A trader studies falling market charts as the 2026 AI selloff deepens. (Image: Pexels)
There is a strange paradox at the heart of this week's market chaos. Artificial intelligence has never been bigger — companies are spending more on AI infrastructure than at any point in history — and yet the stocks most closely tied to AI are being sold off with a violence that has traders talking about the worst stretches since 2008.
On Tuesday, October 20, South Korea's KOSPI fell 3.73% to 5,552.74, its lowest close in more than six months. Japan's Nikkei dropped 1.3%. Bitcoin slipped to around $81,400. Even gold, the traditional safe haven, fell nearly 1%. This is not a one-day event. It is the latest act in a selloff that has been building for months — and it is asking one uncomfortable question of every investor: what if the money being poured into AI never produces the returns being priced in?
What happened on Tuesday: the damage in numbers
Start with Seoul, because the Korean market is the purest expression of this trade. Samsung Electronics and SK Hynix — the two memory-chip giants at the base of the global AI supply chain — account for roughly 44% of the KOSPI. When chip stocks fall, the entire Korean market falls with them.
On Tuesday, Samsung fell 2.98% to 211,500 won. SK Hynix lost 4.23% to 1.22 million won, possibly its lowest close since April. Of 915 traded issues, 773 declined and only 118 advanced. Foreign investors sold a net 530 billion won ($400 million), institutions sold a net 1 trillion won, and only individual retail investors were buying.
The theme ran wider than Seoul. Japan's Nikkei 225 fell 1.3% to 64,075.42, with chip suppliers Kioxia and TDK among the losers. Taiwan's Taiex lost 1.13%. One of the day's ugliest scenes belonged to Elice Group, a Korean AI data-center company, whose market debut opened 16% below its offer price and fell as much as 33% — an IPO disaster that captures the current appetite for AI risk perfectly.
Bitcoin dipped to around $81,400, according to CoinDesk, even though US spot Bitcoin ETFs pulled in $526.6 million on Monday — their largest daily inflow in nearly a month — with puts outtrading calls more than two to one on the Deribit options exchange. Not everything fell: Brent crude climbed another 2% toward $111 on Houthi supply risks, and long-dated US Treasuries attracted defensive flows, with the TLT bond ETF up 0.5%.
This is not a one-day dip
Tuesday looks bad. The trend behind it looks worse. At its June record close of 9,114.55, the KOSPI was up roughly 116% on the year. It is now about 39% below that peak, and its 2026 gain has shrunk to around 31%.
The slide gathered pace in July, when the index lost 22%, and October erased another 17% through October 16. The KOSPI fell seven straight sessions through October 15 — then stronger-than-expected US inflation data hit, and it plunged 7.57% in a single day. Between October 1 and October 16 alone, Samsung fell 19.74% and SK Hynix fell 28.38%, according to the JoongAng newspaper.
Why the selling won't stop: four forces behind the AI selloff
Markets rarely fall for one reason. This one is falling for four, all at once.
1. The AI credit scare. The spark of the current phase came from the bond market. Credit worries around AI spending grew after bond prices fell at cloud provider CoreWeave, whose debt now carries yields of roughly 15% — a level that signals serious stress. When the company building the AI infrastructure can't borrow cheaply, investors question whether anyone can afford this build-out. As Daishin Securities analyst Lee Kyoung-min put it this week: "The sustainability of the recent rebound was limited as credit and profitability concerns around AI investment remained not fully resolved."
2. The rate squeeze. September's US CPI came in at 3.7% — hot enough to end remaining hopes of near-term relief. The Fed's next meeting is October 27–28, with markets in a blackout period and no guidance from dovish-but-noncommittal Chair Warsh. Fed funds sit at 3.75–4%, with futures pricing a 92% chance of a hike by December. High borrowing costs compress growth-stock valuations, and no sector's valuations depended more on cheap money than AI. Korea's own central bank is expected to hold rates on Thursday after two straight hikes, with a Reuters poll forecasting one more increase next month.
3. Concentration. The KOSPI's vulnerability is structural. With Samsung and SK Hynix at 44% of the index, Korea's market is a leveraged bet on memory chips — which is a leveraged bet on AI data-center spending. When that single theme turns, there is nowhere to hide inside the index.
4. Contagion into every risk asset. The most striking feature of this selloff is how far it has spread. Bitcoin's drop despite record ETF inflows shows that "institutional adoption" has turned it into an asset that sells off with tech rather than against it. Gold's dip suggests even defensive positions are being liquidated to cover losses elsewhere. This is classic de-risking: when fear takes over, correlations go to one.
My read: the market is pricing a capex cycle, not an AI funeral
Here is the distinction most commentary is missing. Nothing about the technology has changed. AI models keep improving and adoption keeps rising. What changed is the market's willingness to pay 2021-style multiples for 2026-style uncertainty about who actually profits.
Every industrial revolution has a phase where the picks-and-shovels builders raise enormous sums and the market briefly stops believing the gold mine is real. Railroads went through it. Telecom went through it in the late 1990s. In both cases, the infrastructure got built anyway, the debt holders got hurt, and the winners emerged one layer up the stack. CoreWeave's 15% bond yields and Elice Group's 33% IPO collapse are the bond and equity markets saying the same thing: show us the revenue, not the roadmap.
That is also why the pain is concentrated in the infrastructure layer while the broader market holds up. The Nasdaq rose 1.11% to 26,430.886 on Monday, and the Philadelphia Semiconductor Index snapped an eight-session losing streak — its longest since October 2008. Monday's bounce is what made Tuesday's Asian reversal so jarring: it suggests the relief was temporary, not a bottom.
And then there is Bitcoin. A $526.6 million ETF inflow day that still ends with the price falling is a signal, not noise. It tells you the marginal buyer has shifted from conviction to momentum: institutions are allocating to an asset they now treat as high-beta tech, hedging it with puts at a two-to-one ratio. The old dream of Bitcoin as uncorrelated "digital gold" is having its hardest test yet — and it is failing it.
The other side: why the bulls are not dead yet
It would be a mistake to write the obituary of the AI trade. Beneath the red screens, several developments point the other way.
JPMorgan has included Bitcoin in its long-term capital market forecast for the first time, projecting it could outperform US stocks and gold over the next decade — while warning that volatility and unstable correlations still limit its role in standard portfolios. Switzerland's FINMA has approved the first retail-accessible Bitcoin ETF, citing Bitcoin's "long trading history and particularly high trading volume." BlackRock has filed for an iShares Bitcoin Managed Protection ETF that would use options on its existing IBIT fund to cap investor losses — and IBIT alone drew $665 million in a single day this week.
And earnings season is here: roughly 70 S&P 500 companies report this week, with General Motors and United Airlines among Tuesday's reporters. Strong results from the companies actually spending on AI would stabilize the trade faster than any analyst note.
What to watch next
Five things decide where this goes. First, the Fed's October 27–28 meeting — any hint of higher-for-longer deepens the valuation squeeze; any dovish surprise triggers a sharp relief rally. Second, big-tech earnings: markets will watch capital-expenditure guidance, not revenue — whether the giants keep spending or start to blink. Third, credit markets: watch CoreWeave's bonds and any new AI-infrastructure issuance. Fourth, the Bank of Korea's rate decision on Thursday. Fifth, politics — France's escalating protests and the Middle East tensions keeping oil near $111 all feed the risk-off bid.
Frequently Asked Questions
Why are AI stocks falling when AI itself is booming?
Because the market is no longer pricing the technology — it is pricing the returns on the money being spent to build it. Hundreds of billions are going into data centers and chips, financed increasingly with debt, and investors are asking when that spending turns into profits. Add hot inflation, higher-for-longer rates, and credit stress at names like CoreWeave, and the most expensive stocks fall the hardest.
Is this a market crash or a correction?
By the numbers, it is a severe correction in a specific theme — not a broad market crash. The KOSPI is down about 39% from its June record, a bear market for that index, but the S&P 500 and Nasdaq are holding up far better. A crash would require the selling to infect the whole market. Watch the US indices and credit markets: if they crack, it becomes something bigger.
Why is Bitcoin falling even though ETFs are buying?
Two reasons. First, the marginal ETF buyer is an institution that treats Bitcoin as high-risk tech, not digital gold — so it sells off with stocks. Second, options data shows heavy hedging: puts outtrade calls more than two to one on Deribit. Record inflows with a falling price tell you the selling is coming from holders with bigger positions, not from a lack of demand.
Should I buy the dip in chip stocks?
The framework is this: the stocks are cheaper because the risks are higher, not because the risks are gone. Samsung and SK Hynix are cyclical businesses riding a structural wave. If AI data-center demand holds through 2027, current prices may look attractive in hindsight. If the capex cycle is peaking, this is a repricing, not a dip. Wait for earnings guidance and credit-market stabilization before deciding. (This is not financial advice.)
When does the selling end?
Historically, theme-driven selloffs end when valuations reset to levels value buyers defend, when the fundamental question gets answered (tech earnings showing AI spending still growing), or when central banks turn dovish. The next real decision points are the Fed's October 27–28 meeting and the coming wave of big-tech earnings. Until then, expect volatility.
The bottom line
The 2026 AI selloff is the market's way of asking the question it avoided for two years: who actually gets rich from all this spending? The chipmakers are being revalued as cyclical companies again. Bitcoin is discovering that institutional adoption cuts both ways. And everyone is waiting on the Fed.
None of this means AI is a bubble bursting. It means the trade grew up — from pricing perfection to pricing reality. The infrastructure is still being built, the debt is still being issued, and the models are still improving. But the easy money — owning anything with "AI" in the pitch deck — is over. What comes next depends on earnings, rates, and whether the revenue ever catches up to the rhetoric.
Watch the Fed meeting on October 27–28. Watch the tech earnings. And watch the credit markets — they called this turn before the equity screens did.
For more in-depth analysis of global markets, energy, and technology, visit abdulrahimmughal.com.