From HBM to Cold Storage: Storage Giants at a Crossroads, What Will Drive the Next Market Cycle?

Written by: DaiDai, MSX Maitong

Edited by: Frank, MSX Maitong

Core Overview:

• This round of storage market is not just an ordinary price cycle, but a gradual expansion of AI infrastructure demand from HBM to server DRAM, enterprise SSDs, NAND, and HDDs. The market trading objects have also shifted from "faster storage" to "more, cheaper, and long-term storage needs."
• The six companies correspond to different investment logics: SK Hynix (SKHY.M) trades on its leading position in HBM and order certainty; Samsung focuses on technology catch-up and market share recovery; Micron is the most direct HBM and DRAM proxy in the US market; SanDisk is most sensitive to price increases in NAND and enterprise SSDs; Seagate and Western Digital benefit from high-capacity HDDs, data lakes, and cold storage demand.
• Since 2025, all six companies have undergone significant revaluation, but as of mid-July 2026, they have generally retreated about 20%-30% from their peak levels. The sector is transitioning from "demand explosion and profit recovery" to a phase of high expectations, high valuations, and high volatility.
• Historical performance shows that the true catalysts for sustained market trends are usually not product launches or samples, but rather earnings reports exceeding expectations, core customer certifications, large-scale deliveries, locked orders for the next fiscal year, and upward revisions of revenue, price, and profit margin guidance.
• The core of the next market cycle lies in who can convert demand into long-term contracts, higher product prices, controllable capacity expansion, and sustained profit and free cash flow growth.

In June 2026, Micron (MU.M) released a nearly flawless earnings report.

The company reported quarterly revenue of $41.456 billion, a non-GAAP gross margin of 84.9%, adjusted free cash flow exceeding $18.3 billion, and raised its revenue guidance for the next quarter to around $50 billion. In other words, whether it is product prices, profit margins, or order visibility, they all point to the same conclusion: AI storage demand remains strong.

However, the market's reaction was not as straightforward as before.

After initially surging post-earnings, Micron quickly gave back its gains, and SanDisk, SK Hynix, Samsung Electronics, Western Digital, and Seagate also retreated from their June highs. As of July 22, the six companies had generally retreated about 20%-30% from their peak levels, with some exceeding 30%.

Ironically, the fundamentals did not suddenly weaken, and AI data centers did not stop expanding; investors simply stopped questioning whether AI would create greater storage demand and began to ask how much longer product prices could rise. Would new capacity change the supply-demand relationship again in 2027-2028?

Thus, the storage sector has reached a new crossroads.

1. Expansion of AI Storage Market Since 2025

The storage market since 2025 was initially ignited by HBM.

Large model training requires GPUs, and large-scale GPU clusters need high bandwidth and lower latency memory systems. Therefore, HBM, by vertically stacking multiple layers of DRAM, significantly increases data throughput capacity, quickly becoming one of the most critical, highest-margin, and tightest supply components in AI accelerators.

SK Hynix (SKHY.M), with its leading HBM3E products and major customer relationships, was the first to complete the revaluation of its profit center.

By the third quarter of 2025, the company had indicated that discussions on HBM supply for the next fiscal year were basically complete, with HBM4 set to ship in the fourth quarter. The next year's DRAM and NAND production were also largely covered by customer demand, and by the first quarter of 2026, the company's revenue, operating profit, and net profit continued to set records.

Micron (MU.M) became the core target for trading HBM and server DRAM supply-demand in the US market, especially as HBM4 entered large-scale shipments, server memory prices rose, and multiple long-term customer agreements collectively boosted revenue, gross margins, and cash flow.

Thus, for US investors, the significance of its earnings report is no longer limited to the company itself, so when Micron raises revenue, price, and profit margin guidance, the market typically reassesses the profitability of the entire storage industry.

Samsung Electronics took a different path.

In 2025, the market primarily priced in its lagging progress in HBM products, customer certifications, and yield compared to SK Hynix. Entering 2026, as HBM4 began commercial deliveries, HBM4E entered the sampling stage, and the profitability of the storage business significantly improved, Samsung's valuation logic shifted from "lagging" to "catching up."

The company's preliminary performance for the second quarter of 2026 showed quarterly sales of approximately 171 trillion won and operating profit of about 89.4 trillion won. Of course, this data includes smartphones, foundry, display, and consumer electronics businesses and cannot be directly viewed as the independent performance of the storage division.

In fact, what truly drove the sustained expansion of this round of market was not just HBM itself.

When manufacturers allocate more wafers, capital expenditures, and advanced packaging resources to high-value AI products while maintaining relatively strict supply discipline, the supply-demand pattern for server DDR5, ordinary DRAM, and NAND also improves.

Meanwhile, AI is moving from training to large-scale inference, and storage demand is beginning to spread from around GPUs to the entire data system.

Model weights, vector databases, key-value caches, inference contexts, and high-frequency access data are driving growth in capacity and performance demand for enterprise SSDs; training data, videos, multimodal materials, inference logs, historical model versions, and compliance archiving require lower-cost, scalable large-capacity storage.

SanDisk (SNDK.M) has become the most resilient target in the sector during this phase.

After completing its spin-off from Western Digital (WDC.M) in February 2025, SanDisk transformed from a mixed HDD and flash company into a purer NAND and SSD target in the US market. In the latest quarter, the company's data center revenue grew by 233% quarter-on-quarter, and it signed multiple new business model agreements with financial guarantees, while announcing a stock buyback of up to $6 billion after repaying debt.

As AI data continues to accumulate, demand is also spreading from high-speed storage to large-capacity storage.

Because model training data, videos, multimodal data, inference logs, historical versions, and compliance archiving do not all need to be stored long-term in high-cost SSDs, a large amount of low-access-frequency data that must be retained long-term will ultimately still need to enter nearline HDDs and hierarchical storage systems.

Western Digital's latest quarterly non-GAAP gross margin has reached 50.5%, with free cash flow of $978 million, and it expects nearline HDD Exabyte shipments to maintain mid-double-digit growth over the next three to five years.

Seagate (STX.M), on the other hand, has gained a more obvious technological premium through the commercialization of HAMR, with its latest quarterly non-GAAP gross margin reaching 47%, free cash flow of $953 million, and its highest capacity product, Mozaic 4+, which has begun bulk shipments to two leading hyperscale customers.

Therefore, this round of storage market cannot simply be summarized as a price increase; it is more like a layered expansion of storage demand radius following the deepening of AI infrastructure construction, where HBM addresses the bandwidth issues around GPUs, server DRAM handles workloads during the computing process, enterprise SSDs support high-frequency data access, and HDDs accommodate the ever-expanding data lakes and long-term archiving.

To some extent, the market initially traded on "AI needs faster storage," and then began trading on "AI also needs more storage." This industrial logic of spreading from HBM to server DRAM, NAND, enterprise SSDs, and HDDs is also fully reflected in the stock performance of the six companies.

According to a unified adjusted basis, from the beginning of 2025 or the first trading day after the spin-off, as of July 22, 2026, all six companies have undergone significant revaluation, but the magnitude of the increase and the pace of initiation are not the same:

• SK Hynix and Micron benefited first from the upturn in HBM and server DRAM.
• Samsung then began trading on HBM technology catch-up and market share recovery; SanDisk, after completing its spin-off, became the most resilient target in this round of market with its purer NAND and enterprise SSD exposure.
• Western Digital and Seagate took over the upward trend in the first half of 2026, as the market began to incorporate nearline HDDs, high-capacity products, and cold data demand into the valuation system of AI infrastructure.

If we break down the market over the past year and a half, it can roughly be divided into four phases:

• In the first half of 2025, the market first traded HBM. SK Hynix and Micron became the most direct beneficiaries, as investors began to confirm that HBM was not just a short-term inventory replenishment but a long-term high-margin product formed alongside the expansion of AI accelerators.
• In the second half of 2025, the market expanded to ordinary DRAM and NAND. The occupation of advanced wafers, packaging, and testing resources by HBM tightened the supply of server DRAM and other storage products, prompting the market to raise profit expectations for companies like Samsung and SanDisk.
• In the first half of 2026, enterprise SSDs and HDDs took over. As AI moved from training to inference, storage demand was no longer just around high-speed memory near GPUs; training data, inference logs, and long-term archiving needs drove HDDs and cold storage to be repriced.
• Entering June-July 2026, the market shifted to high expectations and high volatility. The peak stock prices of the six companies were almost all concentrated in June, and even after Micron and others continued to deliver strong earnings reports, the sector failed to maintain the previous unilateral upward trend.

This round of pullback indicates that the market's pricing logic is changing, as the previous stock price increases primarily relied on three reinforcing consensus: AI storage demand continues to expand, industry supply remains tight, and product prices and profit margins still have room to rise.

However, after experiencing significant revaluation, investors began to consider several other questions, such as whether current valuations have already discounted future growth, whether capital trading is overly crowded, whether new capacity in 2027-2028 will gradually be released, and how much strong demand can ultimately be converted into long-term orders and free cash flow.

Therefore, this collective pullback is more like a sign that the storage market is transitioning from the first phase to the second phase, with subsequent developments depending more on who can provide longer order visibility, stronger pricing power, and more sustainable profit growth.

II. Different Stories of Six Companies in the Same Storage Sector

If we only look at stock prices, the six companies seem to belong to the same AI storage market trend.

However, from the perspective of industry position, profit elasticity, and the next stage of catalysts, they actually represent six different pricing logics.

SK Hynix remains the strongest company in terms of industry position and order certainty among the six.

Its core advantage lies not only in its leading market share in HBM but also in its ability to lock in customers, prices, and next year's capacity earlier. Compared to the release of ordinary products, early coverage of orders makes it easier to enhance the visibility of future revenue and profit margins, which also facilitates the formation of a sustained market trend.

However, after SKHY went public on NASDAQ in July 2026, the company added a new trading variable. SKHY was issued at $149 and closed at around $168 on its first day. Due to the limited circulation of shares in the initial U.S. market, ADR once formed a significant premium relative to Korean common stocks. Of course, as the conversion and arbitrage mechanisms gradually open up, this premium may converge.

Therefore, it is necessary to analyze the company's fundamentals separately from the supply and demand of ADR.

Korean common stocks reflect more on HBM orders, prices, and profit margins, while SKHY is also influenced by U.S. capital access, circulation, ETF allocation, and arbitrage mechanisms. This means that even if the company's operations remain strong, ADR may experience fluctuations independent of fundamentals due to increased supply.

Micron is the most direct U.S. proxy and the easiest target for pre-trading.

Its advantage lies in covering HBM, server DRAM, NAND, and enterprise SSDs, supported by U.S. manufacturing and supply chain policies. When Micron raises product prices, gross margins, and performance guidance, the market often adjusts the profit expectations for the entire storage sector in sync. Its financial reports serve not only as a catalyst for the company itself but also as an important price signal for the global storage industry.

However, its problems are equally apparent: the market adjusts its expectations for it very quickly.

In simple terms, when valuations are low and industry expectations are weak, an earnings report that exceeds expectations can lead to a sustained revaluation. Once entering a high-expectation phase, merely maintaining strength is not enough; the stock price needs higher prices than the most optimistic market predictions, longer order visibility, and more significant profit margin upgrades.

This explains why Micron's stock price remains highly elastic after its earnings report, but the sustainability of the trend begins to decline.

Samsung Electronics is relatively lagging in terms of price increase, but its share recovery elasticity is the most evident.

It is not the most fundamentally pure company in this round of market trends. After all, its wafer foundry, smartphones, consumer electronics, and display businesses dilute the impact of storage profit improvements on the group's overall profitability. The slower progress in HBM customer certification has also led to its stock price performance lagging behind SK Hynix and Micron.

However, looking at it from another perspective, Samsung has the clearest "expectation gap" among the six companies. The market already knows that SK Hynix is the leader in HBM but still cannot determine how many high-end HBM orders Samsung will ultimately secure.

Therefore, the truly meaningful catalyst for Samsung in the next stage lies in whether core customer certifications are completed, whether products enter large-scale procurement, whether the proportion of HBM revenue increases, and whether market share and storage business profit margins can improve in sync. Once these indicators begin to materialize, Samsung's narrative will shift from "technological catch-up" to "share recovery."

SanDisk is gradually shifting from being a beta for NAND price increases to a visibility-driven trading model.

It is the company with the highest stock price elasticity in this round of market trends and also the one with the greatest risk when expectations decline. On one hand, its NAND and enterprise SSD businesses are highly pure and extremely sensitive to product prices and industry inventory changes; on the other hand, the independent valuation system post-spin-off, data center revenue growth, and long-term business agreements further amplify the potential for profit upgrades.

However, a noteworthy change is occurring with SanDisk. In the past, the market mainly viewed it as a high-elasticity proxy for NAND price increases. Recently, as agreements including minimum procurement, financial guarantees, or long-term cooperation mechanisms have increased, investors have begun to assign higher valuations to its revenue and cash flow visibility.

The commencement of BiCS10 sampling is an important product milestone, but it may not directly create a sustained market trend. What truly needs to be observed is whether the products can pass enterprise customer certifications, enter bulk procurement, and ultimately reflect in average selling prices, data center revenues, and gross margins.

After all, for high-valuation cyclical stocks, technological advancement is just the starting point; order fulfillment is the endpoint.

Western Digital and Seagate also benefit from HDDs, but their revaluation paths differ, and their core logic varies.

Western Digital's revaluation leans more towards financial structure. After completing the spin-off of its flash memory business, WDC has become a purer HDD company. The increase in the proportion of high-capacity products, improvement in pricing per TB, industry supply discipline, and operational leverage collectively drive up gross margins and free cash flow.

Therefore, WDC's stock price may not always fully reflect on the first day after earnings reports. The market needs to confirm that the increase in profit margins is not due to inventory, exchange rates, or one-time factors, but rather from sustained changes in long-term orders, product structure, and industry supply discipline.

Seagate's revaluation, on the other hand, leans more towards technological and capacity value. HAMR can increase the capacity of a single disk without significantly increasing the physical number of hard drives, reduce the unit cost per TB, and enhance the sales value of each hard drive. As products over 40TB enter bulk delivery, the market is no longer just trading HDD shipment growth but also the penetration rate of the new generation of products, scarce capacity, and long-term profit margins.

This is why Seagate's earnings report catalysts often have stronger sustainability. Investors are not only adjusting the current EPS but also the capacity upgrades and profit centers for the coming years.

III. What to Watch for in the Next Market Cycle?

Overall, in the past year and a half, the valuation reconstruction of storage companies has mainly accomplished two things—the market has confirmed that AI storage demand is not a one-time inventory replenishment and has begun to acknowledge that the profit center of storage companies may be higher than traditional cycles.

However, as all six companies have experienced significant increases, the threshold for the next stage has clearly risen.

First, demand must translate into binding long-term orders.

The market will increasingly focus on contract amounts, minimum procurement volumes, execution periods, price adjustment mechanisms, and customer prepayments. Simply announcing strategic cooperation, joint research and development, or long-term partnerships is not enough to enhance future profit forecasts. Only contracts that can lock in quantities, prices, and cash flows can truly reduce cyclical fluctuations and achieve higher valuation multiples.

In this dimension, SK Hynix benefits the most from long-term HBM supply arrangements; Micron can enhance revenue visibility through strategic customer agreements; and SanDisk needs to prove that new business cooperation models can mitigate the impact of NAND spot cycles on profitability.

Second, product prices need to continue to translate into profit margins and free cash flow.

A common misconception in the storage industry is to discuss quotes without discussing profits. The increase in prices for HBM, DRAM, NAND, and HDD can ultimately translate into higher gross margins, depending on product mix, yield rates, depreciation, capital expenditures, and customer agreements.

For SK Hynix, Micron, and Samsung, the core is the average selling price, yield rates, and revenue proportion of HBM4; for SanDisk, the core is NAND contract prices, the proportion of enterprise SSDs, and inventory; for Western Digital and Seagate, it is pricing per TB, penetration rates of high-capacity products, and free cash flow.

What the market ultimately buys is not a price increase story but an upward revision of profit forecasts.

Third, capacity expansion cannot outpace demand fulfillment.

Currently, HBM, advanced DRAM, NAND, and HDD all benefit from supply discipline, but high profits will also stimulate capital expenditures.

If new wafer fabs, packaging capacities, and high-capacity HDD production lines are concentrated and released in 2027-2028, the market will reassess how long supply tightness can last. Therefore, capital expenditure plans, equipment installation progress, capacity ramp-up speeds, and customer booking coverage periods will gradually become more important indicators than single-quarter revenues.

For Samsung, the market is concerned about whether new capacity can accompany share increases; for SK Hynix and Micron, the core is whether capacity expansion still lags behind HBM demand; for Seagate and Western Digital, it is necessary to observe whether high-capacity HDD capacity continues to be pre-booked by customers.

It is worth noting that the end of July to mid-August will also become a new concentrated verification window for the storage sector: Seagate will release its performance on July 28, SK Hynix will announce its quarterly financial report on July 29; SanDisk and Western Digital both plan to release their performance on August 5, and SanDisk will also hold an Investor Day on August 13.

At that time, the questions the market truly needs to answer will definitely become: To what extent have orders for 2027 been discussed? How much more can product prices increase? When will new capacity be released? Can high profits continue to translate into free cash flow? How does management plan to allocate this cash through buybacks, dividends, or capacity expansion?

Final Thoughts

From HBM to server DRAM, from enterprise SSDs to HDDs, this round of market trends has proven that AI's impact on the storage industry goes far beyond just a high-bandwidth memory surrounding GPUs.

As models grow larger, inferences become more frequent, and generated data increases, the entire storage hierarchy will gain new demand.

However, certain industrial demand does not equate to guaranteed stock returns.

When the market is still skeptical about AI storage demand, investors are buying exposure, but when demand has become a consensus, what the market buys is fulfillment—whoever can lock in orders early, raise product prices, control new capacity, and truly translate revenue into profits and free cash flow will continue to receive premiums.

Disclaimer: This content is provided for general branding and informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online events, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets or to use any services. Crypto assets are highly volatile and may result in loss. WEEX services and online events may not be available in all regions and are subject to applicable laws, regulations, and eligibility requirements. You are responsible for ensuring that your use of WEEX services complies with local laws and for carefully assessing the risks before participating in any crypto-related activities.