원문 보기
Chip stocks just had their worst week in over a year. The Philadelphia Semiconductor Index fell 11.4% in five sessions. And the headlines say Wall Street's two biggest banks are on opposite sides of the trade. Goldman Sachs sees another 15% drop. Morgan Stanley says buy the dip. Here is the problem.
Neither headline is quite true and the truth is more useful. This video what the banks actually said. What actually happened in the sell-off and SOXX versus SMH head-to-head? Because after a drop like this, the question is not just buy or wait. It is also which one. Start with Goldman. The another 15% number is real, but it is not a semiconductor call and it is not a forecast.
Goldman strategist Louis Miller pointed at a divergence. Credit spreads on the big AI hyperscalers are at their widest this year, but their share prices have not fallen to match. The last time spreads were at these levels, those stocks traded about 15% lower than they do now. So either the bond market is too pessimistic or AI linked equities carry roughly 15% of catchdown risk if spreads keep widening conditional AI stocks broadly not a chip sector target and Goldman is not one voice.
Its own semiconductor analysts wrote on October 5th that riskreward had turned more favorable after the sector's pullback. On Monday, Goldman upgraded Coreweave, an AI infrastructure stock down a quarter in a week, calling the pullback a buying opportunity. The bank's only specifically bearish chip call was a hedge on Taiwan semiconductor exposure, where positioning sat at record highs.
Now, Morgan Stanley, there is no Morgan Stanley note telling clients to buy the semiconductor dip. What exists is a broad market call. In early and midocctober, Wilson's team said the pullback had created a better entry point for quality stocks with rising earnings estimates. Quality stocks, not chips.
The bank restored Nvidia as a top pick on October 2nd and defended two storage names. And back in June, Wilson's team warned that semiconductor momentum might be topping. constructive on the market, selective on names, cautious on chips at the top. The sell-off itself, all prices as of Monday's close October 19th.
The chip index fell 11.4% last week, its worst week since April 2025 after eight straight down sessions. SOXX fell 11.8% on the week, down 24.9% from its peak. SMH fell 10.5% on the week, roughly 20% off its high. and the context. Even after all that, SOXX is still up about 67% this year. What drove it?
Three things. First, a hot September inflation print. CPI up 3.7% year-over-year. Revived bets on rate hikes. Second, credit fears around AI financing. Coreweave lost about a quarter of its value in a week. Third, positioning. Hedge funds sold tech at the fastest pace since 2016. Semiconductors were 57% of it.
Goldman's framing, profit taking on winners, not abandonment of the AI theme. TSMC beat earnings and raised its guidance and its stock still fell that day. That is what crowded positioning does. On Monday, both funds bounced. SOXX up 1.7%. SMH up 1.6%. One green day after eight red ones is a bounce.
It is not a trend. So, SOXX versus SMH, same sector 0.97 correlation. But they are built differently. Then the difference is one stock. SMH is a leader bet. Nvidia is 20.5% of the fund. Add TSMC at 9.4% and two companies are roughly 30% of your money. The top 10 are 68% of the fund. SOXX is the capped broader basket.
Its index caps single names, so Nvidia is only 8.0%. Third place, AMD leads at 9.3%. Then Intel at 8.3% and Micron at 7.6%. More second tier, more memory, 34 holdings in all. fees are a wash. 0.33% for SOXX, 0.35% for SMH. Over the last year, SOXX1, up about 75% versus 61% for SMH because the 2026 rally broadened beyond Nvidia into memory and equipment.
Over 3, 5, and 10 years, SMH1 about 57, 33, and 33% a year because Nvidia's mega run dominates the long windows. Same sector, different engine. Now, valuation, the part of this story that genuinely changed. After the sell-off, both funds trade at roughly 18 times forward earnings. 18.3 for SOX, 18.1 for SMH.
The S&P 500 trades at about 19 times forward. So, for the first time in this cycle, chip ETFs carry no forward premium at all. But the word forward is doing heavy lifting. The trailing multiple is still about 32 times, and the forward number is only cheap if the estimates survive. Goldman's own model says hyperscaler spending growth peaks around now at about posing 116% year-over-year then slows to roughly 20% by the end of 2027 with its earnings contribution turning negative by 2028 if estimates get cut 18 times gets expensive again fast meta reports tomorrow October 21st Microsoft and Alphabet on the 28th Intel on the 29th AMD on November 3rd and Nvidia the biggest print of in mid- November.
So, a buy, hold or wait, it depends on who you are already holding. So, XXX or SMH. This is a sizing review, not a panic. The AI spending thesis is intact. TSMC's numbers were strong, but these are roughly 1.8 beta funds. If the run quietly made semiconductors a fifth off your portfolio, that is a sizing question, not a sell signal.
New money with a long horizon. This is a legitimate buy the dip candidate. Staged not allin. down 20 to 25% from the highs. Forward multiples back at market level. Goldman's conditional 15% scenario is exactly why you buy in trenches instead of going all in on the first green day. There is no prize for being first.
Need the money soon or need certainty? Wait, one bounce is not a trend. Credit spreads are still wide and the biggest catalyst has not printed yet. Trading it following knife rules. Defined risk small size and one warning. The leveraged version, SOXL, took in a record 7 billion dollars during the worst week while falling about 35% in seven sessions.
Leveraged ETFs reset daily. You can be right about direction and still lose money. That is a trading tool, not an investment. An SOX or SMH, that is not a quality question. It is a concentration question. Want a governor on single stock risk? SOXX. Believe the leaders keep taking share SMH. and accept that one Nvidia earnings report moves your whole fund.
Final caveats, uh, this story is live. The prices and bank views in this video are as of October 20th, 2026, and this market moves daily. A 25% draw down can become a bigger one. Nothing here says the bottom is in. Past performance does not guarantee future results. This video is educational content only.
It is not financial advice and it is not a recommendation to buy or sell any security. The headline said, "Goldman versus Morgan Stanley, bears versus bulls." The research says something narrower and more useful. Credit markets are nervous, earnings have to prove the forward numbers, and your real decisions are position size and concentration. Get those two right and the headline takes care of itself.
Comments 0