Our numbers can't tell you whether a bubble is popping. They can show how far the sell-off moved AI stocks against their own history, and Friday's part of it was smaller than the headlines suggest: Intel fell 29% on October 30, and applying that drop to its Thursday reading moves its EV/Sales rank only from the 90th to about the 89th percentile of its own five-year range. The bigger change came over the month. Of 45 AI chip, memory and power stocks we rank, 22 sat in their historically high band at the October 2 close and 6 did at the October 29 close.
One thing before the numbers. Our weekly readings stop at Thursday's close, October 29, 2026; we computed them the morning of October 30, before Friday's open. Friday's prices enter the next weekly update. Where this piece uses a Friday move, it says so and shows the arithmetic.
What happened on Friday
The Nasdaq Composite fell 2.9% on October 30, its largest one-day percentage drop since June 5 (Reuters, CNBC). Two earnings reports set it off. Intel forecast fourth-quarter revenue of $15.0 billion to $16.0 billion and said it sees the PC slowdown continuing at least through the first half of next year (Intel). Analysts had expected $17.1 billion, and the shares fell 29%, their worst day since the 1970s (CNBC; Reuters put the drop at 28.6%). Amazon kept its 2026 capital spending plan at about $220 billion and reported negative free cash flow for a second straight quarter (Amazon, CNBC).
One bad day barely moves a five-year rank
A five-year percentile ranks this week's multiple against 261 weekly readings. To fall a band, a stock has to drop below a lot of past weeks at once, and one day rarely does that.
Intel shows it. At Thursday's close of $92.07, its EV/Sales was 8.1, higher than 90% of its weekly readings over the last five years: historically high, right at the edge. Take 29% off that price, keep revenue, share count and net debt where they are, and EV/Sales falls to about 5.7. That's still higher than 89% of its weekly readings, which moves it to above its usual range. Every week with a higher reading falls between mid-April 2026 and now. From November 2021 through the end of 2025, Intel's EV/Sales never went above 4.5, and its five-year median is 2.9.
So a 29% fall, Intel's worst day since the 1970s, shifts its rank by one point. The rank only sees how far the stock still is above its own past, and after a run that more than doubled the price this year, Intel is still far above it.
The takeaway: A single day, even a brutal one, moves a five-year rank by a point or two unless the stock was sitting right on a band line. Intel was, so it slips from historically high to above its usual range on our arithmetic. The official reading updates next week.
Which AI stocks got cheaper in October?
Measured against their own history, many of them, and well before Friday. Between the October 2 and October 29 closes, the median stock among the 45 fell 17% and its EV/Sales rank fell 8 points. Sixteen left the historically high band:
| Company | Theme | 5y rank, Oct 2 | 5y rank, Oct 29 | Band, Oct 29 |
|---|---|---|---|---|
| Hubbell | AI power | 95th | 64th | Within its usual range |
| Eaton | AI power | 96th | 75th | Above its usual range |
| Arm | AI infrastructure | 92nd | 77th | Above its usual range |
| nVent | AI power | 94th | 77th | Above its usual range |
| Valmont | AI power | 91st | 81st | Above its usual range |
| Micron | AI memory | 96th | 82nd | Above its usual range |
| Lam Research | AI infrastructure | 98th | 84th | Above its usual range |
| Corning | AI infrastructure | 93rd | 85th | Above its usual range |
| KLA | AI infrastructure | 97th | 86th | Above its usual range |
| Coherent | AI infrastructure | 94th | 86th | Above its usual range |
| Powell Industries | AI power | 91st | 86th | Above its usual range |
| Lumentum | AI infrastructure | 90th | 87th | Above its usual range |
| Arista | AI infrastructure | 99th | 88th | Above its usual range |
| Bloom Energy | AI power | 96th | 88th | Above its usual range |
| Applied Materials | AI infrastructure | 97th | 89th | Above its usual range |
| Seagate | AI memory | 93rd | 89th | Above its usual range |
Where these stand now
Current readings update separately from the article’s dated analysis.
Above its usual range
INTC's EV/Sales of 5.3x is higher than 88% of its weekly readings over the last five years.
Historically high
AMD's EV/Sales of 19.2x is higher than 92% of its weekly readings over the last five years.
Above its usual range
MRVL's EV/Sales of 16.9x is higher than 83% of its weekly readings since March 2022.
Four names fell to the bottom of their ranges: Super Micro (25th to 5th), Applied Digital (74th to 7th), CoreWeave (21st to 4th) and Celestica (40th to 1st), all historically low now. CoreWeave's history only starts in March 2026 and Celestica's in October 2025, so their ranks rest on less than a year of readings. Nvidia moved from the 20th to the 15th percentile and stays below its usual range.
The takeaway: By Thursday's close, most of the AI trade had already come off the top of its own history. Fifteen of the sixteen names that left the high band now sit above their usual range, not in the middle of it.
Still historically high at Thursday's close
Six names stayed in the top band. The last column is arithmetic, not a forecast: how far Thursday's price would have to fall, with revenue and debt unchanged, for the rank to drop below the 90th percentile.
| Company | Theme | EV/Sales, Oct 29 | 5y rank, Oct 2 | 5y rank, Oct 29 | Price fall to leave the band |
|---|---|---|---|---|---|
| Everpure | AI memory | 11.0x | 100th | 99th | about 31% |
| NetApp | AI memory | 6.6x | 100th | 99th | about 35% |
| AMD | AI infrastructure | 22.5x | 100th | 97th | about 28% |
| Dell | AI infrastructure | 2.2x | 98th | 94th | about 35% |
| Marvell | AI infrastructure | 21.4x | 98th | 93rd | about 9% |
| Intel | AI infrastructure | 8.1x | 97th | 90th | about 1% |
The takeaway: The storage makers Everpure and NetApp barely moved in October and sit at the 99th percentile. AMD, Dell and Marvell would need large further falls to leave the band. Intel needed about 1%, and Friday delivered far more.
What this doesn't tell you
- Whether the stocks fall further. A rank describes where a multiple sits against its own past. It doesn't forecast anything, and a low rank after fast growth can also mean investors doubt the growth will last.
- Whether spending pays off. EV/Sales ignores margins and cash flow, which is exactly what Friday's worry was about. Check P/E and EV/EBITDA on each stock page too.
- What to buy. We don't make that call. Nothing here is a recommendation.
Check the live numbers
Each theme page ranks every member against its own five-year range and refreshes as new weekly readings arrive: AI infrastructure stocks ranked by valuation percentile, AI memory stocks and AI power stocks. For the names above, open Intel's EV/Sales history, AMD's EV/Sales history, Marvell's EV/Sales history, Dell's EV/Sales history, NetApp's EV/Sales history and Everpure's EV/Sales history. How we ranked the whole chain on October 5 is in AI infrastructure stocks against their own valuation range, and what a valuation percentile is explains the method.
Data: SEC EDGAR XBRL filings and weekly closing prices, computed by StockResearch on October 30, 2026 from closes through October 29, 2026. Friday, October 30 prices are not in these readings. October 2 ranks use the five years of weekly readings available on that date. The Intel Friday estimate applies the reported 29% drop to Thursday's price with revenue, share count and net debt unchanged. Price falls needed to leave a band use the same assumption. Ranked members exclude foreign filers and names under data review. Alex leads research at StockResearch and writes the monthly valuation report. Nothing here is investment advice or a recommendation to buy or sell any security. Do your own research.