Storage profit margin of 80% still faces sell-off: Has the market stopped believing in the storage supercycle?

On November 5, the storage sector presented a striking contrast that is hard to ignore. SK Hynix reported an operating profit of 83.33 trillion Korean won for the third quarter, with an operating profit margin close to 80%. Samsung Electronics' quarterly operating profit exceeded 107 trillion Korean won, with the semiconductor business contributing the vast majority. Both companies are setting record profits, yet their stock prices are moving in the opposite direction of their financial performance, with SK Hynix dropping 9.46% in a single day and Samsung Electronics falling 7.57%, while U.S. storage stocks also faced concentrated sell-offs.

SK Hynix closed at 1.264 million won, down 57.7% from its peak of 2.987 million won on June 25. The market sees the profits but is questioning whether this profit is the starting point of a new long-term growth cycle or just a highlight moment left at the peak of the cycle.

The issue facing storage stocks has shifted; whether profits are high enough is just the first half, while the speed at which profits can grow is the second half. The market is no longer trading on profit levels; the slope of profitability is the main focus.

The core controversy of this market cycle has shifted from whether there will be demand in the future to whether locked-in orders can maintain current prices and how much additional unconfirmed demand can grow.

1. Prices are still rising, but the market is starting to worry about stagnation

The profitability elasticity of the storage industry comes from prices; prices are still rising, but the rate of increase has slowed down.

DRAM and NAND are typical cyclical products; once supply and demand are out of balance, price increases will quickly reflect on the profit statement. When the rate of price increase begins to narrow, even if the quotes are still rising, stock prices may preemptively trade on the peak of profit growth.

Data from TrendForce has already shown this change path: the contract price for conventional DRAM rose by 90% to 98% quarter-on-quarter in the first quarter, dropped to 58% to 63% in the second quarter, and further narrowed to 13% to 18% in the third quarter (as reported). From nearly doubling in a single quarter to growing by just a few percentage points, the speed of price increases has clearly slowed.

Of course, one cannot conclude that the cycle has reversed based solely on this set of numbers. On October 30, TrendForce revised its forecast for the fourth quarter PC DRAM price increase to 13% to 18%, up from the previous forecast of 10% to 15%, driven by suppliers continuing to tilt capacity towards server memory, which may further tighten the supply of ordinary PC DRAM. The industry's judgment circulating in the market is that manufacturers like Samsung and SK Hynix may cut more than 40% of PC DRAM supply next year. The trend of price increases remains, but the support has changed sources, relying on capacity allocation and structural shortages; the phase of synchronized expansion of terminal demand across all categories has passed.

Kioxia's financial report revealed another side of demand to the market. The company's revenue for the July to September quarter was approximately 2.48 trillion yen, with an operating profit of about 1.99 trillion yen, and an operating profit margin exceeding 80%, showing very strong profitability (the company disclosed this on October 30). However, the revenue guidance for the December quarter was only about a 1% quarter-on-quarter increase, with management also mentioning weak demand for mobile phones and PCs, and that end customers have limited capacity to bear higher memory costs. On the day of the announcement, Kioxia's stock price closed down.

This is where the market's real tension lies: upstream manufacturers can continue to push prices higher through production cuts and product structure adjustments, but downstream customers may not be able to accept unlimited increases. The higher the storage prices, the greater the cost pressure on terminal manufacturers of mobile phones, PCs, etc. If the final price increase is passed on to consumers, demand may be further suppressed; if manufacturers choose to absorb the costs themselves, profit margins will be compressed.

The storage industry has entered a more complex stage, with the server side still tight, while the consumer side begins to bear high prices. Suppliers' profits continue to expand, but terminal demand's response to prices has become more important than ever.

2. The midterm elections have added a new variable to AI server demand

The pressure on consumer electronics is already reflected in data and company statements, and the AI server, the strongest leg, has also added a policy variable.

Counterpoint data shows that global smartphone shipments fell 20% year-on-year in the third quarter, marking the lowest level for the same period since 2012 (as reported on October 26). Qualcomm stated in its earnings call on November 4 that memory pricing is compressing the scale of the mobile phone industry, with the company's mobile business revenue down 26% year-on-year for the quarter (as reported).

Apple's statement is even more representative. John Ternus, who took over as CEO not long ago, said in the earnings call on November 2, "What we're seeing with memory pricing right now is unprecedented." The company provided a gross margin guidance of 45% to 46% for the December quarter and pointed out that higher memory costs are a significant reason for the pressure on gross margins (as reported). If Apple needs to raise product prices to cope with rising costs, the pressure on small and medium-sized mobile phone and PC manufacturers will only be greater.

The elasticity of demand on the consumer side is decreasing, making the growth focus of the storage industry more reliant on AI servers. This was originally the most solid support for this super cycle, but after the midterm elections in November, this demand line has encountered new policy variables.

The Democratic Party's victory in the House and Senate means that the legislative agenda for the last two years of the Trump administration will face stronger restrictions, and congressional investigations and oversight may also increase. More importantly for the storage industry is the local political backlash surrounding data center construction that emerged during the election. According to Ballotpedia's statistics as of November 4, of the 47 local votes related to data centers that have been decided, 46 results lean towards restricting data center construction, with varying methods of restriction, including banning new constructions, setting construction moratoriums, increasing public voting procedures, and imposing stricter conditions on projects enjoying tax incentives. Another organization reported that out of 44 related votes on that day, 38 passed restrictions, with different organizations having varying scopes and timelines for their statistics.

Sanders also reiterated the call for a nationwide pause on data center construction after the election results were announced, citing that 43 out of 44 communities support local restrictions. This remains a political initiative and has not entered the congressional review process. Democratic control of Congress does not mean they can directly halt data center projects across the U.S.; Trump still holds presidential veto power, and local planning and project approvals are dispersed among state, city, and relevant regulatory agencies.

For the storage industry, the real issue to focus on has shifted; beyond the signed orders, the costs and timelines for new projects may have changed. If local governments raise approval thresholds, requiring developers to bear more costs for electricity, water, and community support, or tighten tax incentives, project financing and construction timelines may be extended. If data center projects are delayed, server procurement, network equipment deployment, and associated DRAM and SSD orders may also be postponed.

The impact on the storage market has a clear time lag. Long-term supply agreements already signed will not automatically disappear due to an election result; however, new data center projects planned for 2027 and beyond may need to reassess investment returns and construction progress. Current signed orders will not suddenly drop to zero due to an election, but the market is concerned that incremental demand in the coming quarters may not be as certain as previously expected.

This also explains why long-term demand remains strong while stock prices have started to decline. Goldman Sachs' industry meeting minutes remain optimistic about HBM demand, believing that the supply gap may continue until 2028 (as reported on October 28). The capital market has begun to break down AI infrastructure investment into two issues: whether confirmed orders can be fulfilled and whether the next round of new capital expenditures can be implemented as planned. The former supports current profits, while the latter determines whether valuations can continue to expand.

3. A 3.5x PE looks cheap, but it may be the most expensive time for cyclical stocks

The calculation of being cheap is not wrong; the mistake is using profits at the cycle peak as the denominator.

J.P. Morgan's report on November 4 laid out the current valuation contradictions of storage stocks. The reported data shows that major storage stocks have retreated an average of about 45% from their 2026 peaks, while the Philadelphia Semiconductor Index has retreated about 29% during the same period. The expected price-to-earnings ratio for 2027 is about 3.5 times, while another media report cites it as about 3.6 times, with both figures being presented. Meanwhile, the market's expectation for the compound annual growth rate of EPS from 2026 to 2028 remains as high as 98% (as reported).

A 3.5x PE paired with nearly doubling profit growth looks very cheap at first glance. The denominator of the price-to-earnings ratio is profit; when storage prices rise rapidly, capacity utilization is high, and product structure continuously shifts towards HBM and server memory, the company's profits will be rapidly boosted. At this time, even if stock prices have risen significantly, as long as profit growth is faster, the PE may still remain at a low level. This represents another matter: the market is willing to believe that current profits can be sustained for years, which is what makes it seem cheap.

If storage prices enter a plateau or consumer demand shrinks further due to rising costs, profit growth may change before revenue does. Assuming profits fall from the cyclical peak, what originally looked like a 3.5x PE will quickly become not cheap as the denominator shrinks. What the market really needs to judge is whether these companies can maintain profit levels close to today's in two years; this year's performance is merely a ticket to entry.

This is also why stock prices may decline ahead of record earnings reports. Stock prices discount the future, while already published quarterly results are merely past performance. The compression of storage stock valuations essentially reflects the market's discount on the sustainability of profits.

However, shorts cannot ignore the supply-side constraints either. SK Hynix officially disclosed that over 50% of its capacity for 2027 to 2028 has already been signed under long-term supply agreements, and the company also expects memory shortages to continue until 2030 (the company disclosed this on October 27). The company emphasized that the agreements will reflect market conditions and help stabilize future profits. Samsung is also continuously expanding the supply of HBM4, server DRAM, and enterprise-grade SSDs. Demand for AI servers remains strong, and the expansion of advanced processes, packaging, and testing capabilities takes time; supply cannot rapidly catch up in the short term.

Both bulls and bears do not deny the existence of AI demand; the contention lies in how long high profits can be sustained. Bulls bet that the structural shortage of HBM and server memory will continue until 2028, believing that long-term contracts and capacity restrictions can protect profits. Bears are concerned that the phase of rapid price increases has already passed, and that terminal cost pressures, data center project approvals, and power restrictions will gradually transmit in the coming quarters. Micron's forward P/E ratio is also less than 5 times, with CEO Sanjay Mehrotra stating that supply will only gradually improve by 2028 (reported on November 3).

The current market gives a P/E of 3.5 times, reflecting another issue: capital is unwilling to pay today's profit levels for the high valuations of the coming years.

IV. Capacity has been booked for years, but profit margins cannot be locked in

Capacity locking addresses who to sell to, long-term contracts address how much to sell, but what truly determines valuation is still the selling price and profit margin.

If the capacity for the next few years has already been locked in by customers, what exactly is the market still worried about? Here, three concepts must be distinguished:

SK Hynix officially disclosed that more than half of its capacity for 2027 to 2028 has been signed under long-term agreements, which is a strong fundamental signal. Some media summarized the management's statements as a substantial sellout of supply over the next three years, which is a paraphrase; the hard numbers in the announcement indicate that over half of the capacity has been signed under long-term contracts. The specific binding force of long-term contracts depends on the contract terms, such as whether there is a minimum purchase volume, how prices are adjusted, whether customers can defer or cancel, whether there are advance payments, and how high the breach costs are. Guaranteed shipment volumes do not mean that every chip can maintain the current selling price and profit margin.

This is also where the policy risks of the midterm elections need to be reinterpreted. Local governments in the U.S. are raising the approval thresholds for data center projects and increasing requirements for power and water supply, which will not automatically cancel the storage long-term contracts already signed by cloud vendors in the short term. What is truly affected are the data center projects that have not yet materialized and the next round of new server purchases. If projects are delayed, cloud vendors may first adjust their new capital expenditures, which will then affect the renewal and expansion of future long-term contracts. Policy shocks are more likely to impact incremental demand, while directly destroying already locked orders is relatively rare.

There is also a time lag on the supply side. SK Hynix, Samsung, and Micron are still expanding HBM and server memory supply, but the release of capacity for advanced products takes time. Meanwhile, the ordinary DRAM and NAND markets may also be affected by domestic manufacturers' capacity expansions. Changxin Memory's fifth-generation DRAM process has already entered mass production, with capacity expected to double by 2028 (according to TechInsights, reported on October 5 and 19), and Yangtze Memory is also continuously promoting NAND product upgrades. The supply and demand cycles for different products are not completely synchronized; just because HBM is still in short supply does not mean the entire storage industry can maintain current profit levels.

V. To judge whether the storage cycle has truly peaked, three signals are more important:

First, observe whether contract prices shift from narrowing increases to actual declines. The fourth quarter PC DRAM price increase forecast has been revised upward, indicating that supply tightness still exists. However, it is necessary to distinguish between server DRAM, PC DRAM, HBM, and NAND; one cannot directly equate the price quotes of one category with the entire storage market. What truly needs to be monitored is whether price increases slow down and then turn into declines, while channel inventory increases and customer purchasing willingness weakens.

Second, observe whether capital expenditures for AI data centers are affected by policies and power restrictions. After the midterm elections, local restrictions on data center projects have transformed from campaign slogans into actual cases of bans, suspensions, public votes, and tax incentive conditions. Next, it is important to watch whether large cloud vendors adjust their capital expenditure guidance, whether project approvals and power access cycles are extended, and whether the construction plans for new data centers are delayed. Only when these changes transmit to server purchases will they truly impact storage vendors' orders.

Third, observe whether the coverage of long-term contracts and actual profit margins can be maintained simultaneously. Long-term agreements can improve capacity utilization and shipment certainty; if the adjustment range of contract prices decreases, or if the supply of high-profit products gradually increases, profit margins may still decline. The upcoming financial reports from SK Hynix and Samsung need to simultaneously observe the product structure, average selling price, gross margin, and capital expenditures of HBM, server DRAM, and enterprise SSDs; merely looking at whether revenue and operating profit reach new highs is not sufficient.

The storage industry is still in a structural opportunity brought about by the expansion of AI infrastructure, and the demand for HBM and server memory has not disappeared due to an election cycle. The market is no longer willing to factor all growth expectations into stock prices in advance; whether the consumer end can bear high prices, whether data center projects can be implemented on time, and whether suppliers can maintain current profit margins—these questions are beginning to become more important than the next record-breaking financial report.

An 80% operating profit margin is the result already realized in this super cycle. A 3.5 times expected P/E reflects the market's skepticism about the sustainability of future profits. The wall that storage stocks have hit is built from pricing, and the height of the wall does not depend on the absolute level of profits, but on how much the market is willing to pay for years that have not yet been realized.

Disclaimer: The content of this article is for general information and market commentary purposes only, compiled based on publicly available information as of the time mentioned in the text. Relevant market data, expectations, and probabilities may change with market conditions. The views and investment strategies of third-party institutions, analysts, or other individuals cited in the text only represent the views of the relevant third parties at a specific time and do not represent the views or recommendations of BIT. This article does not constitute investment advice, investment research, solicitation, or recommendation for any securities, investment products, or trading strategies, nor should it be used as the basis for any investment decision. Financial markets carry risks, and securities prices and market performance may fluctuate; past performance and historical market trends do not represent or guarantee future results. Investors should independently assess the relevant risks based on their own circumstances and seek professional advice when necessary. BIT's Hong Kong stock trading services are not provided to residents of Hong Kong or users in other restricted jurisdictions.