股价创历史新高，但存储类股的估值仍在低档

Original author: Jialiu

Today, Micron delivered a historic earnings report that has boosted confidence across the semiconductor sector.

FY2026 Q3 revenue reached $41.46 billion, nearly $6 billion above market expectations. A memory company, long labeled as a "low-margin commodity," posted a gross margin guidance typically seen in software companies. Its stock price surged 13% to 14% in after-hours trading, pushing its market cap to $1.16 trillion.

Micron's gains this year have already been remarkable. Closing at $1,211.38 on June 22, it has more than tripled year-to-date and surged over 850% in the past 12 months, making it the third-best performing stock in the S&P 500 in 2026, behind only SanDisk and Western Digital—both also memory companies. The entire sector is moving at this magnitude. SK Hynix has risen over 800% in the past 52 weeks, and Samsung has surged over 400%.

Seeing such gains, many people's first reaction is naturally, "It's too expensive." But in reality, a high stock price doesn't necessarily mean expensive valuations. From many perspectives, memory remains a very "cheap" hot sector.

Stock Price Up 9x, But PE Has Barely Moved

Judging whether a company's stock is expensive or not, one of the most common indicators is PE, or price-to-earnings ratio.

Simply put, PE measures how much the market is willing to pay for every dollar of profit a company earns. A PE of 10 means investors are willing to pay $10 for $1 of annual profit. A high PE usually indicates strong future growth expectations; a low PE might suggest the stock is cheap, or it could mean the market believes the company's current profits are just a cyclical peak that will soon decline.

This is where memory stocks are most counterintuitive right now: stock prices have risen significantly, but PEs remain very low.

According to FactSet data, a MarketWatch report in mid-June provided these figures: Micron's forward 12-month PE is around 9x, SK Hynix and Samsung around 6.5x. Barron's believes Micron's forward PE is about 9.74x, compared to the Nasdaq Composite Index around 25.5x and the S&P 500 around 20.3x. Data from GuruFocus on June 21 shows Micron's forward PE at 9.90x, SK Hynix at 5.92x (end of May), and Samsung at about 5.45x.

This means most data sources estimate that the forward PEs of the three memory giants are all in the single digits or just above 10x.

Within the entire AI industry chain, these figures are among the lowest tier.

Nvidia's forward PE is around 23x, Broadcom around 30x, AMD around 25x, TSMC around 20x, and the median for the semiconductor industry overall is around 36x. In other words, the valuation level of the three memory giants is roughly one-third of Nvidia's and about one-quarter of the semiconductor industry median.

But ironically, an increasing share of AI industry profits is being captured by the memory segment.

An AI server isn't just a GPU. Every high-end AI accelerator card needs HBM, every inference server needs large-capacity DRAM, and KV cache, model weights, local cache, and data throughput all rely on SSDs. Without HBM, there is no GPU training cluster; without server DRAM, there is no inference cluster; without high-capacity NAND, the storage and caching costs for AI applications can't be reduced.

Memory is no longer just a standard component in the AI industry chain; it has become a physical bottleneck that all AI capital expenditures must navigate. A figure from Micron's latest earnings report illustrates this: single-quarter data center revenue was $25 billion, with enterprise SSD revenue at $5 billion, accounting for 20% of data center revenue.

It's evident that this bottleneck is now even starting to impact consumer electronics.

AI data centers have driven up the capacity and prices of HBM, DRAM, and NAND, ultimately forcing even a company with strong bargaining power like Apple to face cost pressures and pass some of the price increases onto consumers. In the past, discussions about AI profitability immediately focused on Nvidia. But it's becoming increasingly clear that a large portion of the AI bill is flowing to memory manufacturers.

Memory stock prices have risen significantly, but profits have risen even faster.

Micron's just-announced Q3 EPS was $25.11, compared to $1.91 in the same period last year, a more than tenfold increase year-over-year. SK Hynix's Q1 2026 operating profit was 37.61 trillion won, up 405% year-over-year. Samsung's semiconductor division saw operating profit in Q1 surge more than eightfold year-over-year. Stock prices rose several times, but profits rose even more, preventing PEs from being pulled up.

The flow of AI money into the income statements of memory manufacturers is very real and substantial.

Sector Releases Positive News in a Wave, Micron is Just the First Shot

Micron's earnings report is the starting gun for this round of memory earnings season.

Next, the memory sector enters a month of high information density: TSMC on July 16, Samsung on July 23, and SK Hynix and Western Digital on July 29.

And the strength of Micron's report has already set the tone for the others. Its most critical information wasn't just the single-quarter beat, but the Q4 guidance of $50 billion in revenue and an 86% gross margin.

This guidance essentially tells the market that price increases haven't peaked but are accelerating. The subsequent four companies will, in different markets and with different product mixes, either validate or refute the same trend revealed by Micron's guidance.

Let's look at TSMC, reporting on July 16.

TSMC doesn't manufacture memory, but it is the foundation of the entire AI chip supply chain. Nvidia's GPUs, Broadcom's custom accelerators, and AMD's data center chips all come from its production lines. The question TSMC answers is more fundamental: has the capacity bottleneck for AI chips been resolved? Q1 revenue was $35.9 billion, up 40.6% year-over-year, with a gross margin of 66.2%, and advanced process nodes accounting for 74% of wafer revenue. Q2 guidance is for revenue between $39 billion and $40.2 billion.

There is a multiplier relationship between TSMC and memory. Every additional advanced process wafer TSMC sells results in more AI accelerators downstream, and each additional accelerator requires several HBM stacks. The HBM capacity paired with a single GPU on Nvidia's Vera Rubin platform is multiple times that of the previous generation. The more TSMC ships, the tighter memory capacity becomes.

July 23 is Samsung's earnings report.

Fifteen brokerages expect Samsung's Q2 operating profit to be around 88.3 trillion won, with an operating profit margin flat or even higher than Q1's 66%. For a conglomerate that also makes mobile phones, panels, and home appliances, its profit margin has been pulled to this level solely by the memory division.

But the most important aspect of Samsung's earnings isn't the profit number, but HBM4. Samsung holds only about 17% of the HBM market, far behind SK Hynix's 62% and Micron's 21%. The generational shift to HBM4 represents Samsung's only window to close the gap. During its Q1 conference call, it made several specific statements: HBM sales in 2026 are expected to more than triple year-over-year, and from Q3 onwards, HBM4 will account for over 50% of HBM sales. Micron just disclosed that its HBM4 36GB 12-Hi has begun mass production and shipments. The competitive positioning among the three in HBM4 will be the key battle to watch in the second half of the year.

On July 29, SK Hynix and Western Digital report earnings on the same day.

SK Hynix's Q1 was textbook: quarterly revenue of 52.6 trillion won, up 198% year-over-year, operating margin of 72%, and net profit margin of 77%. A hardware manufacturer achieving a 77% net profit margin is remarkable—Apple's is around 25%, Nvidia's about 58%. For Q2, some brokerages predict the operating margin could approach 80%. Micron has already taken the lead with an operating margin of 81.2%, surpassing TSMC. As the leader in HBM market share, SK Hynix is highly likely to deliver a Q2 report comparable to Micron's. The combined operating profit of Samsung and SK Hynix for Q2 is expected to exceed 150 trillion won. Adding Micron, the total single-quarter profit for the three giants will set a new record.

Western Digital reports its Q4 on the same day, without DRAM or HBM, purely NAND and SSD. It provides another dimension of AI storage demand: the KV cache for inference requires high-capacity SSDs. Q3 Cloud revenue grew 48% year-over-year, with a record gross margin of 50.5%. It's worth noting that Western Digital and its spun-off SanDisk are the two best-performing stocks in the S&P 500 in 2026, ahead of Micron. While NAND line growth hasn't been as explosive as DRAM's, the direction is perfectly aligned.

AI Has Transformed Memory from a Commodity to a Luxury

New stock price highs, yet low PEs. Earnings reports are increasingly astonishing.

Despite all this, some may still wonder if this is sustainable or just another cycle of euphoria destined to crash.

We can look at the analysis from Jukan, a semiconductor analyst at Citrini Research.

As early as Q1 2024, when SK Hynix and Micron were still struggling with post-pandemic DRAM inventory gluts and low stock prices, the Citrini team called for them to outperform. These stocks subsequently increased several times to nearly tenfold. They have been almost perfectly right about this memory cycle. A detail from early June this year illustrates his position in the market: he retweeted a SemiAnalysis report on adjusting the memory configuration for Nvidia's Rubin server, which immediately put visible pressure on Micron and SK Hynix's stock prices.

Jukan's core argument for being bullish on memory isn't a short-term judgment like "prices will rise." Instead, he believes: AI has transformed memory from a commodity into a luxury.

First, HBM has broken a sixty-year trend. From 1957 to 2020, the cost per Gb of DRAM decreased by roughly an order of magnitude every five years; prices were always falling. This is a fundamental law of the memory industry, and the entire industry's competitive model and valuation framework were built on this trend. Jukan points out that the AI-driven demand for HBM has completely shattered this law. Manufacturers are shifting capacity towards the more complex, silicon-intensive HBM, squeezing traditional DRAM supply.

Currently, no manufacturer plans to convert HBM production lines back to traditional DRAM. The reason is simple: HBM profit margins far exceed those of regular DRAM. Rational manufacturers wouldn't swap high-margin lines for low-margin products. This has turned supply tightness from a cyclical phenomenon into a structural condition that won't reverse as long as AI demand persists.

Therefore, the sustained price increase for HBM storage will be long-term.

Annual HBM volume and pricing are largely negotiated at the beginning of the year, providing manufacturers with strong earnings visibility. TrendForce data confirms this: In Q1 2026, the contract price for traditional DRAM rose 90% to 95% quarter-over-quarter, the largest single-quarter increase on record, with further increases expected in Q2. In a normal DRAM cycle, the price increase phase typically peaks after 4 to 6 quarters. This cycle has been rising for nearly 8 quarters without stopping. JPMorgan even predicts DRAM prices could rise for four consecutive years, something unprecedented in the industry's history.

So, it's almost fair to say memory has transformed from a commodity to a luxury.

The biggest difference between luxury goods and commodities lies in pricing. Commodity prices are determined by marginal cost; anyone can expand production, and profits are eventually competed away, hence a low valuation. Luxury prices are determined by scarcity and pricing power; supply is controlled, and profits can be maintained at high levels long-term, hence a premium. The old rule that "low PE equals peak" presupposes that earnings will revert to that long-term declining trend line. But if the trend line itself has turned, where the reversion happens becomes an open question.

Let's return to the contradiction at the beginning. Stock prices are at all-time highs, but valuations are at historical lows. This anomaly exists because the market is still using the old framework for commodities to price an industry that has become a luxury. Micron, with its 84.9% gross margin earnings report and 86% gross margin guidance, just dealt a heavy blow to that old framework. If a return to the old framework is impossible, then the current 5 to 10x PE is wrong.

Therefore, we believe that despite reaching new highs, memory stocks are still not expensive.

Beyond HBM, Is NAND the Real Main Course?

Every major bull run reaches a midpoint where the market asks the same question: After the leaders have run, who's next to take the baton?

HBM and DRAM have been the absolute protagonists of this memory cycle, with the massive surges of the three giants attributed to them. NAND has been consistently treated as a supporting role.

However, a closer look at the supply-demand structure reveals a counterintuitive fact: NAND, long seen as a garnish, might itself be the main course, with its shortage being, in some ways, even more severe than HBM's.

First, why is HBM so hot? HBM is standard on AI accelerator cards, with high unit prices, fat margins, and high technical barriers. SK Hynix achieved a 62% market share and 77% net profit margin because of it. These are facts. But HBM has one characteristic: although supply is tight, the path to capacity expansion is clear. The three giants are all pouring capital into expanding HBM production, with Samsung and Micron chasing SK Hynix across generations from HBM4 to HBM4E and beyond. Supply is increasing at a visible rate, just temporarily unable to keep up with demand.

In contrast, NAND manufacturers haven't significantly expanded capacity for several years.

This is because the NAND price crash in 2022-2023 scared all players. Capital expenditures on NAND by Kioxia, Western Digital, Samsung, and SK Hynix were slashed to extremely low levels, and new production lines have been repeatedly delayed, with the earliest expected around 2027.

The three giants are prioritizing wafer capacity and capital expenditure for HBM and high-end DRAM, leaving fewer resources for NAND. Micron even directly shut down its consumer-grade Crucial business to free up all capacity for enterprise and GPU-grade storage.

Despite the supply shortage for NAND, demand is immense.

Large model inference requires massive amounts of KV cache and data throughput, directly fueling explosive demand for enterprise SSDs (eSSDs). In Q1 2026, global eSSD revenue grew 86% quarter-over-quarter. Another factor is the HDD shortage. The supply of hard disk drives is also tight, forcing data centers to substitute HDDs with high-capacity SSDs, transferring a portion of HDD demand to NAND.

The CEO of Phison Electronics stated, "Every NAND manufacturer tells us that their entire 2026 capacity is sold out." Kioxia also confirmed that its entire NAND capacity for 2026 is fully booked. The price of a 1Tb TLC NAND chip rose from around $4.8 in July 2025 to around $10.7 by the end of 2025, more than doubling in a few months.

HBM faces a shortage, but supply is increasing deterministically; NAND faces a shortage, but supply-side has almost no incremental growth. The HBM shortage has a cure, just slow-acting; the NAND shortage is temporarily incurable because no one is making the medicine. From this perspective, the supply-demand gap for NAND is more rigid than HBM's, and its pricing sustainability could be stronger.

This is why the two best-performing stocks in the S&P 500 in 2026 are not HBM leader SK Hynix or Micron, but pure NAND and SSD players, Western Digital and its spun-off SanDisk. The market has already voted with its feet, quietly moving NAND from the supporting role to the lead, even if most haven't noticed