Micron and Sandisk at 5x Earnings: Bargain or Peak-Cycle Trap?

A 5x P/E in Memory Is Not the Same as a 5x P/E in Software

Micron and Sandisk have reached the kind of valuation multiples that normally make investors stop and look twice. After Thursday’s semiconductor rout, Micron was trading at roughly 4.8 times estimated forward earnings and Sandisk at about 4.9 times, versus roughly 18.5 times for the S&P 500. On the surface, the discount is extreme. It implies an earnings yield of more than 20% for both memory stocks compared with about 5% for the broader index.

But memory stocks have a long history of becoming optically cheapest when earnings are closest to a cyclical peak. The denominator in the P/E ratio can expand much faster than the market is willing to capitalize it. If DRAM and NAND prices later normalize, a stock that looked inexpensive at five times forward earnings can become expensive without its share price moving at all. That is the value-trap argument now being applied to both companies.

What makes the current setup more complicated is that management guidance is not signaling an imminent collapse. Micron expects fiscal first-quarter 2027 revenue of about $61.5 billion and non-GAAP gross margin of roughly 86.25%. Sandisk expects fiscal second-quarter revenue of $12.2 billion to $12.6 billion and non-GAAP gross margin of 84.5% to 85.5%. The market is therefore discounting a downturn that sits beyond the next reported quarter rather than one management can already see.

Figure 1. Forward P/E multiples reported during the Nov. 5 selloff. Source: DaveFinances.com

Micron’s fiscal fourth quarter shows why investors are nervous about extrapolating the present. Revenue reached $54.23 billion, up from $11.32 billion a year earlier. GAAP gross margin expanded to 86.8% from 44.7%, while diluted earnings reached $32.87 per share. For the full fiscal year, revenue rose to $133.19 billion from $37.38 billion, and net income jumped to almost $85 billion.

Sandisk’s numbers are just as extreme. Fiscal first-quarter 2027 revenue rose 378% year over year to $11.03 billion, gross margin reached 85.0%, and diluted EPS came in at $52.60 on a pre-split basis. Datacenter revenue alone reached $4.80 billion, up from $269 million a year earlier, as both sales volumes and revenue per gigabyte increased sharply.

These are not merely good semiconductor results. They are peak-like profitability numbers in businesses that historically suffered when supply caught up with demand. That is precisely why the market can look at record earnings and assign them a lower multiple rather than a higher one. Investors are asking how much of the current margin structure is durable and how much is the product of temporary scarcity.

Figure 2. Year-over-year gross-margin expansion at Micron and Sandisk. Source: DaveFinances.com

Figure 3. Quarterly revenue comparison using company-reported results. Source: DaveFinances.com.

The Selloff Is Pricing Two Risks at Once

The first risk is the familiar memory cycle. When demand runs ahead of supply, pricing rises quickly and incremental revenue drops through at very high margins. When manufacturers add capacity or customers slow purchases, that operating leverage can reverse just as violently. Micron’s move from a 44.7% quarterly gross margin to 86.8% in one year illustrates how fast the cycle has changed. Sandisk’s move from 29.8% to 85.0% is even more dramatic.

The second risk is new: whether political and financing constraints slow the AI data-center buildout that helped create the shortage. The November 5 technology selloff came after Democrats won control of both chambers of Congress and after a series of state and local election results intensified scrutiny of data-center construction. Reuters reported that dozens of ballot initiatives could restrict building and that investors were increasingly concerned about permitting, power-grid pressure and water usage.

That policy risk is real, but it is not the same as saying AI memory demand disappears. Data centers can move between states or countries, approved projects can continue, and the largest hyperscalers still have strategic reasons to expand compute capacity. The immediate market reaction is therefore best understood as a reduction in the assumed growth rate of AI infrastructure, not evidence that the memory cycle has already broken.

Figure 4. One-month drawdowns reported during the Nov. 5 selloff. Source: DaveFinances.com.

Sandisk’s most interesting defense against the value-trap thesis is not its latest EPS figure. It is the structure of its customer contracts. The company disclosed $88.9 billion of transaction price allocated to remaining performance obligations at the end of the October quarter, with roughly 20% expected to be recognized as revenue over the next twelve months. Those obligations are entirely tied to Sandisk’s new business model agreements.

Management has been trying to replace the traditional spot-like memory cycle with longer-duration commercial arrangements that provide better revenue visibility. Sandisk said it now has eight of these agreements. The contracts do not eliminate pricing risk, and only a portion of the remaining obligations will be recognized in the next year, but they make the business less dependent on purely transactional NAND pricing than in prior cycles.

The mix is also changing. Datacenter revenue increased 61% sequentially in the latest quarter and represented more than 43% of total sales. Consumer revenue, by contrast, fell 39% year over year. That means the extraordinary profitability is increasingly tied to enterprise and AI infrastructure rather than a broad consumer electronics boom.

Micron Is Still Guiding Up, Not Down

Micron’s forward guidance makes the five-times-earnings valuation especially unusual. After posting $54.23 billion of fiscal fourth-quarter revenue, the company guided to roughly $61.5 billion in the next quarter, with diluted non-GAAP EPS of about $38.15. HBM4 is already in high-volume shipment for a lead customer, while Micron says it has shipped qualification samples to multiple additional customers and expects HBM4E production in calendar 2027.

That matters because high-bandwidth memory is less interchangeable than commodity DRAM and is closely tied to leading AI accelerators. The tighter qualification process and packaging constraints can support pricing and customer relationships for longer than a conventional PC-memory cycle. Micron is also seeing strong demand across server DRAM and data-center SSD products, so the AI exposure is broader than HBM alone.

The bearish case is that these numbers represent maximum earnings power. The bullish case is that AI has structurally increased memory content per server and tightened the supply chain enough to keep profitability above historical norms. At a forward multiple around five times, investors do not need margins to remain at 86% forever. They do, however, need to believe the eventual normalization is much less severe than the share prices now imply.

The clearest warning would be a change in company guidance rather than another sharp move in the stocks. If Micron begins guiding to lower bit shipments, falling HBM pricing or a significant gross-margin decline, the market’s low multiple would be validated. For Sandisk, investors should watch the pace at which its $88.9 billion order base converts to revenue, whether customers continue signing new long-term agreements, and whether datacenter pricing begins to weaken.

Capital spending across the hyperscalers is the second signal. The election has added uncertainty around new data-center construction, but the relevant question for memory suppliers is whether that uncertainty turns into delayed GPU deployments, lower server shipments and reduced memory orders. Political headlines matter only when they change procurement schedules.

The third risk is supply. Memory companies have destroyed shareholder value in past cycles by adding capacity into strong pricing environments just before demand softened. If Micron, Sandisk and their Asian competitors all respond to today’s shortages with aggressive supply expansion, the current margins could prove temporary even if AI demand continues growing.

The Market Is Pricing a Much Worse 2027 Than Management Is

At roughly five times forward earnings, Micron and Sandisk are not being valued like companies whose current economics persist. They are being valued as if earnings estimates are close to a high-water mark and the next phase of the cycle brings a substantial decline. That is a reasonable concern in memory, but it creates an unusually wide gap between what the companies are reporting today and what the market is discounting tomorrow.

The difference between a bargain and a value trap will not be settled by whether the stocks rebound 10% after a 30% decline. It will be settled by the durability of pricing, the pace of AI data-center deployment and the ability of both companies to maintain margins as supply responds. For now, the operating data still look stronger than the stock prices.

That makes the next two quarters more important than the headline P/E. If Micron and Sandisk continue to guide revenue higher, preserve gross margins and convert AI demand into contracted cash flow, a five-times multiple will look increasingly difficult to justify. If pricing begins to roll over before capacity commitments are absorbed, the market will have been right to treat today’s record earnings as the top of the cycle.

Original Post