TL;DR

• According to Citrini analyst Jukan, South Korea saw significant year-on-year growth in multiple storage export amounts and kilogram prices before June 20, but this data remains preliminary and is derived from social media summaries.
• This set of data reinforces the judgment that AI memory demand is spilling over, but MCP cannot be directly equated with HBM, and kilogram prices do not equate to multiple times increases in the price of single chips.
• Related entities: SK Hynix, Samsung, Micron, Nvidia.

According to analyst Jukan of Citrini, South Korea's storage export amounts and kilogram prices have seen significant year-on-year growth before June 20, prompting the market to reconsider whether storage manufacturers are obtaining AI infrastructure bottleneck premiums.

The importance of this matter lies not just in another set of semiconductor export figures, but in its simultaneous touch on two variables that concern investors the most: increasing shipment amounts and rising unit weight corresponding to export values. The former points to demand intensity, while the latter indicates price and product structure migrating towards higher value products. For storage stocks, this holds more significance than simply "selling more," as it affects revenue, gross margins, and the potential for EPS upward revisions.

Over the past year, the market has accepted HBM (High Bandwidth Memory) as a scarce resource in AI servers. The debate lies in whether this scarcity is merely a price increase for a few high-end products or if it has begun to spill over to the broader DRAM, NAND, and SSD storage chains. If it is the former, storage stocks will still resemble a cyclical repair trade. If it is the latter, the valuation anchors for SK Hynix, Samsung, and Micron may shift from "inventory cycle" to "AI infrastructure bottleneck" portions.

The South Korean data provides strong signals but is not conclusive. Especially the subcategory and kilogram price data before June 20 are more suitable as preliminary observations under the social media summary framework, and cannot be directly considered as official complete confirmations. Its value lies in advancing a somewhat narrative question to a stage where it can be cross-verified using trade amounts, price indicators, and company guidance.

South Korea's exports have given the market a price signal

This set of data's most direct implication is that the storage economy may not only be recovering in terms of shipment volume; prices and product combinations are also becoming more expensive.

In the preliminary export data from South Korea between June 1 and 20, export amounts for various categories like DRAM, NAND/Flash, MCP, and SSD saw significant year-on-year growth. Notably, non-module DRAM export amounts rose nearly four-fold year-on-year, while module-inclusive figures grew over three-fold; NAND/Flash and SSD export amounts also increased sharply. Of greater market concern is the kilogram price, with some DRAM and NAND-related categories seeing year-on-year increases exceeding 500%.

These numbers need to be viewed with context. The data from the first 20 days resembles a mid-month snapshot of South Korean trade data that can indicate direction and slope, but does not represent final full-month figures. The classification of subcategories may not fully align with investors' understanding of product definitions, making it unsuitable for direct extrapolation into annual profit models.

A more stable reference comes from the already published May data. According to South Korean media reports based on official data, total exports in May reached $87.75 billion, a year-on-year increase of 53.2%; semiconductor exports were $37.16 billion, a year-on-year increase of approximately 169%, setting a monthly record, accounting for 42.3% of total exports. Exports of computers and related equipment also surged, with the media linking this to AI server SSD demand. The preliminary export figures from June 1 to 10 were similarly strong, with total exports at $28.6 billion, a year-on-year increase of 86%, and semiconductor exports approximately $11 billion, more than tripling year-on-year.

This makes the social media summary data from before June 20 no longer an isolated signal. It forms continuity with the previous official export trends. For investors, continuity is more important than single-month spikes, as it determines whether earnings revisions can transition from one-off surprises to multiple quarters of model adjustments.

Soaring kilogram prices do not equal fivefold increases in chip prices

The most easily misread aspect of this data is the direct translation of soaring kilogram prices into "price increases for each chip." A more accurate statement is that kilogram prices reflect price increases, product structure advancement, and statistical definitions working together.

Some categories in the South Korean export data calculate average unit prices based on weight. This metric is easily understood for commodities. However, for semiconductors, the value of the same kilogram of goods can vary significantly. A kilogram of low-end storage chips and a kilogram of HBM, high-capacity DRAM, or complex packaged products do not have the same value density. The rise in kilogram prices may be attributed to either price increases in similar products or a shift in export structure towards higher value products.

This precisely addresses the core of the AI trade. AI servers require memory systems with higher bandwidth, larger capacity, and lower latency, where the value density of HBM and high-end DRAM far exceeds that of ordinary storage products. As the share of these products in the export structure rises, the average export value per kilogram will be elevated. What the market sees is not that all storage chips uniformly increased fivefold in price, but rather that the growing proportion of high-end products combined with price increases is changing the income quality of the storage chain.

The MCP metric also needs to be approached with special caution. The market often observes MCP as a proxy indicator related to HBM, given that HBM typically involves multi-chip stacking and packaging. However, MCP (multi-chip package) is not synonymous with narrow HBM and may also encompass other multi-chip packaging products. Strengthening MCP export amounts and unit prices can support the directional judgment that "high-end packaging memory demand is strong," but cannot be directly written as HBM export amounts.

This limitation does not diminish the data's value; rather, it makes it more suitable for investment judgments. The truly useful conclusion is not how much a particular product has precisely increased, but rather that multiple storage categories have simultaneously experienced increases in amounts and unit values, indicating that AI demand may no longer be confined to HBM alone. It is influencing a broader storage pricing system through capacity allocation, product structure, and customer procurement.

HBM shortages are changing storage manufacturers' pricing positions

If we only look at HBM itself, the market is already well aware of its scarcity. The new question is why HBM shortages would affect DRAM, NAND, and SSD.

The mechanism is not complex. Storage manufacturers have limited advanced capacity, R&D resources, and customer certification capabilities. When Nvidia and cloud manufacturers continue to lock in high-value products like HBM and high-capacity DRAM, manufacturers prioritize directing their resources toward areas with higher returns and stronger order visibility. This leads to a continuous tight supply of high-end products and may indirectly squeeze the supply elasticity of ordinary DRAM, NAND, and SSD.

SK Hynix is the most direct beneficiary of this logic. The market generally believes that its share of HBM is in a leading position. According to industry reports and broker analyses, SK Hynix has high visibility regarding its HBM capacity by 2026, with customer demand exceeding supply capabilities and high-value product sales increasing. For a storage manufacturer, customers locking in capacity and the growth of high-end product sales change not just the next quarter's income but also the market's assessment of its pricing power. The core issue for traditional cyclical stocks is how long prices can rise, while the core issue for bottleneck assets is how much premium customers are willing to pay for guaranteed supply.

Samsung and Micron's logic is slightly different. Samsung has more scale in NAND and overall storage capacity while still striving to catch up with high-end HBM customer certification. Micron benefits from the expansion of high-end memory demand and diversification of its supply chain. For these two companies, the market is not trading on whether they have completely replicated SK Hynix's HBM pricing power but rather on how, if HBM shortages spill over into high-end DRAM, enterprise-level SSDs, and NAND prices, their gross margin elasticity will be stronger than in the last cycle.

Intel CEO Pat Gelsinger mentioned in an interview with No Priors that the AI infrastructure bottleneck is spreading from GPUs to memory, CPUs, optical interconnects, power conversion, advanced packaging, and materials. The focus here is not to rewrite the issue as Intel's strategy but rather to outline a larger context: the constraints of AI data centers are no longer limited to a single GPU; any factor that limits cluster expansion and efficiency may gain new pricing power.

Memory is one of the segments observed early on through trade data. No matter how powerful a GPU is, it still requires sufficient memory bandwidth and capacity to feed data. As inference and intelligent agent tasks increase, the system's demand for memory, storage, and scheduling resources becomes more complex. The value of South Korean export data lies in its ability to ground the macro judgment of "spreading AI infrastructure bottlenecks" to changes in storage export amounts and unit values.

Storage stocks must still accept cyclical constraints

For investors, this round of storage increases appears more as a combination of "real economic acceleration and future earnings re-evaluation" rather than simply telling a story. Export data indicates that demand and prices already have real support, and what the market is truly buying is whether revenue, gross margins, and EPS for 2026 will continue to be revised upward.

If subsequent financial reports validate this trajectory, SK Hynix's valuation premium becomes the easiest to explain: leading HBM share, customer order locks, and high-value product boosts jointly make for improved visibility. Samsung's key will depend on whether its pursuit of high-end HBM translates into actual orders, as well as whether NAND and SSD prices can form broader support. Micron needs to prove that price increases in high-end DRAM and data center storage can penetrate to gross margins and guidance.

Risks also lie here. Storage remains a strongly cyclical industry, and supply expansion, inventory changes, and customer procurement rhythms can all affect pricing. The preliminary export data from the first 20 days can signal steepening slopes but cannot definitively prove yearly certainties. Rising kilogram prices may indicate increased value density, yet they cannot entirely separate the proportion of average selling price increases and product structure changes. Stronger MCPs can serve as proxy signals related to HBM but cannot be directly equated to HBM export.

Another risk comes from AI capital expenditure itself. If investments in power, thermal management, packaging, or overall computing capacity slow down, storage demand will also be affected. The spreading bottlenecks are both the reason for storage gaining premiums and a potential constraint. When other segments in the system get stuck first, the pace of memory demand release may also be delayed.

Earnings reports will determine whether valuation anchors can switch

This round of re-evaluation ultimately needs to be reflected in company statements, not just remain within trade data. The full export data for June will first provide the market a more comprehensive confirmation: whether the high growth of the first 20 days continues, whether price indicators remain high, and whether the strength of NAND and SSD is merely driven by short-term bulk orders.

A more critical validation comes from the Q2 and Q3 earnings reports of SK Hynix, Samsung, and Micron. The market needs to see continued realization of HBM shipments and prices, synchronized improvement in average prices for DRAM and NAND, and margin increases driven by data center SSD demand, rather than just reflected in revenue scale. If gross margins and guidance do not keep pace with the slopes presented by export data, the re-evaluation will quickly revert to cyclical trading.

The current more prudent judgment is that the South Korean storage export data for the first 20 days is strong enough to support the market's upward revision of storage manufacturers' profit elasticity and to reconsider the AI infrastructure bottleneck premiums. However, it is not sufficient to prove that the storage industry has escaped the cycle. What will determine whether the valuation anchor can shift is not how high a single year-on-year number is, but whether prices, product structures, and profit margins can all hold up in the coming quarters.

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。