The Short Answer
Memory and storage stocks are falling while the semiconductor complex rises because this is a position unwind concentrated in one cluster, not a market-wide reduction in risk. The distinguishing evidence is simple: if investors were cutting risk broadly, the index would fall with the group. It did not. As of 10:34 ET on August 7, 2026 (session still open), SPY was +0.50% and the semiconductor ETF SMH was +1.04%, while Seagate was -9.74%.
A second, less-noticed fact makes this concrete: every storage name opened higher and was sold anyway. The group did not gap down on bad news. It gapped up, then gave the gap back and more within sixty minutes.
What Actually Happened in the First Hour
All figures below are live intraday readings pulled at 10:34 ET (22:34 CST), roughly 64 minutes into the session. These are not closing prices and will change before the close.
| Ticker | Gap at open | As of 10:34 ET | Move from the open |
|---|---|---|---|
| Seagate (STX) | +1.14% | -9.74% | -10.76% |
| SanDisk (SNDK) | +3.94% | -4.42% | -8.05% |
| Western Digital (WDC) | +1.31% | -5.47% | -6.70% |
| Micron (MU) | +2.63% | -2.85% | -5.34% |
Against the same clock, the broader complex held:
| Ticker | Gap at open | As of 10:34 ET | Move from the open |
|---|---|---|---|
| NVIDIA (NVDA) | +1.18% | +1.83% | +0.64% |
| Broadcom (AVGO) | +1.95% | +1.24% | -0.70% |
| SMH (semis ETF) | +2.12% | +1.04% | -1.06% |
| QQQ | +0.78% | +0.77% | -0.01% |
| SPY | +0.32% | +0.50% | +0.18% |
The spread between Seagate and SMH in a single hour is over 10 percentage points, with both starting the day green.
Why This Is Dispersion, Not De-Risking
The market-structure question worth asking is not "is it going down" but "is one cluster going down, or is everything going down together." The two have different causes and different resolutions.
- Broad de-risking shows up as high index-level volatility and rising correlation between names. Everything falls together because the seller is reducing gross exposure, not expressing a view.
- A momentum unwind shows up as high single-stock volatility with low index correlation. Specific crowded positions get liquidated while the index absorbs it.
Today's tape is the second pattern in an unusually clean form. The index is up. Semis are up. Four names are down between 2.9% and 9.7%. Nothing about SPY at +0.50% is consistent with a market cutting risk.
This matters because the two regimes resolve differently. Dispersion tends to end when the crowded position is fully liquidated, which can happen quickly and violently. Broad de-risking tends to end only when the macro driver changes.
Seoul Said the Opposite, Six Hours Earlier
Korea and Hong Kong closed before the US opened, so these are final closing values, not live prints.
At the Korean close on August 7, the memory complex looked like it was stabilizing:
- SK Hynix -4.88%, versus -10.37% the prior session: the decline roughly halved
- Samsung Electronics +0.22%: positive, after -6.30% the prior session
- Hanmi Semiconductor -4.09%
And the rest of the Korean market was firmly risk-on: Samsung SDI +7.49%, LG Electronics +5.59%, LG Energy Solution +4.35%. Money was moving within Korea, not leaving it.
The flow data said the same thing. KOSPI investor-type net flows, in 100 million KRW, with all four categories summing to zero as a validation check:
| Session | Individual | Foreign | Institutional | Other corp. |
|---|---|---|---|---|
| Aug 6 | +33,367 | -32,893 | -1,214 | +740 |
| Aug 7 | +2,675 | -8,651 | +5,854 | +122 |
Foreign net selling fell 74% in one session, and institutions flipped from net seller to net buyer. Within the foreign total, the two memory majors accounted for -4,815 (56% of the foreign net sell, down from 71% the prior session), implying roughly -3,836 across the rest of the market, versus -9,575 the session before, a 60% reduction.
So the direction of the signal persisted, but its intensity collapsed. Reporting only the direction, "foreigners sold for a second consecutive session," would read as accelerating flight. The magnitude says the opposite: they were backing off.
The US open initially agreed with Seoul. That is exactly why every storage name gapped up. Then Wall Street sold the gap.
The divergence between the Asian close and the US first hour is the actual news of the day.
The Seagate Tell: The Last Holdout Broke
Through this entire drawdown, Seagate had been the name that held. Measured to the August 6 close, it was down just 0.82% over one month while SanDisk was down 27.13%. Over three months it was still up 8.54%.
Today it is the worst performer in the group at -9.74%.
That is a meaningful change in character. When the resilient name in a group breaks, it usually means the selling is no longer discriminating between fundamentals within the sector, and has become a decision about sector exposure itself.
A necessary caveat: we found no August 7-specific news catalyst for Seagate. Public coverage through the prior session concerns Western Digital's and SanDisk's fiscal Q4 results released after the August 5 close, where both beat on revenue but SanDisk's next-quarter guidance midpoint landed below consensus. That explains August 6. It does not explain why Seagate, which was not part of that earnings event and had been outperforming, is today's worst performer. We would rather flag an unexplained move than attach a tidy story to it.
The Window Trap: "Profit-Taking After a Huge Run" Depends Entirely on Your Lookback
The most common explanation offered for this selloff is profit-taking after an enormous rally. Whether that is true depends entirely on which window you choose, and the windows disagree violently.
All figures are close-to-close with an endpoint of the August 6 close, so the windows are clean and comparable. YTD is measured from the December 31, 2025 close.
| Ticker | 1 week | 1 month | 3 months | YTD |
|---|---|---|---|---|
| Micron (MU) | +0.78% | -7.10% | +32.26% | +209.0% |
| SanDisk (SNDK) | -1.67% | -27.13% | -10.74% | +430.2% |
| Western Digital (WDC) | -15.29% | -17.95% | -6.52% | +162.3% |
| Seagate (STX) | +0.15% | -0.82% | +8.54% | +210.5% |
| SMH | +6.05% | -3.63% | +3.95% | +58.7% |
| NVIDIA (NVDA) | +12.28% | +7.28% | +5.49% | +17.6% |
| SPY | +3.62% | +3.11% | +5.00% | +13.3% |
Read the YTD column and "profit-taking after a huge run" is obviously right. Read the one-month column and it is obviously wrong: over the last month these names were down 7% to 27% while SPY was up 3.11%. They were not running up into this. They were already falling.
Both columns are accurate. They support opposite conclusions. Any explanation of this selloff that quotes only one window has been chosen rather than discovered, and the honest reading is that the group had a historic year, peaked in late June, and has been de-rating for six weeks.
An Index at Highs, a Group in a Bear Market
The cleanest single frame. Drawdown from each name's six-month high, measured at the August 6 close:
| Ticker | Drawdown from 6-month high | Date of that high |
|---|---|---|
| SanDisk (SNDK) | -46.10% | June 25, 2026 |
| Western Digital (WDC) | -39.49% | June 18, 2026 |
| Micron (MU) | -27.35% | June 25, 2026 |
| Seagate (STX) | -21.98% | June 22, 2026 |
| SMH | -14.57% | June 22, 2026 |
| SPY | -0.36% | August 4, 2026 |
The S&P 500 ETF set its high three sessions ago and sits 0.36% below it. Storage names sit 22% to 46% below highs set in June. This is not a market correcting. It is one group being repriced inside a market that is otherwise fine.
What Would Change This Read
Falsifiable markers, stated in advance:
- Correlation turning up. If index-level implied correlation starts rising while single-stock volatility stays elevated, the regime is shifting from dispersion to genuine de-risking. That is the first real warning, and it would come before the index breaks.
- The index joining. If SPY and SMH start falling alongside storage rather than absorbing it, the "one cluster" framing is dead.
- Storage stabilizing while semis keep rising. That would mean the rotation completed and the unwind is finished.
- A second failed bounce. Today the market gapped storage up and sold it. If that pattern repeats, it indicates persistent supply above the market rather than a completed liquidation.
The distinction to hold on to: a decline losing follow-through is evidence that selling pressure is exhausting. It is not evidence that a rebound has begun. Today supports the first statement only.
Data Notes and Limitations
- US figures are live intraday, captured 10:34 ET / 22:34 CST on August 7, 2026, about 64 minutes into a session that is still open. They are not closing prices and will change. The gap-and-reversal described here has already occurred and is fixed; the closing values are not.
- Korean figures are final closes. Korea and Hong Kong were shut before the US opened.
- Multi-window returns and drawdowns use an endpoint of the August 6 close deliberately, so that no window is contaminated by a partial, still-moving session. Mixing a one-hour reading into a one-month return is a common and invisible error.
- KOSPI investor-type flows are in 100 million KRW and were validated by checking that all four investor categories sum to zero. Note that this data source revises prior sessions, so both sessions above are quoted from a single retrieval rather than compared across retrievals.
- No August 7-specific catalyst was identified for the US move. We report that rather than supply a narrative.
- This is market analysis, not investment advice. It contains no positions, trades, or recommendations.