Micron trades at roughly 6.4x forward earnings, and almost everyone reads that number backwards. A single-digit multiple on a $1 trillion company looks like the cheapest large-cap in the market. In memory semiconductors it has historically meant the opposite: the market applies a trough multiple to peak earnings precisely when it believes those earnings are about to break. MU closed 6 August at $881.47, down 27.4% from its 25 June closing peak of $1,213.56, and still up 688% from the $111.87 it printed a year ago. The question is not whether Micron is cheap. It is whether $138 of annualised EPS is a floor or a ceiling.
Here is what makes this cycle genuinely different, and it is the fact most of the bearish commentary is skipping. Samsung, SK hynix and Micron have reportedly sold out their entire 2027 DRAM and HBM capacity, and Micron has locked in 16 multi-year customer agreements. You cannot have a classic inventory glut in a year where 100% of output is already contracted. That removes the standard memory bear case from the table for 2027 entirely. So the argument is not really about next year at all. It is about 2028, and about whether hyperscalers keep writing the cheques once the current agreements roll. Both sides of this trade are arguing about a year almost nobody is naming.
Key facts
- Share price: $881.47 close on 6 August 2026, down 27.4% from the 25 June closing peak of $1,213.56 – Nasdaq, 6 Aug 2026
- 52-week range: $106.75 to $1,255.00; the stock is up 688% from its August 2025 closing low of $111.87 – Nasdaq, Aug 2026
- Market capitalisation: $1.008 trillion, implying roughly 1.14 billion shares outstanding – Nasdaq, 7 Aug 2026
- Quarterly EPS trajectory: $2.86 (Aug 2025), $4.61 (Nov 2025), $12.08 (Feb 2026), $24.89 (May 2026) – a beat in all four quarters – Nasdaq consensus and actuals, Aug 2026
- Forward consensus: $31.17 for the August 2026 quarter and $34.54 for November 2026 – Nasdaq, Aug 2026
- Street target: mean price target near $1,568.74; BofA’s Vivek Arya at $1,550, Itau BBA at $1,697.09 – TheStreet and Yahoo Finance, Aug 2026
- 2027 capacity: DRAM and HBM output at Samsung, SK hynix and Micron reported fully booked for 2027 – TweakTown, Aug 2026
- HBM market forecast: $246 billion by 2030, with Micron holding 16 multi-year customer agreements – BofA via 24/7 Wall St., Jul 2026
The multiple everyone is misreading
Start with the arithmetic, because it is the crux of the disagreement and it is rarely laid out properly.
Micron’s trailing four quarters of EPS come to $44.44, made up of $2.86, $4.61, $12.08 and $24.89. At $881.47 that is a trailing multiple of about 19.8x, which is unremarkable. But consensus for the August 2026 quarter is $31.17 and for November 2026 is $34.54. Annualise the November run-rate alone and you get roughly $138 in EPS, which puts MU on about 6.4x forward earnings. For a trillion-dollar company that is an extraordinary number.
Retail reads 6.4x and sees a bargain. Anyone who has sat through a memory cycle reads 6.4x and hears an alarm. Memory is the one large-cap sector where a collapsing multiple on rising earnings is the market’s standard way of saying it does not believe the earnings will persist. DRAM producers have repeatedly looked cheapest at the exact top, because the P/E denominator is a cyclical peak rather than a run-rate. That is the single most important thing to understand about this chart, and it is why the gap between the $881 tape and the $1,568.74 mean target is not simply an analyst error waiting to correct.
The honest framing is this: the multiple is not evidence for either side. It is a statement that the market and the sell side disagree violently about the durability of Micron’s earnings, and the price is where that disagreement is being settled. We have made the same argument about multiple compression in the wider semiconductor complex in our Texas Instruments TXN price prediction.
What actually caused the drawdown
The 27.4% fall from the June peak was not one event. It was four pressures arriving together.
The first was straightforward profit-taking. Memory stocks went vertical in the first half of 2026, and MU’s May close of $971.00 was 88% above its April close of $517.16. A move like that in a single month invites de-risking regardless of fundamentals.
The second was the hyperscaler return-on-investment debate. Investors have started asking, out loud, whether the enormous AI capital expenditure programmes will earn an acceptable return, and memory is the most levered way to express doubt because it is the purest capacity play in the AI supply chain.
The third was China. Concerns that domestic Chinese DRAM capacity additions could outpace price growth reintroduced the oldest bear argument in memory: supply catching demand.
The fourth was sentiment contagion. A global tech selloff tied to AI valuation fears, compounded by weak preliminary numbers from Samsung, dragged Micron down alongside Nvidia, AMD, Western Digital, Applied Materials and Marvell. That is a sector re-rating, not a Micron-specific verdict, and readers following that broader unwind may recognise the same dynamic in our AMD stock forecast.
Note what is absent from that list: any deterioration in Micron’s own reported numbers. The company has beaten consensus in each of the last four quarters, and by widening margins – $12.08 against an $8.64 estimate in February, $24.89 against $20.98 in May. The drawdown is entirely an argument about the future.
Supply, demand and the sold-out year
The supply picture is the strongest card the bulls hold, and it deserves precision rather than slogans.
Reporting indicates that AI demand has booked substantially all 2027 DRAM and HBM supply across the three major producers. Micron has 16 multi-year customer agreements underpinning its share, and BofA projects the HBM market reaches $246 billion by 2030. High-bandwidth memory is not a commodity in the way conventional DRAM is: it is co-engineered with the accelerator, qualified per customer, and switching suppliers mid-programme is expensive and slow. That gives HBM revenue a contractual quality that historical DRAM revenue never had.
This is the structural change bulls are pointing to, and it is the same argument that carried Micron past a trillion dollars in the first place, as we covered when Wall Street recomputed semiconductor memory valuation. Memory has always been a spot-priced commodity business valued at commodity multiples. If a growing share of output is sold years ahead under multi-year agreements, the business starts to resemble a capacity-constrained infrastructure supplier, and infrastructure suppliers do not trade at 6.4x.
The bear rebuttal is specific and worth stating fairly. Sold out is not the same as sold out at today’s prices. Contracts have pricing mechanisms, and a multi-year agreement signed into a shortage can reprice into a glut. Chinese capacity is additive and is not bound by those agreements. And 2028 is entirely unspoken for, which matters because equity markets discount roughly eighteen months ahead – meaning the market is already trading 2028 while the bulls are still quoting 2027.
“Bank of America added Micron Technology to its US 1 List,” the firm’s roster of highest-conviction US picks, alongside analyst Vivek Arya’s decision to lift his objective to $1,550 from $1,500, per TheStreet. Conviction lists are where banks put reputational weight behind a call, which makes this a more meaningful signal than a routine target revision.
The bull case: $1,550
The bull target is $1,550, BofA’s published price objective, sitting just below the $1,568.74 street mean. It implies roughly 76% upside from $881.47. Three conditions.
First, the 2027 book has to hold at economics close to current terms. That is the load-bearing assumption, and it is the one with the most external validation: three producers, all sold out, in a market where the buyers are the best-capitalised companies on earth.
Second, HBM has to keep taking share of the mix. HBM carries materially better economics than commodity DRAM, so a rising HBM share lifts blended margin even if bit growth is flat. The path to $246 billion by 2030 implies that mix shift continues for years, not quarters.
Third, the market has to accept a higher multiple on the earnings. This is where most of the return lives. At $138 of annualised EPS, $1,550 is roughly 11x – not a heroic re-rating, simply the market conceding that contracted HBM revenue deserves more than a spot-DRAM multiple. The bull case does not need earnings to grow from here. It needs the market to stop treating them as temporary.
The bear case: $520
The bear target is $520, roughly 41% below the 6 August close. Two independent derivations converge there, which is why we prefer it to a target anchored on the drawdown alone.
The first is the price base. $520 is approximately the 30 April 2026 close of $517.16, the level from which the parabola launched before May’s 88% surge. If the AI-memory re-rating unwinds rather than consolidates, the April shelf is the obvious landing zone, because it is the last price the market set before the melt-up.
The second is normalised earnings. Memory has historically bottomed near 10x trough EPS. If 2028 pricing normalises and EPS settles near $52 – roughly 38% of the annualised peak run-rate, consistent with the amplitude of prior memory down-cycles – then 10x gives about $520. The two methods, one technical and one fundamental, land in the same place.
The bear case does not require the 2027 book to fail. It only requires the market to look past it. If investors conclude in the next twelve months that 2028 brings both Chinese supply and contract repricing, the stock does not wait for the evidence. It de-rates first and confirms later, which is exactly what the last two months have looked like.
Two additional risks belong here. Micron pays a $0.60 annualised dividend, a 0.07% yield, so there is no income support in a drawdown. And at a $1.008 trillion market capitalisation, the marginal buyer must now be an index or a mega-cap allocator rather than a sector specialist – a structurally less price-insensitive bid than the one that carried the stock up. The same liquidity dynamic showed up in our coverage of the SpaceX SPCX share unlock.
Bull versus bear at a glance
| Factor | Bull case ($1,550) | Bear case ($520) |
|---|---|---|
| 2027 supply | Sold out across all three producers | Sold out, but not at fixed prices |
| Forward multiple | 6.4x is an anomaly that must close | 6.4x is the market pricing peak earnings |
| HBM mix | Contracted revenue deserves a re-rating | Co-engineering premium erodes as rivals qualify |
| Earnings momentum | Four consecutive beats, widening margins | Beats are a lagging indicator at cycle peaks |
| China capacity | Years behind on HBM qualification | Additive DRAM supply, outside the contracts |
| Analyst positioning | $1,568.74 mean, BofA US 1 List conviction | Targets 78% above the tape signal lag, not upside |
What happens next
Three calls, with the reasoning attached.
The August quarter is a low bar and that is the risk. Consensus sits at $31.17 after four straight beats, so a beat is close to fully expected and unlikely to move the stock much on its own. The variable that matters is 2028 commentary. Any management language about contract structures or pricing mechanisms extending beyond 2027 is worth more than the printed number, because 2028 is the year actually in dispute.
Watch HBM as a share of revenue, not revenue growth. Total revenue can rise on DRAM pricing alone and tell you nothing about durability. HBM share is the tell for whether Micron is becoming a contracted infrastructure supplier or remains a spot-priced commodity producer. That single ratio decides which multiple the stock deserves, and therefore decides between $1,550 and $520.
Expect the target-versus-price gap to close from the target side first. A mean target 78% above the traded price is not a stable configuration. Historically the sell side capitulates toward the tape more often than the tape rallies to meet the sell side, so treat published targets as a lagging measure of sentiment rather than a forecast. The same pattern is visible across the space complex in our note on how SpaceX dragged every space stock down.
Our read: the sold-out 2027 book is a genuine structural change and it makes the classic glut thesis unavailable for next year, which is why we do not treat $520 as the base case. But the 6.4x multiple is not the bargain retail thinks it is – it is the market’s considered opinion that these earnings are the peak. Resolving that requires evidence about 2028, and no one has it yet.
Frequently asked questions
Why is Micron stock falling in August 2026?
MU is down 27.4% from its 25 June closing peak of $1,213.56 on four combined pressures: profit-taking after an 88% single-month gain in May, investor debate over hyperscaler returns on AI capital expenditure, concerns that Chinese DRAM capacity could outpace price growth, and a broad tech selloff amplified by weak preliminary numbers from Samsung. Micron’s own reported results have beaten consensus in each of the last four quarters.
What is the Micron price target for 2026?
The street mean is approximately $1,568.74. Bank of America’s Vivek Arya carries a $1,550 objective with a Buy rating and has added Micron to the firm’s US 1 conviction list, while Itau BBA raised its target to $1,697.09. Our bull case is $1,550 and our bear case is $520, against a 6 August close of $881.47.
Is Micron stock cheap at 6.4x forward earnings?
It is cheap on the arithmetic and ambiguous in practice. Annualising the $34.54 November 2026 consensus gives roughly $138 of EPS, or about 6.4x at $881.47. In memory semiconductors, however, a low multiple on rising earnings has historically signalled that the market expects those earnings to be a cyclical peak rather than a run-rate. The multiple alone does not settle the question.
Is DRAM and HBM capacity really sold out through 2027?
Reporting indicates AI demand has booked substantially all 2027 DRAM and HBM output across Samsung, SK hynix and Micron, with Micron holding 16 multi-year customer agreements. The important caveat is that being sold out is not the same as being sold out at fixed prices, and 2028 capacity remains uncommitted. Since equity markets discount roughly eighteen months forward, 2028 is the year now being traded.
How much has Micron stock risen in the past year?
MU closed at $111.87 in August 2025 and at $881.47 on 6 August 2026, a gain of 688%. It peaked at $1,213.56 on 25 June 2026, which was an 11-fold increase from the prior year’s low. The 52-week intraday range is $106.75 to $1,255.00, and the company crossed a $1 trillion market capitalisation during the run.
What would invalidate the Micron bull case?
Evidence that 2027 contracts contain pricing mechanisms that reset materially lower, confirmation of Chinese DRAM capacity arriving faster than expected, or hyperscaler guidance pointing to slower AI capital expenditure in 2028. Any of these would move the debate from supply scarcity to price realisation, which is the axis on which memory cycles have always turned.
This article is for information purposes only and does not constitute investment advice. Price data is as of the 6 August 2026 close and pre-market trading on 7 August 2026.