Micron Earnings Preview: What Are the Key Points to Watch?

The most important event in the AI industry this week may well be Micron's earnings report. Everyone will be closely watching how the market moves following Micron's results. From a personal perspective, this Micron earnings report is not simply about verifying "whether it's strong," but rather about seeing how long that strength can last. There are several points worth discussing:

1. Current quarter performance must not only beat the company's guidance but also reach the upper end of the market's most optimistic expectations;

During the last earnings call, Micron provided guidance of $33.5 billion for this quarter, but the market's current consensus revenue expectation has already been raised to around $34.8-$35.5 billion, with EPS in the $19.8-$20.7 range. Simply beating the company's guidance is far from enough; it must also beat market expectations.

The strongest beat would be significantly exceeding market expectations, with a gross margin of 83%, and management clearly breaking down the sources of revenue:

1) How much comes from ASP (Average Selling Price)

Last quarter, Micron said DRAM prices rose 60% QoQ and NAND prices rose 70%, which is the contribution from ASP increases. ASP increases are usually the most cyclical part (price hikes due to tight supply and demand), but also the easiest to reverse;

2) How much comes from bit shipments

The actual storage capacity sold (calculated in "bits"). The core purpose is to let the market clearly see how much performance growth comes from unit price increases and how much from shipment volume growth;

3) How much comes from product mix

Changes in the proportion of high-margin versus low-margin products in total sales. Even if total bit shipments remain unchanged, shifting more product sales to high-priced, high-margin markets (such as HBM, high-end DDR5, enterprise SSDs) will increase overall revenue and gross margin;

4) How much comes from cost reductions

Even if prices and sales volumes remain unchanged, lower costs will automatically improve gross margin. This is one of the "hardest" sources of profit because it results from the company's own efforts and is not easily replicated by competitors immediately.

2. Look not only at current quarter performance, but also at next quarter's guidance

The current quarter is in the past; visibility into the next quarter and 2027 is what truly determines whether sell-side models will continue to be revised upward. Goldman Sachs' current FYQ4 model is very aggressive (revenue around $48.8 billion, gross margin 86.1%, EPS $29.95), while market consensus is roughly around $40.4 billion in revenue and $23.68 EPS.

If next quarter's guidance can reach $4.5-$4.9 billion in revenue and an EPS in the $25-$30 range, high-end sell-side models will have continued support, and the wave of target price revisions upward will likely persist.

If it only reaches around $4.0-$4.2 billion with EPS near $23-$24, the numbers are still strong, but for the current stock price, it might just be "realization" rather than a further catalyst.

If it falls below high-end expectations, and management simultaneously avoids discussing 2027 supply and demand, short-term pullback pressure will increase significantly.

3. "SCA/LTA Long-Term Agreement Mechanism": This is the core of valuation enhancement

Over the past two months, the market has begun to view memory stocks as growth stocks rather than cyclical stocks, with the core reason being long-term agreements. The biggest problem with memory stocks in the past was not that they couldn't make money, but that the market didn't believe high profits were sustainable. We previously discussed the long-term agreement terms disclosed during SanDisk's Q1 earnings. Micron also previously announced it had reached multiple Strategic Customer Agreements (SCAs). If SCAs can bind future supply, customer commitments, pricing mechanisms, and capex returns together, MU's valuation multiple could shift from a "cyclical peak discount" to a "smaller AI infrastructure asset discount."

Last quarter, Micron already confirmed several points:

First, SCAs differ from traditional LTAs, which are typically around one-year terms;
Second, SCAs are multi-year agreements;
Third, they include specific commitments from customers;
Fourth, the goal is to improve business visibility and stability;
Fifth, the company has already signed its first five-year SCA.

But the company also explicitly stated that due to confidentiality reasons, it does not disclose details such as pricing, cancellation terms, or downside protection.

So during this call, we need to hear if management has progressed further than last quarter. The most important thing is not "reiterating robust terms," but whether the following information is provided:

1) Coverage: Does the SCA cover HBM, DDR5, LP DRAM, NAND/eSSD, or only certain AI data center products?

2) Commitment Strength: Is it take-or-pay, capacity reservation, prepayment, minimum purchase volume, or a softer forecast?

3) Pricing Mechanism: Is there a price floor, formula pricing, cost-plus-return, or ROIC-linked pricing?

4) Cancellation Cost: If customer demand slows in 2027-2028, does canceling/postponing orders require compensation?

5) Capex Binding: Is Micron's new capacity expansion premised on customer commitments, rather than proactively leveraging up to expand at price peaks?

If management can only repeat "we have multi-year agreements with robust terms" but cannot convince the market that downside gross margins are protected, then the contribution of SCAs to valuation multiples will be discounted. If management can convince investors that a portion of 2027 profits has already been contractualized, it will clearly support a higher target price.

4. "HBM4/HBM4E Roadmap"

Micron has officially announced that HBM4 36GB 12-Hi has entered high-volume production, targeting NVIDIA Vera Rubin, with bandwidth exceeding 2.8TB/s and over 20% better energy efficiency compared to HBM3E; simultaneously, it has already sampled HBM4 48GB 16-Hi to customers, with a single stack capacity 33% higher than the 36GB 12-Hi. During the last quarter's call, the company also stated that HBM4E development is progressing, with volume ramp expected in 2027, and will utilize 1-gamma DRAM.

But a counter-intuitive point needs to be added here: Micron's strongest gross margin currently may not come from HBM, but from non-HBM DRAM/DDR5. TrendForce mentioned that Micron's management stated last quarter that "non-HBM margins are currently higher than HBM," meaning the current 81%+ gross margin largely comes from extreme tightness in traditional DRAM, rather than HBM independently pulling up the gross margin.

So when listening about HBM on the call, don't just listen for "shipments," "qualification," or "roadmap," but also listen for:

Whether HBM4 yields mature faster than HBM3E;

Whether HBM4/HBM4E can maintain or increase the company's share in HBM;

Whether HBM4E's custom base die/TSMC collaboration brings stronger customer lock-in;

After the HBM mix increases, whether the overall gross margin continues to rise, stays flat, or is diluted by advanced packaging costs;

Whether Micron is the core supplier, or a third-source supplier filling gaps, for NVIDIA Vera Rubin, Rubin Ultra, and other XPU customers.

Competition cannot be ignored either. Samsung has already started sampling HBM4E to customers, and SK Hynix has also sampled 12-layer HBM4E to major customers, with speeds up to 16Gbps/pin and over 20% energy efficiency improvement. This is not necessarily negative for Micron, but it means HBM4E is not a one-sided story; platform qualification and share stability are very important.

5. "NAND/eSSD Data Center Business"

This segment may actually be underestimated by the market and is an important part that differentiates Micron from a "pure HBM story." Last quarter, management was very clear: AI inference, vector databases, and KV cache offloading are driving data center NAND bit demand; data center SSD market share has increased for the fourth consecutive calendar year; in FY2Q26, data center NAND revenue more than doubled QoQ, hitting a significant new high, and is expected to continue growing next quarter. Meanwhile, Micron has already mass-produced PCIe Gen6 data center SSDs; the Micron 9650, targeting AI training and inference, supports up to 28GB/s sequential read and 5.5 million random read IOPS.

The significance of this line is threefold:

First, it makes Micron not just an HBM supplier, but a combined AI memory + storage supplier.
Second, it improves the quality of the NAND cycle, as data center SSDs are in a higher-value pool than consumer NAND.
Third, it can explain why NAND prices are also strong, rather than simply "DRAM rises, NAND follows."

The call needs to examine:

Whether data center NAND revenue again shows significant QoQ growth; adoption of 122TB/245TB high-capacity SSDs;
Customer validation and ramp-up pace for PCIe Gen6 SSDs;
Profitability of QLC in the data center capacity tier;
And whether NAND supply is also entering a long-term tight balance.

The current "AI memory" narrative is not a single HBM story. If the call only talks about HBM, the story is incomplete; if it can clearly articulate the joint growth of AI servers, traditional server refresh, and CPU memory, LPDRAM, and eSSD driven by agentic AI, the valuation logic will be more solid.

Additional Points to Watch

1) Capex discipline must be viewed together with SCAs

An important reason the stock price was pressured after hours last quarter was that strong performance was partially offset by a significant increase in capex. At that time, Micron raised its FY2026 capital expenditure to over $25 billion and expected further increases in 2027, causing market concern that this implies future supply normalization and the re-emergence of cyclicality.

The earnings call needs to ask:

How much of the new capex is for shell/cleanroom, and how much is for equipment?

When will this capacity form effective bit supply?

Is it HBM-related, DRAM-related, NAND-related, or general expansion?

Is it supported by customer SCAs or prepayments?

What is the target ROIC?

If 2027 demand is lower than expected, can capex be postponed?

If capex is described as "long-term investment based on customer commitments," this is positive for valuation; if it resembles expansion at a traditional cycle peak, it is a source of valuation discount.

2) Gross margin isn't simpler the higher it goes; beyond 80%, the focus shifts to "sustainability"

The company's FYQ3 gross margin guidance is about 81%, already far above Micron's historical norm. Last quarter, management also cautioned that at this gross margin level, the marginal improvement from further price increases will diminish.

So the market won't just reward "gross margin being 2 points higher," but will ask:

Can the gross margin remain above 70% long-term into 2027? MarketWatch also mentioned that investors are focused on whether the long-term gross margin can exceed 70%, which would be seen as a strong positive signal.

The key is not whether Q3 gross margin is 82% or 83%, but whether management is willing to hint that even if prices fall, SCAs, mix, cost curves, and the supply-demand structure can keep gross margin at historically high levels.

3) Strong demand does not mean there is no "demand destruction"

Rapid price increases can cause PC, mobile, and consumer electronics customers to cut configurations or delay purchases. Last quarter, management already acknowledged that in price-sensitive markets, high prices could affect some demand, but overall demand remained strong; they also stated that in the medium term, the company could only meet 50% to two-thirds of customer demand.

This time, we need to see:

Are PC/mobile devices starting to reduce memory configurations? Are OEMs postponing purchases? Are automotive/industrial customers accepting price hikes? Are customers pulling in orders early? If significant demand destruction appears, the market will worry that 2027 is the peak.

Currently, MU is already trading near $1,134, with a market cap approaching $1.3 trillion. The market has already priced in a significant portion of optimistic expectations ahead of the earnings report. It is very easy for volatility to occur if the guidance slightly falls short of the market's most optimistic expectations, or if certain business discussions lack sufficient detail, leaving the market without enough expected information and data. However, from a medium- to long-term perspective, it is still necessary to delve into the most fundamental logic of the business to see what the growth drivers are. Where does the performance come from? How sustainable and certain is future performance growth? This is the most critical aspect.