, it must overcome U.S. semiconductor restrictions, Apple's quality and specification certifications, and potential intellectual property litigation risks. The related low-power DRAM also needs to meet requirements for speed, power consumption, and ECC, while Korean and American giants still hold strong technological and patent barriers in advanced mobile DRAM.

Even if Apple tries a small amount of Changxin chips in low-end models like the iPhone 18e, the actual order volume may still be limited. Bank of America believes that the demand for low-end models in the Chinese market is relatively limited, and the contribution to procurement scale is not large. A more realistic impact is that Apple uses this to strengthen its negotiation position with Korean and American giants regarding contract prices in the second half of 2026 or 2027, rather than immediately changing the global supply and demand structure for mobile DRAM.

The long-term impact of Changxin cannot be ignored. China's localization efforts will continue to change some customers' procurement choices. But in terms of the current cycle, it is still not evidence of “a sudden influx of supply.” What the market is truly concerned about is whether Changxin can reliably pass Apple's quality certification, how U.S. restrictions will be enforced, and whether its production capacity can expand from low-end models to higher-spec products.

The 800 Trillion Won Plan from South Korea Is Not New Supply for the Next Two Years

The third concern comes from the South Korean government's announcement of a large-scale semiconductor and storage investment plan. According to information released by South Korea in late June, the scale of this plan is about 800 trillion won, equivalent to approximately 520 billion dollars, involving Samsung, SK Hynix, new wafer fabs, and HBM capacity expansion. Such a figure can easily be interpreted by the market as a new round of major capacity expansion.

However, Bank of America believes that this is not a direct supply signal for the current cycle. The report states that the relevant new clusters and supporting construction are more oriented towards long-term industrial planning, and some projects still have a long way to go before actual quality production, which will not suddenly release a large amount of new capacity in the next two to three years.

The memory industry has previously experienced several instances where “capital expenditure peaks correspond to cycle peaks,” so any large-scale plant construction plan triggers vigilance. However, current enterprise demand is focused on AI-related products such as HBM, SOCAMM, and enterprise SSDs, with constraints on advanced processes, packaging, yield, and customer certification that are stronger than traditional DRAM. Long-term investment plans do not equate to short-term effective supply, especially when high-end memory capacity remains constrained.

Exports and Prices Are Still Rising, but the Market Starts to Question How Long It Can Last

Short-term data still leans towards the bulls. South Korean official data shows that semiconductor exports in June were about 44.8 billion dollars, a year-on-year increase of 199.5%, with total exports of approximately 102.25 billion dollars, a year-on-year increase of 70.9%. This data corroborates the continuing upward trend of memory prices.

Price forecasts also remain strong. TrendForce raised its forecast for the average sales price of DRAM in the third quarter of 2026 to a sequential increase of 13%-18%, compared to a previous expectation of 3%-8%. Bank of America's report estimates the DRAM prices for the second to fourth quarters of 2026 to increase sequentially by 53%, 17%, and 7%, with NAND expected to grow 65%, 13%, and 1% in the same period. The two sets of forecasts are generally consistent in the direction for the third and fourth quarters, with differences mainly in the rhythm of the second quarter and some price assumptions for NAND.

In June, semiconductor exports were about 44.8 billion dollars, a year-on-year increase of 199.5%.

The forecast for DRAM prices in Q3 2026 has been raised to a sequential increase of 13%-18%.

The spot market also indicates that supply remains tight. According to Bank of America, the spot price of 16Gb DDR5 reached a new high of 47 dollars in early July, 16Gb DDR4 is about 75 dollars, and 512Gb NAND wafers are around 20 dollars. Prices had a pullback from April to May but rebounded in June. DRAM manufacturers prioritize the production of HBM, further compressing traditional DRAM supply.

The higher the prices, the more the market worries about a cycle reversal. The current divergence is not whether memory is strong in the short term, but whether this strength has been fully reflected in stock prices and whether AI demand can continue to absorb high prices.

Samsung's Q2 Preview Is a Close Test of Memory Market Conditions

Samsung is expected to release its preliminary results for Q2 on July 7, which will become a short-term window for the market to test memory market conditions. According to Moneycontrol citing Bloomberg, the market is concerned whether its operating profit can alleviate the pressure from the cooling AI transactions.

Overall profits may be affected by factors such as a special bonus included in Q1, the pressure on smartphone business profit margins, etc., potentially falling short of some optimistic expectations. However, Bank of America believes that the operating profit of the memory sector is expected to exceed market consensus, primarily benefiting from the increase in average sales prices.

This is also the delicate part of memory stocks currently. The overall profit of the company may be disturbed by factors like smartphones and bonuses, but the memory business itself is still in a phase of rising prices. If Samsung's memory sector performs significantly better than expected, it will reinforce the judgment that "the July 2 pullback was more about concentrated emotional and worry release." If price transmission or profit margins do not meet expectations, concerns about the cycle peaking will continue to rise.

Valuation cannot completely eliminate risk. After a significant expansion in profitability, memory stocks still have a relatively low P/E ratio, and ROE has improved significantly. However, the sector has already seen a large increase since 2026, and the pullback on July 2 indicates that investors are becoming more sensitive to every potential negative news.

The Controversy Ultimately Falls on AI Capital Expenditures After 2027

The larger background supporting memory demand is that cloud giants' capital expenditures are still expanding. According to media summaries and analyst estimates, large cloud vendors like Amazon, Microsoft, Alphabet, and Meta may reach capital expenditures related to AI of about 700 billion dollars in 2026, a significant increase over the previous year, and this high level may continue into 2027 and 2028. This estimate is not a unified official guidance from the four companies; different statistics may also lead to discrepancies in whether companies like Oracle are included.

Trends in capital expenditures, cloud revenues, and profit margins of large cloud vendors. The capital expenditures in 2026 are expected to be around 700 billion dollars, which is an important support for whether high-end memory demand can continue.

Therefore, this pullback in memory stocks seems more like the market testing three questions in advance: Whether Meta is truly cutting AI infrastructure investment, whether Changxin can shift from a symbolic entry into Apple's supply chain to large-scale substitution, and whether South Korea's long-term investment plan will ultimately bring a new round of supply pressure.

Currently, these questions have not provided answers sufficient to overturn the cycle. Exports, spot prices, price forecasts, and cloud vendor capital expenditures still point to strong demand. However, risks have not disappeared. AI capital expenditures need to continue to materialize after 2027, geopolitical restrictions may change supply chain choices, and the localization process in China will continue to affect negotiation relationships between Korean and American giants and customers.

Whether the memory cycle has peaked is not a conclusion that has been confirmed by data at this time. More accurately, the market has shifted from "focusing only on price increases" to "validating how long price increases can last."

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。