Why AI Stocks Are Falling After a $1 Trillion Four-Day Selloff

Four sessions that erased $1 trillion from chip stocks

Global AI and semiconductor stocks lost about $1 trillion in market value over four trading sessions, from October 12 to 15, and Nvidia, SK Hynix and Samsung Electronics led the slide. Seoul Economic Daily reported that the combined value of the world's 20 largest AI and semiconductor stocks fell by that amount, a drop that reached memory makers and AI infrastructure companies alike.

Investors want to know why AI stocks are falling and whether the AI buildout has peaked. Those are different questions, and the reporting so far answers only the first one in part.

Totals also depend on the measuring window. Benzinga counted six memory and storage makers, including Micron, SanDisk and Western Digital, and put their combined losses at about $1.09 trillion since October 2. That window starts before the four-session run, so the figure is larger and covers a different slice of the market.

Which chip makers took the heaviest losses

Nvidia fell 5.3% over the four sessions, shedding about $291 billion in market value. SK Hynix lost about $243.8 billion and Samsung Electronics about $237.3 billion over the same period, according to the Seoul Economic Daily.

Thursday, October 15, added to the damage. Benzinga reported that SK Hynix fell 11.1% in Seoul, its seventh straight loss. Briefs said Samsung fell 8.73% that day. The pain also reached Korea's main index, which the Kyunghyang Shinmun reported plunged more than 7% on October 15.

For US investors, the exposure runs through a handful of familiar names. Nvidia, Micron and SanDisk are direct single-stock channels, and semiconductor exchange-traded funds spread the losses across many holdings at once. The iShares Semiconductor ETF (SOXX) posted an eighth consecutive daily decline on Thursday, with a cumulative drop of roughly 12.7%. That is its longest losing streak in an 11-year history, according to Futunn News.

The figures do not line up neatly because each outlet chose a different window. The table below sorts them out.

The Nvidia row shows the problem. A separate KuCoin News report put the stock's one-day move on Thursday at minus 1.84%. That reflects a shorter window than the four-day total, so the two numbers describe different stretches of trading.

Why are AI stocks falling? Sources point to three causes

Coverage points to three main explanations: rising interest-rate expectations, higher AI financing costs and competition from Chinese memory makers. One report cited the prospect of further Federal Reserve rate increases. None of the sources reviewed confirms a Fed decision, so the rate story rests on market expectations rather than policy action.

Each explanation reaches a different part of the market. Higher rate expectations lower the present value of future earnings and raise borrowing costs, which matters for companies spending heavily on data centers. Financing worries land on the firms that must fund that buildout. Competition from China hits memory makers, whose pricing depends on how quickly rivals can match their products.

The competition argument has a concrete example. Quartz reported that China's ChangXin Memory Technologies recently announced a chip design that fits more data into less space, which raised concerns for memory makers such as SK Hynix and Samsung.

Quartz also quoted the International Monetary Fund, which warned in its financial stability report: "The consequences of an AI correction could be global." Quartz reported that warning on Tuesday. This article did not review the IMF report directly, so the line reflects Quartz's quotation.

Comments from Nvidia's chief executive reportedly steadied the "neocloud" group of AI infrastructure companies for a time, according to Briefs. The relief did not stop the wider decline. China's proposed export controls on AI model weights and source code add a policy layer to that rivalry, as we reported in China proposes export controls on AI model weights and source code.

Why the July rally makes this drop stand out

The scale of the drop makes more sense against the run-up before it. In July, CNBC reported that the Philadelphia semiconductor index had risen 92% over the 12 months before late July, even after a nearly 20% drop over the prior month. Analysts quoted in that coverage described the earlier selloff as a repricing after a strong rally, not a sign of weak AI demand.

The October reports tell a more complicated story. Rates, financing and competition now all appear in the explanations, with no single confirmed trigger. A sector that gained 92% in a year has more room to give back those gains when expectations shift, which is one reason each new leg of losses draws so much attention.

Is the AI bubble popping? What the sources say

The sources do not agree on how far this goes. One report says prices are unlikely to collapse the way they did in past downturns. Others argue that the AI supercycle may be ending. Four sessions of trading cannot settle that debate.

Two signals would help separate a pause from a turn. The first is whether AI infrastructure buyers can keep financing new capacity at a reasonable cost as rate expectations move. The second is whether memory makers can hold their pricing as Chinese competitors ship denser chips. If both hold, the repricing story gains ground. If financing costs keep climbing while memory makers lose pricing power, the peak argument becomes harder to dismiss.

What US investors can check now

This is general information, not investment advice. Several checks are worth doing this weekend.

• Add up your semiconductor exposure. Look for SOXX or other semiconductor ETFs, plus single names such as Nvidia, Micron and SanDisk, across every brokerage account you hold.
• Check your index and target-date funds. Pension and retirement accounts with heavy technology weightings can carry this risk indirectly, so read each fund's top holdings list.
• Measure the drop against your plans. Ask whether a double-digit decline in those holdings would change your timeline, your contributions or your withdrawals.
• Follow the macro calendar. Inflation reports and Federal Reserve communications are the drivers the sources point to most often.

What to watch next

The Federal Reserve is the first thing to watch. The sources describe rate expectations and do not confirm a decision, so the next inflation data and Fed commentary will carry more weight than any single day's chip move. Policymakers are already weighing AI: USA 24 has reported that the Fed's Beige Book found AI restraining hiring in 5 of 12 districts, in Fed Beige Book Finds AI Restraining Hiring in 5 of 12 Districts.

Valuation is the second test. Our coverage of Anthropic's IPO valuation talks, which fell to $1.5 trillion as growth slowed, shows the same question playing out in private markets.

The next signal will come from financing deals for AI data centers and from the next round of rate data. If the SOXX losing streak ends and memory makers hold their pricing, this selloff will look like a repricing. If the streak runs on while financing costs climb, the case for a peak gets stronger.