VectorialDATA

MU — Micron Technology

TechnologyNorth AmericaUnited StatesBlockchain certified

Price when we picked it (Sep 15, 2026)

$937.81

Target: $1489.57 (+71.9%)

P/E Ratio

19.3

P/E Forward

5.7

Dividend

0.06%

Market Cap

$963.6B

EPS

$44.22

Consensus

Strong Buy

What they do

Micron makes memory chips. If the processor is the brain that thinks, memory is the desk where that brain keeps its papers within reach: without the desk, the brain can't work. Micron lives off three products:

  • DRAM: the working memory of computers, servers, and phones. It's the core business (Micron has ~22% of the global market).
  • NAND: the chips where files are stored permanently (the SSDs in laptops and data centers).
  • HBM: the current crown jewel. It's DRAM stacked into towers and placed right next to the AI chip, to feed it data at extreme speed. Every Nvidia AI accelerator absolutely needs it, and only SK Hynix, Samsung, and Micron know how to make it.

Who buys it? In the latest quarter, data centers and the cloud made up ~61% of sales (Cloud Memory: $13.77 billion; Core Data Center: $11.52 billion); the rest came from phones and PCs ($11.52 billion) and from autos and industrial ($4.63 billion). Founded in 1978 in Boise, Idaho, it has ~53,000 employees and is the only US-based memory maker, something Washington rewards with support for building local factories.

Why we like it

1) There's a global memory shortage, and Micron is one of only three companies that make it. AI data centers are soaking up nearly all the memory available on the planet. Result: DRAM and NAND prices multiplied by ~4x in a year, and the three major manufacturers have their production booked through 2027. Building a new factory takes 4-5 years, so the shortage won't be solved tomorrow.

2) Last quarter's numbers are out of this world. Sales of $41.46 billion (+346% vs. a year earlier), a fifth consecutive record quarter. Gross margin of 84.9%: out of every $100 in sales, almost $85 is left after the cost of manufacturing — software-company level margins in a chip factory. Net profit of $28.24 billion in ONE quarter (the entire fiscal year 2025 was $8.54 billion). And for the current quarter, the company itself expects ~$50 billion in sales with a margin of ~86%.

3) It has already sold the near future. All of its 2026 HBM is sold in advance under price and volume contracts, and its HBM4 (the new generation) is already in volume production for its lead customer. In June it also signed a multi-year supply and co-design agreement with Anthropic, the AI lab behind Claude.

4) The entry price helps. We bought at $866.38, ~31% below the ~$1,255 high from late June, after a sector-wide correction. The forward P/E is 5.7, and the average target from 42 analysts is $1,489.57.

5) It closes the loop on the portfolio. We already own KLA (#126), which inspects the wafers, and Teradyne (#118), which tests the chips: they sell the shovels. Micron is the miner: this year it will spend ~$27 billion buying exactly that kind of equipment. Now we own both sides of the table.

Key Risk

Risk #1 is the memory cycle, and it isn't theoretical. Memory is a boom-and-bust business: when it's scarce, prices soar (like now); when there's a glut, they collapse and manufacturers lose money. It just happened to Micron: in fiscal year 2023 it lost more than $5 billion. Today the entire industry is building capacity at the same time (Micron announced a $250 billion plan for US factories; Samsung and SK Hynix are also expanding), and if there's a memory glut by 2028, prices and that 85% margin could fall fast. The forward P/E of 5.7 looks like a steal, but it's the classic trap of cyclical stocks: the stock looks cheapest right when profits are at their peak.

Other real risks: China — state-owned manufacturer CXMT is preparing an $8.55 billion IPO to fund more subsidized capacity, and it was the trigger for the 8% drop on July 15; China also banned Micron from its critical infrastructure back in 2023. Competition — SK Hynix dominates HBM with ~62% of the market (Micron has ~21%), and Samsung is accelerating. Concentration — if the cloud giants pull back on AI spending, demand could cool off overnight. For all of this, it's a position to size with a cool head, not to fall in love with.

This is educational and informational content, not financial advice. Always consult a qualified financial advisor before making investment decisions.

Vectorial Data picked MU on Jul 16, 2026 at $866.38.

Full Research

What Micron is

Micron Technology was born in 1978 in Boise, Idaho, and today is the only major memory chip maker based in the United States. It has ~53,000 employees and factories in the US, Taiwan, Singapore, and Japan. Its CEO since 2017 is Sanjay Mehrotra, an industry veteran who previously founded SanDisk.

To understand the business: every computer has a processor (the brain that calculates) and memory (the desk where that brain keeps its papers within reach). Nvidia, AMD, or Apple design brains; Micron makes desks. And in the AI era, something curious happened: brains became so fast that the bottleneck is no longer thinking, it's feeding them data fast enough. Memory stopped being a supporting actor and became the lead.

Only three companies in the world dominate advanced memory manufacturing: SK Hynix and Samsung (both from South Korea) and Micron. It's an extremely closed club: getting in requires decades of accumulated know-how and tens of billions in factories.

How it makes money

Micron sells three product families:

  • DRAM: the working memory of servers, PCs, and phones. Historically ~3/4 of its sales. Micron has ~22% of the global market.
  • NAND: permanent storage chips (the SSDs where files are kept).
  • HBM (high bandwidth memory): the star product. It's floors of DRAM stacked like a tower and placed right next to the AI chip, connected by thousands of channels, to feed it data at extreme speed. Every Nvidia accelerator carries several of these towers. Without HBM, the world's most expensive AI chip is left waiting for data, like a star chef with no assistants bringing ingredients.

In the latest quarter, sales broke down as follows: Cloud Memory $13.77 billion (memory for the cloud and AI, the segment that includes HBM), Core Data Center $11.52 billion (traditional data centers), Mobile & Client $11.52 billion (phones and PCs), and Automotive & Embedded $4.63 billion (autos and industrial). In other words: ~61% of the business already comes from data centers.

An important structural shift: memory always used to be sold as a commodity, at the price of the day. Today, HBM and much of advanced DRAM are sold under multi-year price and volume contracts signed in advance. Micron has already sold all of its calendar-year 2026 HBM. That gives a revenue visibility this business never had before.

Last quarter's numbers, and the year's

Fiscal Q3 2026 (ended May 28, 2026, reported June 24):

  • Sales: $41.46 billion, +346% vs. the same quarter a year earlier ($9.30 billion) and +74% vs. the prior quarter. Fifth consecutive record quarter, and it beat the company's own guidance ($33.5 billion) by 24%.
  • Gross margin: 84.9% (adjusted; 84.6% GAAP), an all-time record. Out of every $100 in sales, almost $85 is left after the cost of manufacturing.
  • Net income: $28.24 billion GAAP ($24.67 per share); $28.86 billion adjusted ($25.11 per share).
  • Cash generated from operations: $25.39 billion; adjusted free cash flow: $18.3 billion, after investing $7.1 billion in factories during the quarter.
  • Guidance for the current quarter (fiscal Q4): sales of ~$50 billion (±$1 billion) with a gross margin of ~86% and earnings per share of ~$31. A year earlier, that same quarter had $11.3 billion in sales.

Full fiscal year 2025 (ended August 2025): sales of $37.38 billion and net income of $8.54 billion. Read that again: last quarter alone had more sales than the entire prior fiscal year, and earned three times as much profit in three months as it did in those twelve. That's the size of the leap.

The dividend is $0.15 per quarter: a yield of 0.06%, symbolic. This stock isn't bought for income.

The thesis: why now

1) Structural memory shortage. AI data centers are absorbing nearly all of the world's memory: for the first time ever, they're expected to consume more than half of all DRAM and NAND produced in 2026. DRAM prices rose 55-60% in the first quarter of 2026 alone (TrendForce) and multiplied by ~4x in a year. All three manufacturers are booked through 2027, and a new factory takes 4-5 years to build: real capacity relief won't arrive until late 2027 or 2028.

2) HBM is the fastest-growing market in the sector. From ~$35 billion in 2025 to a projected ~$100 billion in 2028. Micron came late to HBM but has already captured ~21% of the market, its HBM4 (new generation) is already in volume production for its lead customer (Nvidia's platform), and it has shipped samples to more customers. In June it also signed a multi-year agreement with Anthropic (the lab behind Claude): memory and storage supply plus architecture co-design, and it even invested in Anthropic's funding round.

3) The geopolitical card. Being the only US-based memory maker is worth its weight in gold in Washington. Micron announced a $250 billion plan for factories and development in the US (Idaho, with production starting mid-2027; New York, with up to four factories by 2030+). The government wants it to exist and thrive.

4) The entry point. The stock touched ~$1,255 in late June and corrected ~31% down to our $866.38, dragged down by sector-wide profit-taking (SK Hynix's Nasdaq debut and the scare over China's CXMT). We're buying the pullback, not the euphoria.

5) It fits the portfolio. We already own KLA (pick #126), which inspects wafers for defects, and Teradyne (pick #118), which tests finished chips, plus ARM. They sell the tools; Micron is the one buying them: this fiscal year it will invest ~$27 billion equipping factories. With Micron, we go from owning the shovels to also owning the mine, right when the gold (memory) is at its highs.

Valuation: cheap or a trap?

  • Trailing 12-month P/E: 19.3 ($866.38 / $44.22 in earnings per share). For a company that multiplied its sales by 4.5x, that's a modest multiple: the market already doubts it will last.
  • Forward P/E: 5.7. It reads like an extreme bargain, but watch out: in cyclical stocks, the lowest P/E usually shows up near the peak of earnings. That 5.7 assumes earnings keep growing; if the cycle turns, the «E» shrinks and the real multiple ends up much higher.
  • Market cap: $963.6 billion, knocking on the door of the trillion-dollar mark. Three years ago, at the bottom of the cycle, Micron was losing money; today the market treats it as a core piece of AI infrastructure.
  • Analysts: 42 cover it, strong-buy consensus, average price target $1,489.57 (~72% above our entry price). The most bullish (Barclays, among others) have targets of $2,000, betting that margins hold up through 2027.
  • Against peers: SK Hynix (the HBM leader, recently listed on Nasdaq after raising $26.5 billion) and Samsung also trade at single-digit multiples on projected earnings: the market is applying the «this is a cycle» discount to the ENTIRE sector.

Our read: we're not paying for euphoria (19x on real, already-reported earnings), but we're also not buying the 5.7x as if it were permanent. The truth is somewhere in between, which is exactly why position size matters as much as the thesis.

The risks

1. The memory cycle: the risk that rules them all

Memory is the most cyclical business in semiconductors. When it's scarce, prices soar (like now); when there's a glut, they collapse, because manufacturers can't just switch off factories that cost tens of billions. This isn't theory: in fiscal year 2023, Micron lost more than $5 billion with the same people, the same factories, and the same technology. Today the entire industry is building capacity at once (Micron's $250 billion, Samsung and SK Hynix expanding, China subsidizing). If supply catches up with demand by 2028, the 85% margin won't just drop to 80% — it could fall to 30% or less. This risk sits above all the others.

2. Korean competition: Micron is third, not the leader

SK Hynix dominates HBM with ~62% of the market and was the first to qualify each generation with Nvidia; Samsung, the sector's financially strongest player, is racing to catch up with its HBM4. Micron has ~21% and is climbing, but it's fighting two bigger rivals in memory. If either one decides to buy market share by cutting prices, all three companies' margins suffer.

3. China: subsidized competition and a banned market

China's state-owned CXMT is preparing an $8.55 billion IPO to fund even more subsidized DRAM capacity: it was the trigger for Micron's 8% drop on July 15. China competes with state money that doesn't require profitability. On top of that, since 2023 Chinese regulators have banned Micron's products from the country's critical infrastructure, so Micron competes with one hand tied behind its back in the world's second-largest market.

4. Concentration in AI spending

The boom depends on a handful of giants (Nvidia as the top HBM customer, and behind it the cloud hyperscalers) continuing to spend record sums on data centers. If AI returns disappoint and that spending pauses for even two quarters, memory demand could cool off overnight. July's correction was exactly that fear playing out.

5. A massive capital bet

~$27 billion in investment this fiscal year alone, plus a $250 billion long-term plan. If demand keeps pace, that's vision; if not, it's crushing fixed costs. Chip factories can't hibernate: they either produce or burn cash.

6. The cheap-multiple trap

The forward P/E of 5.7 can lure in investors who don't know the sector. Historically, buying memory stocks because they look «cheap on P/E» right at the peak of prices has been the classic way to lose money in this sector. We're buying knowing this: the thesis is that this cycle is longer and more contracted than previous ones, not that the cycle is dead.

Verdict

Micron is the most direct way to own the scarcest input of the AI boom: memory. Last quarter's numbers are unprecedented in the company's history (sales of $41.46 billion, an 84.9% margin, and guidance of $50 billion for the next one), 2026 production is already sold under contract, and we're buying in 31% below the high, with 42 analysts seeing ~72% of upside.

But let's be adults about it: this is the most cyclical pick in the portfolio. The same company earning $28 billion in a single quarter today lost $5 billion in a single year just three years ago. The bet isn't «memory will never be in oversupply again» — it's that between the signed contracts, insatiable HBM demand, and factories that take years to build, this cycle has runway through 2027-2028, and that at 19 times already-reported earnings, we aren't paying for euphoria. It pairs perfectly with KLA and Teradyne: they collect the toll on building factories; Micron collects on the scarcity. This is a position to watch — track DRAM prices and capacity announcements — not one to buy and forget.

More picks in Technology

See all Technology picks →

More picks from United States

See all picks from United States →
Researched: Jul 16, 2026Updated: Jul 16, 2026Next review: Jan 16, 2027

This is not financial advice. Consult a certified financial advisor.

The author may hold positions in the securities discussed.

Past performance does not guarantee future results.