Brutal Semiconductor Selloff Traces Its Roots to This Trend
Are we at capitulation point on semiconductor stocks? Here’s why I feel it’s time to be brave.
You've reached your free article limit
You've read 0 of 1 free Pro articles.

If risk sentiment recovered on Wednesday, it’s well off the table on Thursday. Risk may not even be in the same room as the table anymore …
Major tech stocks are down hard Thursday, particularly semiconductors, feeding into pessimism surrounding the entire Artificial Intelligence (AI) sector. Asian selling set a trend that’s continuing on Wall Street.
There’s been plenty of pessimism surrounding the semiconductor sector for months now. There’s a different negative explanation every day. The hyperscalers are overspending on capex, or they’re not spending enough, or chip prices are so high they’re stifling PC and mobile phone sales, meaning chip revenues are likely to fall.
Sometimes the arguments are circular — hyperscaler capex suggests high spending on semiconductors, but we often see both sides of the equation sell off. High chip prices are good for chipmakers even if they crimp company and consumer spending on electronics. But slowing computer sales because of high chip prices … sell the chipmakers again!
What has produced today’s carnage?
Don’t get too caught up in any midterm-election effect. Yes, Democratic lawmakers may push back on new data-center development. At the same time, no politician wants to be seen to hold AI innovation back, and watch China steal the advantage. I don’t think we can predict the impact of the political shift with any confidence.
As an aside — I was correct in my last column and on the Money Talk podcast in predicting that U.S. President Donald Trump would not contest losses in the mid-terms in face of a “blue wave,” and would instead wash his hands of any “losers.” He’s been surprisingly quiet, and appears willing to work with a Democrat-led Congress.
Instead of Washington ins and outs, I’m watching technology innovations, and earnings, instead of reading too much into election results. Investors were not encouraged by results after Wednesday’s bell from Qualcomm (QCOM). While its relatively new data-center segment posted encouraging results, Qualcomm missed Wall Street estimates on smartphone chip sales.
Its chip unit saw sales shrink 26%, dragging overall chip sales down 8% to $9.0 billion for the company’s Q4, through September 27. QCOM is down 10.2% in early going on Thursday.
But I think we have an even more important catalyst out of Asia. It is well worth watching, and appears to have started Thursday’s selling.
Tracing Back to CXMT
ChangXin Memory Technologies (SH:688825) — better known as CXMT — on Friday reported its first set of earnings as a public company, after listing in late July. The company turned a C¥54 billion ($8.1 billion) profit for the quarter, and converted a C¥5.3 billion loss attributable to shareholders for the first nine months of last year to a C¥132.0 billion profit.
That’s all well and good. We have seen strong earnings from Asian chipmakers in this last round of earnings fail to produce a stock response.
What’s of more interest to me is that CXMT is indicating that it has completed sample validation for its sixth-gen DRAM chips, using the 4F² architecture. I discussed this in mid-October, when we first got word that CXMT was perfecting the 4F² technology, something other Asian chipmakers are also chasing.
CXMT over the weekend, and after its earnings, told Chinese media including the Global Times that its product validation means it can move toward mass production, if its customers also qualify its chips. It aims to introduce chips leveraging the 4F² architecture by the end of the year.
That led to punishing moves Thursday in the shares of SK Hynix (KR:000660) (SKHY), down a whopping 9.5% in Seoul. Its U.S. shares are faring even worse, down 12.0% at the open.
Korean Chipmakers Chasing Same Tech
Samsung Electronics saw its Seoul shares sink 7.6%. Both Hynix and Samsung are chasing 4F² tech, with the current DRAM standard being a 6F² cell structure. The new technology would allow for a denser concentration of cells on a semiconductor.
CXMT has already gained significant market share at the lower end of the chip market. It claimed 10% of the DRAM market for Q2, according to data from Counterpoint. and although that slipped to 9% in Q3, the company would surely continue to win significant business inside China in particular with its innovations in chip efficiency.
Micron Technology (MU), meanwhile, said in October that it is abandoning its efforts to implement the 4F² standard. President and chief technology and products officer Scott DeBoer indicates that the advantage of 4F² is “a bit of a myth,” with how close capacitors can be packed holding back density, not just the footprint of the cell.
Micron will stick with 6F² technology. While it’s possible to build the 4F² architecture, it comes at a higher cost and complexity, DeBoer says.
That said, I trace the selling to CXMT’s advances. When CXMT first discussed this advance in October, I noted that we should be seeing increases in its share price, if that’s the competitor likely to win out as a result. But we had not seen the stock move.
Now we do. CXMT shares raced ahead 13.6% on Wednesday. They have given back 3.5% Thursday, but we have finally seen the company find renewed strength after its incredible 466% first-day pop.
What isn’t yet clear is whether CXMT can commercialize its technology, at affordable cost. It’s also not yet clear whether CXMT will be selling its chips beyond China — Beijing has been tightening its rules on the export of sensitive technology, particularly products that can contribute to AI.
But we do have CXMT promising to bring new chip technology to market, while I also indicated on Tuesday that we had troubling indications in the earnings of Japanese chipmaker Kioxia Holdings (T:285A) this week.
Are We at Capitulation Point Yet?
In Tokyo, Kioxia is down another 13.1% Thursday, meaning it has lost half its value in the last month.
The former Toshiba Memory chip unit indicates that NAND pricing is slowing down, with a 70% price increase in the June quarter slowing to the mid-20% range in the September quarter. Its forecast for the current quarter missed estimates as a result.
The declines for chip stocks are far worse if looking back to their June peaks. Kioxia, briefly Japan’s biggest company by market value, is now down 74.3% from its June 22 all-time high close.
Even Taiwan Semiconductor Manufacturing Co. (TSM) (TW:2330), typically a steadying influence in the chip sector, is plunging 7.6% in early going on Wall Street Thursday. The chip-foundry market leader has seen its shares sell off 24.7% since October 5.
Terrible Times for Memory
I’ve watched my largest holding, the Roundhill Memory ETF (DRAM), take an absolute hammering. It has found support around the $50 mark in recent months, but has broken well below that and even under $40, with a 14.3% decline Thursday. Its trading at $38.67 as I write.
That takes the ETF back to its levels from April, the month that it launched. The ETF launched at $27 on April 2, but quickly accelerated to $40.41 by May 1, as it became the fastest-growing new index fund, hitting $6 billion in assets within a month. It then topped $80 in June.
So it has doubled from spring into summer, then halved from summer through fall. And what a fall …
I’m continuing to catch this falling knife, but it is scary. My holding is averaging just under $50 in terms of purchase price, but after Thursday’s bruising that’s producing an 18.8% decline.
Call Me Foolish
The semiconductor sector did not participate in Wednesday’s post-election Wall Street rally. The Philadelphia Semiconductor Index (SOX) fell 3.1% on Wednesday, and is down a whopping 10.1% Thursday.
That takes the total decline for the SOX to 30.3% since October 6. It’s down 37.2% since its June peak.
I think this is the time to be brave. I’ve added to DRAM under $40, and I’m also adding to my position in Nvidia (NVDA), with those shares down 6.4% today and back to the levels last seen in April.
I exited much of my Nvidia position in April 2025, shortly after the “Liberation Day” tariffs were put in place. My sales price around $94 proceeded to look very foolish in retrospect.
At the time, I was convinced that Trump’s tariffs were going to do lasting damage to the U.S. and global economy. The tariffs have quite rightly been ruled illegal taxes, and they did not hold back the U.S. economy as much as I thought they might.
Nvidia of course continued to benefit from the memory boom, proceeding to come close to tripling its share price over the course of the next year. I feel we will soon hit another turning point from which chip designers and chipmakers can rally — provided they’re a market leader, and making a product that’s in demand.
This is set to be the worst day for semiconductors since Liberation Day. What continues to encourage me about the semiconductor sector in general is that most of its production has been promised to customers over the course of 2027, well into 2028. Hynix is saying its capacity will remain constrained for the rest of this decade.
I plan to hold my semiconductor stocks for all of next year as a result. This appears a strong buying opportunity for investors who can hold out while the selling reaches its panicky peak. The selling may not be done yet, but it may be time to take on risk when it scares you most.
At the time of publication, McMillan was long TSM and DRAM.
