Micron is currently around $1,096, up roughly 5% today, giving the company a market capitalization of about $1.26 trillion. That price matters enormously to this analysis: Micron's business fundamentals have arguably never been stronger, but MU has also undergone one of the largest semiconductor reratings of the AI cycle.The central question is no longer whether AI is benefiting Micron—it clearly is. The question is whether today's extraordinary memory pricing, ~80%+ gross margins and earnings power represent a structurally new Micron or an exceptionally strong peak in the memory cycle.Micron reports fiscal Q4 earnings on September 30, making that report the next major test of the thesis.

The Micron thesis can be summarized in one sentence:

AI is transforming memory from a relatively commoditized semiconductor product into a strategically important bottleneck, potentially raising Micron's long-term growth, margins and valuation.

Historically, Micron was a classic cyclical semiconductor company.

The cycle looked like:

Demand rises → memory shortages → prices rise → manufacturers add capacity → oversupply → prices collapse → Micron earnings collapse.

AI may be changing that structure.

Modern AI accelerators require enormous quantities of extremely fast memory. HBM—High Bandwidth Memory—has therefore become almost as important to an AI system as the GPU itself.

And only a handful of companies can manufacture advanced DRAM/HBM at scale: SK Hynix,  Samsung, Micron

That oligopolistic structure is extremely important.

Micron's latest results demonstrate how powerful the current environment has become:

Q3 FY26 revenue: $41.46B
Q3 FY25 revenue: $9.30B

That's approximately +346% YoY.

Non-GAAP EPS reached $25.11, while operating cash flow reached $25.39B.

Even more strikingly, management guided Q4 to approximately:

$50B revenue
86% gross margin
$31 non-GAAP EPS

That isn't normal memory-cycle profitability.

The structural bull case begins with the amount of memory required by AI. Traditional CPUs rely heavily on conventional DRAM. AI accelerators require vastly greater memory bandwidth, which is where HBM comes in. HBM stacks multiple DRAM dies vertically and places the memory extremely close to the processor, providing enormous bandwidth.

Think:

Nvidia GPU → computation

Micron/SK Hynix/Samsung HBM → feeds data to the GPU

Without sufficient memory bandwidth, extremely powerful accelerators can't operate efficiently.

It isn't just HBM.

AI infrastructure requires:

HBM — accelerator memory

DDR5 DRAM — server system memory

Enterprise SSD/NAND — AI data storage

High-capacity RDIMMs — CPU/server memory

Micron therefore participates in several layers of the AI infrastructure stack.

Micron manufactures memory and storage products across four major business units.

Latest quarterly revenue:

The transformation becomes obvious here.

This includes HBM and other memory used by cloud/AI infrastructure.

Revenue:

$13.77B

versus

$3.39B one year earlier.

That's roughly 4×.

Revenue:

$11.52B

versus

$1.53B a year earlier.

That's approximately 7.5×.

Together, Cloud + Core Data Center produced:

in quarterly revenue.

That's approximately 61% of total Micron revenue.

AI/data-center demand has therefore become the center of Micron's earnings story.

The financial acceleration has been extraordinary.

Q3 FY25:

$9.30B

Q2 FY26:

$23.86B

Q3 FY26:

$41.46B

Q4 guidance:

That implies approximately another 21% sequential increase if management hits the midpoint.

Q3 GAAP EPS:

$24.67

Non-GAAP:

$25.11

Q4 guidance:

Annualizing $31 would theoretically produce:

$124 EPS

But that's not how I would value MU because earnings are still accelerating and memory remains cyclical.

The more important question is what Micron earns through an entire memory cycle.

This is one of the most remarkable aspects of the current Micron cycle.

Cloud Memory:

83%

Core Data Center:

87%

Mobile & Client:

87%

Automotive & Embedded:

79%

And Micron guided overall Q4 gross margin to:

That deserves emphasis.

An 86% gross margin from a company historically treated as a cyclical memory manufacturer is extraordinary.

That's software-like profitability coming from semiconductor manufacturing.

And it's one of the primary reasons MU's valuation has rerated.

The earnings aren't merely accounting profits.

Micron generated:

$25.39B operating cash flow

during Q3.

Net capital expenditures were approximately:

$7.1B

Leaving adjusted FCF around: $18.3 billion in one quarter.

This represents one of the biggest improvements in the Micron investment thesis.

Historically, memory producers could earn huge profits and then pour enormous amounts back into fabs right before the cycle collapsed.

Micron is currently generating enough cash that even enormous capital expenditures leave significant FCF.

Micron finished Q3 with approximately:

in cash, marketable investments and restricted cash.

The balance sheet therefore isn't an immediate concern.

The more important issue is future capital intensity.

Micron expects approximately:

And manufacturing investment will continue.

That is simultaneously bullish and bearish.

Demand is so strong that Micron needs substantially more capacity.

Semiconductor cycles usually end because manufacturers respond to shortages with massive capital spending.

That's why investors should closely monitor:

Capex growth vs. demand growth.

HBM is arguably the most important product in the Micron story.

Micron currently has HBM4 in high-volume shipments for its lead customers platform, while qualification samples have been sent to multiple other customers.

Micron is simultaneously developing HBM4E using its 1-gamma DRAM process, with volume production expected during calendar 2027.

The roadmap essentially looks like:

HBM3E, HBM4, HBM4E, future generations

The important point is that Micron isn't merely participating in HBM. It has managed to remain near the technological frontier.

I'd separate Micron's moat into four areas.

Advanced DRAM manufacturing requires enormous fabs, specialized equipment, extreme precision and decades of process knowledge. A startup cannot simply raise $10 billion and compete with Micron.

HBM adds even greater complexity through stacking, packaging, thermal management, power efficiency and integration with accelerators.

Nvidia and other accelerator companies cannot casually swap memory suppliers. Products must be tested and qualified. Once Micron is designed into a platform, that creates greater customer stickiness than traditional commodity DRAM.

There are essentially three major leading-edge DRAM players: SK Hynix, Samsung, Micron That industry structure is substantially better than a fragmented commodity market.

This does not mean Micron has an Nvidia-like moat. SK Hynix is an extremely capable competitor and has been a leader in HBM. Samsung remains one of the world's largest semiconductor manufacturers. And another longer-term risk is emerging:

China wants domestic memory production. CXMT is becoming increasingly important in conventional DRAM. The immediate HBM threat is limited relative to Micron/SK Hynix/Samsung, but over a 5–10 year horizon, Chinese memory capacity deserves serious attention. If Chinese producers eventually flood conventional DRAM markets with capacity, pricing could weaken even if advanced HBM remains attractive.

HBM gets most of the attention, but Micron continues developing conventional server memory.

On September 15, Micron announced what it described as the world's first 512GB DDR5 RDIMM, targeting next-generation servers.

The module reaches speeds up to 9,200 MT/s while Micron says it reduces operating power by more than 60%.

That's important because the AI thesis isn't just: "sell HBM to Nvidia." It's increasingly: Sell dramatically more memory throughout the entire AI server. That makes the opportunity much larger.

HBM gets most of the attention, but Micron continues developing conventional server memory.

On September 15, Micron announced what it described as the world's first 512GB DDR5 RDIMM, targeting next-generation servers.

The module reaches speeds up to 9,200 MT/s while Micron says it reduces operating power by more than 60%.

That's important because the AI thesis isn't just: "sell HBM to Nvidia." It's increasingly: Sell dramatically more memory throughout the entire AI server. That makes the opportunity much larger.

CEO Sanjay Mehrotra has overseen an extraordinary transformation in Micron's earnings profile.

Management has simultaneously:

The challenge from here is different.

Management must resist the classic temptation of memory cycles: Overbuilding during the boom.

Capital allocation over the next 2–3 years could therefore be more important than what management did during the previous three.

Another important development is Micron's push toward longer-term customer agreements.

The company characterized agreements announced alongside Q3 results as "transformational Strategic Customer Agreements."

This matters enormously for the long-term investment case.

Historically:

Memory price ↑

→ customers order aggressively

→ manufacturers expand

→ customers reduce orders

→ inventories rise

→ memory prices collapse.

Longer-term agreements can potentially improve:

If that becomes structurally widespread, it supports the argument that future Micron cycles could be less violent than historical cycles.

Now we reach the hardest part of the analysis.

At roughly $1,096, Micron's market capitalization is approximately:

Trailing EPS is approximately $44.17, giving MU a trailing P/E around: ~25×.

But trailing earnings are almost meaningless right now because profitability has accelerated dramatically.

The market is valuing MU on forward earnings.

Recent market estimates have put FY2027 EPS around the mid-$150s, although estimates are changing rapidly; one recent compilation cited roughly $155 EPS after estimates had risen sharply over the preceding months.

Using $155:

$1,096 ÷ $155 =

That appears incredibly cheap. But there is a major caveat.

Cyclical stocks often have their lowest P/E immediately before earnings decline.

Imagine MU earns: $155 EPS and trades at:7× That looks extremely inexpensive.

But suppose memory prices normalize and EPS falls to:

$75 Then today's $1,096 stock actually represents: 14.6× normalized earnings.

That's not nearly as cheap. This is why I'd argue the most important Micron metric isn't FY27 EPS.

It's: Normalized EPS. What can Micron sustainably earn across FY2028–2031? That's what ultimately determines today's intrinsic value.

Consider several hypothetical long-term earnings levels:

This table explains almost the entire debate.

At ~$1,096, MU isn't expensive if sustainable EPS is $100–150.

It's much more questionable if normalized earnings eventually return to $60–80.

So your Micron thesis ultimately comes down to:

Has AI permanently raised Micron's normalized earnings power?

This is now the biggest near-term catalyst.

Micron officially reports fiscal Q4 results on:

Management's existing guidance is:

Revenue: $50B ± $1B

Gross margin: ~86%

GAAP EPS: $30.73 ± $1

Non-GAAP EPS: $31 ± $1

But those headline numbers aren't what I'd watch most closely.

I'd focus on six things:

The stock is already pricing enormous growth.

Therefore simply "beating earnings" may matter less than what management says about 2027.

This is the most important company-specific news risk right now. Micron employees in Taiwan are pushing for a permanent profit-sharing system. Taiwan matters because it is a major production center for Micron's: DRAM and HBM.

The union has warned that it could move toward strike action. However, as of the latest reliable reporting: no strike had been called and production had not been affected. This isn't currently thesis-breaking. But it's absolutely something investors should monitor. A meaningful production interruption would be much more important given how tight memory supply already is.

This is arguably the most important industry variable.

Right now the environment is exceptionally favorable:

AI demand ↑

HBM demand ↑

DRAM demand ↑

constrained leading-edge supply

=

memory pricing ↑

That produces Micron's extraordinary margins. But manufacturing expansion is underway.

Micron is expanding: DRAM production, HBM packaging, Idaho manufacturing, New York manufacturing,  Singapore capacity

Some new capacity begins contributing during 2027. That's exactly what investors need to monitor. If demand continues growing faster than capacity: Bullish.

If capacity growth catches demand: Margins normalize. If capacity exceeds demand:

classic memory downturn.

MU's price action itself tells you something important.

At approximately $1,096 today, the stock is up roughly 5% intraday and has moved dramatically higher during 2026.

It is no longer a neglected semiconductor stock. Micron has become one of the market's premier AI infrastructure trades.

That has two consequences.

Institutional investors increasingly recognize memory as an AI bottleneck rather than just a commodity.

Expectations have risen dramatically. The market now expects: huge earnings, huge margins, continued shortages,  successful HBM4 ramp, strong FY27 guidance

That makes earnings reactions potentially more volatile.

Micron sits at the intersection of several macro themes.

Most important. Microsoft, Meta, Google, Amazon, Oracle and other hyperscalers continue investing heavily in AI infrastructure.

That spending eventually flows through:

Data center

→ accelerator

→ HBM

→ DRAM

→ storage.

Higher rates affect valuation multiples across technology. But Micron's extremely low forward earnings multiple makes it somewhat different from extremely high-multiple AI stocks.

PCs, smartphones, automotive and conventional servers remain economically sensitive. A recession could hurt those businesses even if AI remains strong.

Semiconductors remain central to U.S.–China technology policy. Export restrictions and China's effort to develop domestic semiconductor capacity create both risks and opportunities.

I'd organize the catalyst stack like this:

September 30 earnings — immediate.

FY27 guidance — arguably more important than Q4 results.

HBM4 production ramp — verifies technological competitiveness.

HBM4E qualification — critical for maintaining position into the next generation.

Continued AI capex growth — supports memory demand.

Strategic customer agreements — potentially reduce cyclicality.

New manufacturing capacity — allows Micron to capture greater demand.

Additional capital returns/buybacks — possible use of extraordinary FCF.

The full bear case deserves equal attention.

Still the biggest risk. If supply catches demand, memory pricing can decline extremely quickly.

86% gross margin is extraordinary. Even falling from:

86% → 65%

could dramatically reduce earnings despite continued revenue growth.

Samsung has enormous financial and manufacturing resources. A successful HBM push could intensify competition.

SK Hynix remains arguably Micron's strongest direct HBM competitor.

Long-term conventional DRAM competition could increase materially.

Micron increasingly depends on AI infrastructure spending. Any hyperscaler capex slowdown would matter.

Micron and competitors are spending billions adding capacity. Today's shortage can become tomorrow's oversupply.

Production concentration creates geopolitical and now labor-related risks. The current union dispute has not disrupted production, but it bears monitoring.

The stock has already rerated enormously. At this point, merely reporting "good" results may not be sufficient.

Instead of anchoring on FY27 peak EPS, I'd value Micron using normalized earnings.

Normalized EPS:

$70

Multiple:

10×

Value:

What gets us there?

AI infrastructure growth slows, HBM supply catches demand, DRAM pricing declines, gross margins normalize sharply and conventional memory cyclicality returns.

Normalized EPS:

$105

Multiple:

12×

Value:

This assumes AI permanently raises memory intensity and Micron's normalized profitability, but margins fall substantially from today's extraordinary levels.

That would imply roughly 15% above today's ~$1,096 price.

Normalized EPS:

$140

Multiple:

14×

Value:

This requires a genuinely structural transformation:

HBM demand remains extremely strong, memory supply remains disciplined, HBM4/HBM4E succeed, strategic agreements reduce cyclicality and Micron sustains materially higher margins than historical norms.

This is something I'd include in every stock report you eventually publish.

Watch for:

FY27 revenue estimates rising

HBM sold out well into future periods

HBM4E customer wins

Gross margins remaining >75–80%

Strategic agreements becoming increasingly common

Capex remaining disciplined

AI memory demand continuing faster than supply

Watch for:

DRAM prices falling

Inventory rising

Samsung aggressively adding supply

HBM pricing weakening

Gross margin guidance falling rapidly

Hyperscaler capex cuts

Customer agreement cancellations

Micron dramatically accelerating capacity beyond demand

Those are far more important signals than whether MU drops 10% because the Nasdaq has a bad week.

If I were listening to the earnings call, my checklist would be:

HBM demand through 2027

→ Is supply still effectively committed?

HBM4

→ How quickly is production ramping?

HBM4E

→ How many customers are qualifying it?

Gross margins

→ Does management expect ~86% to persist?

DRAM pricing

→ Is conventional DRAM still tight?

NAND

→ Is storage pricing following DRAM?

Capex

→ How aggressively will Micron expand FY27 capacity?

Strategic agreements

→ Are customers locking in more long-term supply?

China

→ Is CXMT affecting conventional DRAM pricing?

Taiwan

→ Any expected production impact from labor negotiations?

Those answers will tell us much more than the headline EPS beat/miss.

/

Micron's fundamentals today are extraordinary. Revenue has gone from $9.3B to $41.5B YoY, Q4 guidance calls for approximately $50B, gross margins are approaching 86%, quarterly adjusted FCF reached $18.3B, HBM4 is shipping, and HBM4E is progressing toward 2027 production.

Yet MU at ~$1.1K and ~$1.26T market cap is a very different investment than MU before its enormous rerating. The apparent ~7× FY27-style earnings valuation looks extremely inexpensive only if a large portion of current earnings power survives normalization.

So the number I would obsess over isn't FY27 EPS.

It's 2028–2030 normalized EPS.

If AI has permanently moved that figure above roughly $100, today's valuation can be justified relatively easily. If Micron eventually returns toward $60–80 of normalized earnings, the stock is considerably less cheap than its headline forward P/E suggests.

The September 30 earnings report is therefore unusually important, because FY27 supply/demand, HBM commitments, gross-margin guidance and capex commentary should give us much better evidence about which of those worlds we're moving toward.

Rating: BUY
12-Month Price Target: $1,500
Current price: ~ $1,080–$1,100
Implied upside: ~37%

The $1,500 target assumes AI-driven HBM/DRAM demand remains structurally strong, Micron sustains unusually high margins through FY2027, and HBM4/HBM4E keep the company competitive with SK Hynix and Samsung. The primary risk is that current earnings prove closer to peak-cycle earnings than sustainable earnings. Wall Street's current consensus target is around $1,513–$1,564, with recent targets including RBC at $1,500 and TD Cowen at $1,600.

Newsletter verdict: BUY — $1,500 12-month target.

This rating and valuation are an equity-research framework based on the assumptions above, not individualized investment advice.