Memory chips - DRAM, NAND, and now HBM (High Bandwidth Memory) - are commodity-like products. When demand is strong, Micron prints money. When supply catches up, prices collapse and so does the stock. This cycle has repeated for decades, and the data shows it clearly.
We pulled 10 years of monthly price data for MU, SMH (semiconductor sector ETF), and QQQ (Nasdaq 100) to quantify exactly how cyclical Micron is compared to both its sector and the broader tech market.
The chart tells the story
All three rebased to 100 in January 2016. MU (orange) has returned nearly 70x - but the path was violent. Notice the sharp drawdowns in late 2018, late 2021 into 2022, and late 2024. Each time, MU fell 40-46% while QQQ held up far better.
The drawdown chart: where it hurts
The rebased chart above makes MU look like a rocket ship, but it hides the pain. This chart shows MU's drawdown from its running all-time high. Every time the line touches 0%, MU made a new high. Every dip below is a loss from the peak. Four distinct craters drop below -40%, and MU spent most of 2018-2023 in a drawdown of some kind.
Cyclicality scorecard: MU vs SMH vs QQQ
The numbers below are computed from monthly closing prices, January 2016 through May 2026. MU dominates every volatility and risk metric.
| Metric | MU | SMH | QQQ |
|---|---|---|---|
| Annualized Volatility | 48.8% | 27.7% | 18.6% |
| Beta (vs QQQ) | 1.34 | 1.20 | 1.00 |
| Max Drawdown (2016-2026) | -46% | -40% | -32% |
| Months with >10% Move | 44% | 23% | 7% |
| Total Return (2016-2026) | +6,841% | +2,401% | +638% |
| 12M Return Range | -46% to +693% | -34% to +141% | -32% to +69% |
The standout: 44% of all months had a move greater than 10%. For QQQ, that number is 7%. Almost every other month, MU is making a double-digit move in one direction or the other.
The rolling 12-month return range is even more telling. In any given year, MU could be up 693% or down 46%. QQQ ranges from -32% to +69%. That is a 738 percentage-point range vs 101 for QQQ - roughly 7x wider.
Major drawdowns since 2016
MU has suffered four drawdowns exceeding 20% in the last 10 years. QQQ had one (the 2022 bear market at -32%). Each MU drawdown had a memory-specific catalyst, not just general market weakness.
| Period | Peak | Trough | Drawdown |
|---|---|---|---|
| May 2018 - Dec 2018 | $56.17 | $30.95 | -45% |
| Feb 2021 - Oct 2021 | $89.27 | $67.49 | -24% |
| Dec 2021 - Sep 2022 | $91.07 | $49.15 | -46% |
| Jun 2024 - Apr 2025 | $130.61 | $76.77 | -41% |
The full history: MU has done this before
The 2016-2026 data only captures part of the picture. Going back further, MU has experienced even more extreme cycles. After the dot-com boom, MU dropped 99%. During the GFC, it fell 88%.
| Cycle | Peak | Trough | Drawdown |
|---|---|---|---|
| Dot-com Boom/Bust | ~$97 (2000) | ~$0.65 adj (2003) | -99% |
| Great Financial Crisis | ~$13 (2007) | ~$1.60 adj (2009) | -88% |
| 2014-2016 Downcycle | ~$25 (2014) | ~$9 (2016) | -64% |
| 2018 Oversupply | $56 (2018) | $31 (2018) | -45% |
| 2022 Rate Shock | $91 (2021) | $49 (2022) | -46% |
For context, QQQ's worst drawdown was roughly -80% during the dot-com bust and -50% during the GFC. MU exceeded even those in four separate cycles.
Why does this keep happening?
Memory chips follow a textbook commodity cycle - sometimes called cobweb dynamics:
- Shortage phase:Demand outpaces supply. DRAM/NAND/HBM prices spike. Micron's margins expand dramatically and earnings surge.
- Capex boom: High prices attract investment. Micron, Samsung, and SK Hynix all announce massive fab expansions. New capacity takes 18-24 months to come online.
- Overcapacity:New supply hits the market just as demand normalizes. Memory prices crash. Micron's earnings collapse - sometimes into losses.
- Capex cut: Producers slash spending. Supply tightens. The cycle restarts. Each full rotation takes 3-5 years.
The P/E trap
This cycle creates a counterintuitive valuation pattern. MU often looks cheapest (low P/E) right at the peak because current earnings are temporarily inflated. The market is not pricing the current quarter - it is pricing the coming downturn. Conversely, MU looks most expensive (high P/E or even negative earnings) at the trough, which is usually the best time to buy.
What about HBM?
HBM (High Bandwidth Memory) is a specialized type of stacked DRAM used in AI accelerators - Nvidia's H100 and B200 GPUs, Google's TPUs, and AMD's MI300. HBM has higher margins and a more concentrated customer base than commodity DRAM. The current MU rally (+722% since January 2025) is largely driven by HBM demand from AI datacenter buildouts.
The bull case is that HBM breaks the cycle - it has structural demand that does not follow the traditional PC/smartphone replacement cycle, and only three companies (Micron, Samsung, SK Hynix) can produce it. The bear case is that this is the same pattern as every other memory boom - demand is real but capacity expansion will eventually catch up, and AI capex could slow if hyperscaler spending cools.
The data does not predict which outcome wins. But it does tell you that every previous MU boom - including ones with strong fundamental justification - ended with a 40%+ drawdown. Knowing that history helps size positions accordingly.
The leveraged ETF top signal
When Wall Street rushes to launch a themed ETF on a hot sector, it often marks peak retail enthusiasm. The Roundhill Memory ETF (DRAM) launched on April 2, 2026 at $27.00 and hit $52.80 in just 26 trading days - a 96% gain. It pulled in $5 billion in assets and added $1 billion in a single day. CNBC called it "the hottest ETF since bitcoin-mania."
That comparison is doing more work than CNBC intended. Here is what happened after previous themed ETF launches:
| ETF | Launch | Drawdown After |
|---|---|---|
| BITO | Oct 2021 | -78% |
| MJ | Jan 2018 inflow mania | -63% |
| TARK | May 2022 | -34%* |
| DRAM | Apr 2, 2026 | +96% |
BITO (ProShares Bitcoin Strategy) launched October 18, 2021 at $11.01. Bitcoin peaked at $69K within weeks. BITO bottomed at $2.63 thirteen months later - a 78% drawdown from its launch-week high of $11.84.
MJ (cannabis ETF) saw massive inflows in January-February 2018, pulling in $401 million in two months. Cannabis stocks peaked in September 2018 at $387 and ground down to $141 by November 2019 - a 63% decline.
TARK (2x leveraged ARKK) launched May 2, 2022. By then ARKK had already crashed 70% from its $155 peak, so you might call this a contrarian buy signal. But ARKK fell another 34% after TARK launched, bottoming at $31 in December 2022. The total peak-to-trough was -80%.
DRAM launched into a 96% rally with volume exploding from 2.5 million shares on day one to 47 million shares by early May - a 19x increase. The $1 billion single-day inflow on May 8 was the largest for any non-index ETF in 2026. This is the kind of data point that historically marks the moment of maximum enthusiasm, not the beginning of a move.
Leveraged ETFs do not cause tops. But they are a symptom of the same dynamic: retail demand peaking, issuers rushing to monetize the theme, and breathless financial media coverage. The base rate from BITO, MJ, and TARK says the underlying asset drops 34-78% within 6-18 months of the euphoria peak. DRAM may be different - but the pattern is worth watching.
Key takeaways
- MU is 2.6x more volatile than QQQ even though it is a tech stock. The extra volatility comes from memory chip commodity cycles, not just market beta.
- Every previous boom ended with a 40%+ drawdown - including booms with strong fundamental justification. Size positions accordingly.
- Low P/E on MU is a warning, not a buy signal. Earnings are cyclical, not stable. The stock looks cheapest right before the bust.
- HBM may break the pattern, but the base case should assume the cycle repeats until proven otherwise. Three companies producing HBM today does not guarantee supply discipline forever.