The AI infrastructure sell-off has left 13 core AI-infrastructure stocks a median 17.1% below their 2026 closing highs as of the October 19, 2026 close. The valuation damage runs deeper than the price damage: in the five names that reported higher earnings after their peaks, trailing P/E multiples fell further than the shares, by three times as much in $MU's case. $QQQ, $SMH and $NVDA all closed October 19 below their 50-day averages but above their 200-day averages. That is a short-term trend break, not yet a long-term one.
Below: how far each name has fallen, what investors pay for it now, and the dated drivers. More AI coverage sits on the Traders On Trend Technology page.
The Week the AI Trade Broke
The hard part of the move came in five sessions. From the October 9 close to the October 16 close, FMP end-of-day data show:
- Nasdaq Composite: -4.48%
- $QQQ: -5.52%
- $SMH, the semiconductor ETF: -10.53%
- $NVDA: -6.61%
- $ORCL: -9.31%
- $CRWV: -25.21%
- $SPY: -1.97%
- $RSP, the equal-weight S&P 500: +0.49%
That last line is the tell. The equal-weight index rose while the AI complex got hit. This was a factor sell-off, not a market crash.
Monday, October 19 brought a bounce. $NVDA rose 1.38%, $MU 2.97%, $AMZN 4.98% and the Nasdaq 1.11%. $CRWV still closed 0.44% lower. One up day does not repair a broken week.
The Scoreboard: Drawdowns and Multiples
Trailing P/E uses the October 19, 2026 close divided by the sum of the last four reported quarters of diluted EPS. Forward P/E uses the FMP analyst consensus for the next fiscal year not yet reported, and that year is named in each row. NM means not meaningful (losses).
| Stock | Oct 19, 2026 close | 2026 closing high (date) | Off high | Trailing P/E | Forward P/E (FMP consensus FY) |
|---|---|---|---|---|---|
| $NVDA | $217.09 | $239.24 (Oct 6) | -9.3% | 27.4 | 23.4 (FY2027, ends Jan 2027) |
| $AVGO | $334.59 | $481.57 (Jun 2) | -30.5% | 42.7 | 28.7 (FY2026, ends Nov 2026) |
| $AMD | $538.05 | $649.42 (Oct 6) | -17.1% | 138.0 | 70.4 (FY2026, ends Dec 2026) |
| $MU | $904.90 | $1,213.56 (Jun 25) | -25.4% | 12.2 | 5.2 (FY2027, ends Sep 2027) |
| $ORCL | $129.59 | $248.15 (Jun 1) | -47.8% | 20.3 | 15.9 (FY2027, ends May 2027) |
| $CRWV | $61.12 | $137.98 (May 6) | -55.7% | NM | NM (FY2026, ends Dec 2026) |
| $VRT | $206.08 | $376.23 (May 14) | -45.2% | 46.6 | 30.7 (FY2026, ends Dec 2026) |
| $ANET | $182.95 | $216.74 (Oct 9) | -15.6% | 57.7 | 44.5 (FY2026, ends Dec 2026) |
| $SMCI | $32.10 | $50.17 (Jun 2) | -36.0% | 9.8 | 7.4 (FY2027, ends Jun 2027) |
| $MSFT | $533.52 | $543.45 (Oct 13) | -1.8% | 29.7 | 27.0 (FY2027, ends Jun 2027) |
| $META | $705.44 | $777.59 (Sep 24) | -9.3% | 26.6 | 23.0 (FY2026, ends Dec 2026) |
| $GOOGL | $344.18 | $402.62 (May 13) | -14.5% | 17.3* | 16.7* (FY2026, ends Dec 2026) |
| $AMZN | $262.45 | $284.02 (Aug 3) | -7.6% | 21.1* | 20.5* (FY2026, ends Dec 2026) |
*$GOOGL and $AMZN multiples are flattered by one-off items. FMP data show $GOOGL booked $37.7 billion and $98.0 billion of net non-operating income in the March and June 2026 quarters, and $AMZN booked $53.4 billion in the June 2026 quarter. The FMP consensus for fiscal 2027, which sits below fiscal 2026 for both, puts $GOOGL at 22.5 times and $AMZN at 24.6 times. Smaller distortions also exist: $NVDA had $16.4 billion of net non-operating income in its April 2026 quarter, $MSFT had $10.0 billion in its December 2025 quarter, and $META's September 2025 quarter came in at only $1.05 a share. $CRWV has lost money in every one of its last six reported quarters, and the FMP consensus still has a loss for fiscal 2026.
Two groups stand out. $NVDA, $AMD, $ANET, $MSFT and $QQQ all set their 2026 highs between October 6 and October 13, so for them this is a fresh drawdown. $ORCL, $CRWV, $VRT, $AVGO, $SMCI and $MU peaked in May or June. For those six, October is the second leg of a decline that was already months old.
Prices Fell. Multiples Fell Harder.
This is the part the headline numbers miss. For the names that peaked in the spring, earnings kept arriving after the top. The table compares trailing P/E at each stock's 2026 high, using only quarters filed before that date, with trailing P/E at the October 19, 2026 close.
| Stock | Price change from high | Trailing EPS change | P/E at high | P/E Oct 19 | P/E change |
|---|---|---|---|---|---|
| $MU | -25.4% | +250.3% | 57.3 | 12.2 | -78.7% |
| $SMCI | -36.0% | +72.5% | 26.5 | 9.8 | -62.9% |
| $AVGO | -30.5% | +52.9% | 94.1 | 42.7 | -54.6% |
| $ORCL | -47.8% | +14.5% | 44.6 | 20.3 | -54.4% |
| $VRT | -45.2% | +11.1% | 94.5 | 46.6 | -50.7% |
$MU is the extreme case. Its stock is down a quarter from the June 25 high, yet trailing diluted EPS went from $21.19 to $74.22. The multiple collapsed from 57 to 12, and the FMP consensus for fiscal 2027 prices it at 5.2 times. The math says the market is discounting how long those earnings last, not whether they exist.
$ORCL's trailing figures include a $2.10 quarter in November 2025 that carried $1.6 billion of net non-operating income, according to FMP data. At about 16 times FMP consensus fiscal 2027 EPS, $ORCL trades well below most of the group. The credit market, covered below, is where its price is being argued.
For the October peakers, no new quarter has been reported since the high, so the P/E fell exactly as far as the price. $AMD still trades at 138 times trailing and 70 times FMP consensus fiscal 2026. That is the stretched end of the high-multiple trade.
The Drivers, Dated and Sourced
Rates set the stage. The US Treasury's daily par curve shows the 30-year yield at 4.86% on January 2, 2026, reaching a 2026 high of 5.67% on October 7 and sitting at 5.48% on October 19. The 2-year moved from 3.47% to 4.77% over the same span. $QQQ and $NVDA set their highs on October 6, the day before the long bond peaked. Background: the 30-year Treasury yield breakout.
The twist: long yields eased during the worst week. The 30-year went from 5.56% on October 13 to 5.43% on October 16 while $SMH lost 10.5%. By then the selling was about credit, not just rates.
The hot September CPI, October 14. CNBC reported headline CPI up 3.7% year over year and 0.6% month over month, with gasoline up 9.1%. Per the CME FedWatch figures cited, the odds of a 25-basis-point hike at the October Fed meeting rose to about 37% from 22% the day before. The same report noted Broadcom CEO Hock Tan warning on October 12 that access to capital could become the industry's next constraint. That session, $SMH fell 3.98% and $CRWV fell 15.46% (FMP data). More in our CPI coverage.
CoreWeave credit stress, week of October 12. CNBC reported on October 16 that a CoreWeave 2031 bond issued at 9.75% in April yielded almost 16% that week. Five-year CDS topped 1,000 basis points, 178% above a year earlier, implying roughly 60% odds of default within five years, according to Bloomberg. On October 15, Nvidia CEO Jensen Huang called CoreWeave's financial position stable. CNBC's October 19 credit report put the CDS at about 1,082 basis points. See the CoreWeave credit breakdown.
Oracle's debt insurance, October 19. CNBC reported Oracle's five-year CDS at 295 basis points on the morning of October 19, then 315 basis points later that day, 311% above a year earlier. S&P rates Oracle BBB-, one notch above junk. FactSet data cited by CNBC show $118 billion of long-term debt plus $39 billion of capital and lease obligations, with consensus free cash flow of -$23.7 billion this year.
What the Trend Tools Say
Simple moving averages as of the October 19, 2026 close:
- $QQQ: 1.1% below its 50-day ($726.68) and 6.4% above its 200-day ($675.43). Below the 50-day since October 16.
- $SMH: 5.2% below its 50-day ($578.35) and 7.1% above its 200-day ($512.29). Below the 50-day since October 14.
- $NVDA: 2.6% below its 50-day ($222.94) and 6.9% above its 200-day ($203.12). Below the 50-day since October 14.
For all three, the 50-day average is still above the 200-day: $QQQ by 7.6%, $SMH by 12.9% and $NVDA by 9.8%. None of the three is near a death cross. The spring peakers are another matter. $AVGO, $ORCL, $CRWV and $VRT all have their 50-day below their 200-day, which is the classic long-term downtrend setup.
Breadth tells the same story of rotation rather than collapse. As of October 16, 51.7% of S&P 500 members closed above their 200-day average, up from 40.4% on September 30. Only 45.1% were above their 50-day. Technology stood at 56.5% above the 200-day, behind healthcare's 83.3%. The equal-weight $RSP gained 0.5% from October 6 to October 19, while $SPY lost 1.6% and $SMH lost 13.3%. The method is explained in our 200-day breadth guide.
Bull/Bear Verdict
Bull Case: The long-term trend is intact for $QQQ, $SMH and $NVDA, which are all above their 200-day averages with the 50-day still above the 200-day. Earnings kept growing through the drawdown, so multiples at $MU, $AVGO and $SMCI are now far below their spring levels. The equal-weight $RSP rose during the sell-off week.
Bear Case: This is now a credit story. CoreWeave CDS above 1,000 basis points and Oracle near 315 basis points at a BBB- rating show bond investors are pricing financing risk that equity multiples like $AMD at 70 times forward do not reflect. The 50-day averages have broken, and four of the spring peakers have their 50-day below their 200-day.
FAQ
How far have AI infrastructure stocks fallen in the October 2026 sell-off?
As of the October 19, 2026 close, the 13 names above sit a median 17.1% below their 2026 closing highs. The range runs from -1.8% for $MSFT to -55.7% for $CRWV. In the week of October 9 to October 16, $SMH fell 10.53% and $CRWV fell 25.21%.
Are AI stocks cheaper after the sell-off?
On trailing multiples, many are much cheaper. $MU fell from 57.3 to 12.2 times, and $AVGO from 94.1 to 42.7 times, because earnings rose while the prices fell. $AMD at 138 times trailing and 70.4 times FMP consensus fiscal 2026 remains the most expensive name in the table on both measures.
What caused the AI infrastructure sell-off?
Dated drivers include a 30-year Treasury yield that peaked at 5.67% on October 7, the hot September CPI on October 14 that lifted Fed hike odds, CoreWeave CDS above 1,000 basis points by October 16, and Oracle CDS at 315 basis points on October 19. Back to the Traders On Trend homepage for daily coverage.
Is the Nasdaq 100 in a downtrend?
Short term, yes: $QQQ closed October 19 1.1% below its 50-day average. Long term, no: it remains 6.4% above its 200-day average, and the 50-day is still 7.6% above the 200-day.