Advanced Micro Devices (AMD) fell about 6% to $519.50 on October 22, 2026, leaving the stock 20% below its October 6 peak of $649.42 and oversold, with a daily RSI near 25. The weekly uptrend is still intact, as price remains above its 40-week average. The balance of evidence leans positive for the next two to six weeks, with high conviction, though the shares have yet to reclaim the 50-day average at $541.84.
Why it matters
The last company news is from late September. AMD agreed to buy World Labs, Fei-Fei Li's spatial-intelligence startup, for $8.2 billion in stock, with closing expected by year-end and Li joining as chief scientist. On September 30, HPE won a $1.2 billion Helios order from Vultr, a first commercial deployment of AMD's rack architecture. Both items landed after a huge run. AMD had briefly passed a $1 trillion market value after a nearly 13% rally, and the shares are still up 142.6% this year and 71.2% over six months. Good news was already in the price by the October 6 peak.
Nothing in the news we have explains the slide since then or today's drop; the newest article is dated September 30. The tape does show where the pressure sits. Over five days AMD is down 7.6%, against 0.7% for the S&P 500 and 4.8% for the semiconductor ETF SOXX. The market explains almost none of it, the sector explains most, and the stock lost a further 2.8 points beyond SOXX. Today the pattern repeats: AMD -6.0%, SOXX -4.0%, SPY -0.8%. This looks like a chip-sector sell-off that AMD is amplifying, not a company-specific break. Over 50 days AMD still leads SOXX by 5.1% while lagging the S&P 500 by 3.9%.
The argument about AMD is about how much of the AI build-out it can capture and at what price. The bulls point to demand: Jim Cramer says the run isn't over, citing agentic AI, and Morningstar argues the real prize is recruiting Li. The skeptics start with the currency. Reuters Breakingviews notes AMD is using its high-priced shares to pay for a robotics bet, in a market JPMorgan sizes at $2.5 trillion by 2035 but where chips and software are only about 6% of a robot's bill of materials. Then there is the competitor: Forbes puts Nvidia's share of AI accelerator revenue at 75% to 81%.
Valuation is where the two camps meet. Fifty analysts rate the stock a strong buy, and the mean price target of $637.61 sits 18.5% above the current price. The targets range from $365 to $1,250, though, which shows how little agreement there is on what AMD's share of AI spending will be worth.
What the evidence says
The long-term thesis is intact. Price is above its 40-week average and far above the 200-day average at $400.61, and it is 172% above the 52-week low of $190.95. Against the S&P 500, AMD leads by 24.9% over 200 days. The stock is in a correction phase inside that advance: it has fallen below both its 10-week average and the 50-day average, and has given back a fifth of its peak in about two weeks.
For the next two to six weeks the balance of evidence leans positive, with high conviction: three of four points line up. A pullback inside a weekly uptrend has reached oversold. AMD is leading SOXX over 50 days. History after today's events also favours the trend side. After one-day drops of more than 5%, the stock was higher a month later in 70.9% of 151 cases, against 61.8% typically, with a median gain of 5.38%. The oversold RSI signal alone adds little, since 61.4% of 303 cases were higher a month later, in line with the typical 61.8%. The edge comes from the sharp drop.
The strongest case against is the one point that does not line up. AMD trails the S&P 500 by 3.9% over 50 days, which runs against the trend, and SOXX is down 14.5% in a month with no sign yet that the selling has ended. A stock that has lost 16.7% in a month can keep falling before it turns. That is why conviction is not a full four of four. It does not tip the balance because the stock's weakness is mostly shared with its sector, which it still leads, and because the long-term trend has not been touched.
Levels that matter
The structure holds above $500, a round number sitting just over the swing zone at $494.87. A break under $500 would be the first crack. The trend ends on a weekly close below $461.71, a swing zone; under it the next supports are $450 and $438.87. Today's drop has put price within reach of the first of those lines.
The upside map starts at $541.84, the 50-day average. A daily close above it on volume above its 20-day average would be the first sign that buyers are back; volume so far today is 70.4% below that average, though the session is young. Above it sit $550, then $562.23 (swing zone), $600 and the $649.42 high. From $541.84 against a stop at $493.00, which is $500 minus a quarter of a day's average range, the first target gives 0.4 to 1 and the 52-week high gives 2.2 to 1. The math does not reach 3 to 1. It would take a target beyond the 52-week high, meaning fresh highs, or a stop much closer to entry, to get there. A breakout would be wrong below $493.
What would change the picture
A weekly close below $461.71 would turn this from a pullback into a broken trend. The non-price fact most likely to do it is the earnings report on November 3: a data-center AI outlook that disappoints, with Helios demand slower than the HPE and Vultr order implies, would make that level far more likely to give way.
What to watch
- Earnings, November 3, 2026: the number that matters is the data-center AI revenue guidance for the next quarter, since the whole debate is about how fast that business scales.
- The sector: SOXX fell 4.0% today and 14.5% over a month, and it is moving AMD before AMD's own news does. A stabilizing SOXX would matter more than any AMD headline.
- World Labs: the all-stock deal is expected to close by year-end, so the share count and the stock price both feed into its cost.
- Valuation: at $519.50 the stock is 18.5% below the $637.61 mean target, inside a $365 to $1,250 range of analyst opinion.