Tech Stocks Tumble on AI Jitters as Kospi Sinks 3.73% [2026]

Global technology stocks took a hard hit on October 20, 2026, as South Korea’s Kospi Index sank 3.73% to close at 5,552.74 points, erasing two days of gains and dragging down chip bellwether SK Hynix as investors grew uneasy about the payoff on the AI infrastructure boom. The selloff rippled across Asia and into currency markets, with the U.S. dollar slipping against most major peers just as Wall Street braces for a crowded earnings week headlined by Meta Platforms and Tesla.

The pullback is less a single shock than a pressure test. After weeks of record highs built on AI capital expenditure, traders are now asking whether the spending can actually generate returns, and whether chipmakers, cloud platforms and model developers can keep justifying valuations that assume years of uninterrupted growth. Tuesday’s price action in Seoul gave an early answer: not everyone is convinced yet.

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

Kospi’s 3.73% Drop Leads the Regional Selloff

South Korea’s benchmark Kospi Index fell 215.27 points, or 3.73%, to close at 5,552.74 on October 20, after touching an intraday low of 5,493.24. The decline wiped out a two-session rebound and came on top of a brutal stretch in early October, when the index had already dropped a combined 20.7% over seven trading sessions between October 6 and October 15 before clawing back some ground. Tuesday’s reversal shows just how fragile that recovery was.

The proximate trigger, according to reporting from Yonhap and the Korea JoongAng Daily, was profit-taking combined with renewed doubt about the scale, financing and profitability of investment across the broader AI value chain. Those are three distinct worries bundled into one selloff: how much is being spent, who is funding it, and whether the spending will ever show up as earnings. When all three questions surface at once, index-level declines of 3% to 5% in a single session tend to follow, and that is roughly what Seoul saw.

MSCI Inc.’s Asia Pacific Index fell 1% in the same session, a smaller but still notable move that shows the pressure was not confined to Korea. The regional decline also stood in sharp contrast to Monday’s session on Wall Street, where the Philadelphia Semiconductor Index rose 1.6%, snapping an eight-session losing streak. That divergence, chips up in the U.S. on Monday and chip stocks down hard in Asia on Tuesday, underlines how jumpy sentiment has become around the same underlying story.

SK Hynix Hits Its Lowest Close Since April

SK Hynix, the memory chipmaker that has become one of the clearest proxies for AI infrastructure demand, fell 4.23% to close at 1.22 million won, its lowest closing level since April. Samsung Electronics, the other pillar of Korea’s chip sector, declined 2.98% to 211,500 won. Together the two companies make up a large share of the Kospi’s weighting, so their losses explain much of the index’s slide on their own.

The irony is that SK Hynix’s underlying business has rarely looked stronger on paper. Just four days before the selloff, on October 16, the company presented its next-generation AI infrastructure lineup at the OCP Global Summit 2026, highlighting HBM4 high-bandwidth memory and SOCAMM2 modules that summit visitors could see installed on live server demonstrations. The company has also been spending heavily to keep up with demand: in August it approved roughly 54 trillion won for two new fabs, 35.2 trillion won for the Yongin Y2 site and 19.1 trillion won for the Cheongju M17 site, both aimed at expanding DRAM and NAND capacity for AI workloads. It also accelerated a 40 trillion won share buyback and cancellation program and set a goal of returning more than half of free cash flow to shareholders.

None of that insulated the stock on Tuesday. That is the part worth sitting with: a company executing well against AI demand still got sold hard, because the question investors are asking is not about SK Hynix’s execution, it is about whether its customers, the hyperscalers and model labs buying all that memory, can turn their own AI spending into profit. Readers tracking the chip side of this story in more detail can see our earlier coverage of Nvidia, Intel, Micron and Monolithic Power being named buy picks ahead of Q3 earnings, which lays out the bull case Tuesday’s selloff is now testing.

The Dollar Slips as Safe Havens Shift

Currency markets moved in sympathy with the equity selloff. The U.S. Dollar Index stood at roughly 101.72 to 101.92 on October 20, down around a quarter of a percent on the day, according to Charles Schwab’s market desk. Against specific pairs the move was similarly modest but consistent: the dollar eased to 158.14 Japanese yen from 158.46 yen, and the euro ticked up to $1.1211 from $1.1227. Separate reporting noted the dollar weakened against all G10 currencies as investors reassessed risk tied to technology stocks.

A quarter-point move in a dollar index is not dramatic on its own. What makes it notable is the timing: it happened on the same day Asian tech shares were falling hard, suggesting some of the capital that had piled into dollar-denominated AI trades was rotating out, at least temporarily. Some of that safe-haven positioning was also tied to unrest in France, which added a layer of geopolitical risk aversion on top of the AI-specific worries. Untangling exactly how much of the dollar’s dip came from AI jitters versus other factors is not possible from Tuesday’s reporting alone, and outlets covering the move were careful not to draw a single-cause line.

Wall Street’s Early Tuesday Picture Was Calmer

The U.S. market open told a more mixed story than Asia’s rout. The S&P 500 fell 0.1% early Tuesday and the Nasdaq composite dropped 0.5%, both modest declines by comparison. The Dow Jones Industrial Average actually rose 80 points, or 0.2%, showing that the selling was concentrated in the same AI-adjacent names rather than spreading across the broader market. Meta Platforms, Micron and Intel all fell in early trading, a direct line from the overnight Kospi rout straight into U.S. premarket sentiment.

That pattern, a severe move in Asia followed by a milder echo in the U.S., is typical of sentiment-driven selloffs rather than a change in fundamentals. If Tuesday’s move reflected new information about AI returns, it should have shown up just as forcefully in New York. Instead, it looked more like Asian investors moving first and more aggressively on the same worry that U.S. traders were pricing in more cautiously, with one eye on this week’s earnings calendar.

Elice’s Rough Seoul Trading Debut

The broader market mood showed up most dramatically in one single-stock story: Elice Inc., a Korean technology company, fell as much as 33% in its Seoul trading debut. A new listing losing a third of its value on day one is an unusually harsh verdict, and it landed on the same day the broader index was tumbling. New listings are often treated by traders as a referendum on overall risk appetite, and Tuesday’s appetite in Seoul was clearly thin.

Why Investors Are Suddenly Nervous About AI Spending

The skepticism did not appear out of nowhere. For more than a year, hyperscalers and model developers have been locked in a spending race that shows no sign of slowing, even as questions mount about the return on that capital. The pace of model releases alone illustrates the intensity: in the final week of September and first week of October 2026, Google shipped Gemini 4 Argon, OpenAI expanded its GPT-6 family, Anthropic added Opus 5.5 and Sonnet 5.5 to the Claude line, and Mistral introduced Large 4. Anthropic followed on October 7 with Claude Haiku 5.5, and Mistral opened a public preview of Large 4 on October 6.

Meta added its own entry on October 12, introducing Muse Wonder and Muse Wonder Full, which the company called its most capable models to date. Muse Wonder is available through Meta’s API and through Muse Code subscribers, while Muse Wonder Full is restricted to Muse Code subscribers only. That launch followed Meta’s September 8 consumer rollout of Muse, a personal AI agent, and came alongside continued investment in Ray-Ban Meta hardware, including the Ray-Ban Meta Audio glasses that began shipping October 13 after preorders opened at $349. Our earlier report on how Meta’s stock jumped on its Muse Wonder frontier bet captured how the market initially rewarded that announcement, which makes Tuesday’s reversal in sentiment toward AI-exposed names feel especially abrupt.

Every one of those releases requires enormous compute, which in turn requires chips, memory and data center capacity, which is exactly the spending investors are now scrutinizing. The logic driving Tuesday’s selloff is straightforward even if the fix is not: labs keep shipping frontier models at a pace of roughly one major release every week or two, each one assumed to need more infrastructure than the last, and at some point markets want to see that spending convert into revenue rather than simply funding the next release cycle.

Meta and Tesla Earnings Loom as the Next Test

The timing of Tuesday’s selloff is no accident. Netflix was scheduled to report earnings the same day, with Meta Platforms, Tesla and IBM following on Wednesday, October 21. That lineup makes this one of the most closely watched earnings weeks of the year for AI-linked stocks, because each of the three companies represents a different piece of the AI spending puzzle: Meta as an infrastructure buyer and model developer, Tesla as a company betting its valuation partly on autonomous software and robotics, and IBM as an enterprise AI vendor.

Meta’s results will draw particular attention to its AI infrastructure spending and whether rising capital expenditure is beginning to show up as revenue growth, rather than simply funding more model releases like Muse Wonder. Tesla, meanwhile, is scheduled to publish its Q3 financial results after market close on October 21, followed by an investor webcast. Shares had fallen nearly 20% year to date as of October 19, even after recovering from a difficult July earnings report, which puts extra weight on Wednesday’s numbers. Readers can find a fuller breakdown of what to expect in our Tesla Q3 2026 earnings preview, published ahead of the report.

It is worth being precise about what is known and not known heading into Wednesday. Tesla’s Q3 production and delivery figures were already released on October 2: the company reported producing 464,391 vehicles and delivering 486,532, with Model 3 and Model Y accounting for 478,237 of those deliveries, alongside 13.7 GWh of energy storage deployments. Those operating numbers are confirmed. What is not yet confirmed is the financial outcome, revenue, margins and management guidance, which will only become public after Wednesday’s close. Analyst consensus compiled ahead of the report had pointed to roughly $28.3 billion in revenue, but that figure is a forecast, not a result, and should be read as such until Tesla’s own numbers land.

How This Selloff Compares to Past AI Jitters

This is not the first time in 2026 that AI-linked valuation worries have hit Korean tech stocks hard. The Kospi’s combined 20.7% slide across seven sessions in early October, before Tuesday’s renewed drop, shows that this pattern of sharp pullback, partial recovery, and renewed selling has repeated multiple times this year rather than representing a single isolated event. Each cycle tends to follow a similar script: a burst of optimism around new model releases or capacity announcements, followed by a reassessment of whether near-term earnings justify the spending, followed by a partial recovery once results come in better than feared.

What distinguishes Tuesday’s move is how directly it is tied to a specific earnings catalyst just one day away, rather than a vaguer macro worry. Markets had roughly two weeks to digest the early-October drop and partially recovered; Tuesday’s renewed selling, by contrast, happened on the eve of the actual evidence, Meta and Tesla’s numbers, that will either support or undercut the AI spending thesis. That makes this round of jitters more of a pre-earnings positioning move than a verdict on the AI trade itself.

Industry Data on the Scale of the AI Bet

The scale of spending that investors are now questioning is not a matter of speculation; industry researchers have been tracking it for several years. IDC’s Worldwide Semiconductor Technology and Supply Chain Intelligence service projected in 2025 that worldwide semiconductor revenue would reach $800 billion for the year, representing 17.6% year-over-year growth from $680 billion in 2024, with the compute segment alone forecast to grow 36% to $349 billion, according to IDC’s research published in 2025. A separate 2024 IDC report had projected the global semiconductor market would grow 15% in 2025, driven by demand for artificial intelligence and high-performance computing, according to IDC’s 2024 supply chain intelligence report.

Corporate appetite for more AI spending has not cooled despite the volatility. In a 2025 workplace survey, McKinsey found that 92% of companies plan to increase their AI investments over the next three years, with 55% expecting those investments to rise by at least 10% from current levels, according to McKinsey’s 2025 report on AI in the workplace. Confidence inside the semiconductor industry itself has also stayed high: 86% of semiconductor executives surveyed expected their companies’ revenue to grow over the coming year, according to KPMG’s 2025 Global Semiconductor Industry Outlook. Those figures help explain the gap at the center of Tuesday’s selloff: the people building and selling AI infrastructure remain broadly optimistic, while public equity markets are pricing in more doubt, at least for now.

Market Impact Across the AI Supply Chain

The practical effect of Tuesday’s selloff is a repricing of risk across the entire AI supply chain, not just at the chip layer. Memory suppliers like SK Hynix and Samsung sit at one end, selling the HBM4 and DRAM capacity that model training and inference depend on. Model developers like Meta, OpenAI, Anthropic, Google and Mistral sit in the middle, racing to release ever-larger systems like Gemini 4 Argon and Muse Wonder Full. And companies like Tesla sit further downstream, betting that AI capability translates into product advantages, in Tesla’s case through autonomous driving software and robotics programs such as Optimus and Cybercab.

A selloff that starts at the memory layer, as Tuesday’s did with SK Hynix, tends to ripple upward through that chain because every layer above it depends on the one below staying investable. If chipmakers pull back on capacity expansion in response to weaker share prices, that eventually constrains what model developers can build, which eventually constrains what product companies can ship. None of that has happened yet. SK Hynix’s capacity plans, including the 54 trillion won fab investment approved in August, remain in place as of October 20. But a sustained period of this kind of volatility is the mechanism by which stock market sentiment could eventually feed back into real investment decisions.

The Open Source AI Angle

One thread running underneath the spending debate is the competitive pressure from lower-cost, often open-weight alternatives. Mistral’s Large 4 opened in public preview on October 6 at a steep discount to prior pricing, with open weights promised for release by the end of the month. That kind of pricing pressure matters to Tuesday’s story because it raises the stakes on whether premium, closed frontier models from the biggest spenders can sustain the margins that justify their infrastructure costs. If cheaper, openly available alternatives can match a meaningful share of frontier performance, the return-on-spending questions driving Tuesday’s selloff get sharper, not softer. Readers following that side of the story can see our coverage comparing Gemini 4 Argon against Claude Opus 5.5 on coding performance and our report on Claude Haiku 5.5’s price cut intensifying the AI pricing war, both of which show the same spend-versus-return tension playing out model by model rather than just stock by stock.

What This Means for AI Funding and IPO Plans

Volatility of this kind tends to show up fastest in the financing plans of companies that have not yet gone public or that are raising fresh capital on the premise of continued AI growth. Scrutiny of AI valuations has already become a live political and financial issue well beyond Tuesday’s trading session; earlier this month, New York City’s comptroller publicly urged Anthropic to pause its reported IPO plans over concerns tied to the broader AI investment picture, a story we covered in detail in our report on the NYC comptroller’s push for Anthropic to pause its IPO. Separately, speculation about Anthropic’s next model, a possible Claude Opus 6, has been circulating alongside reports that OpenAI may be preparing a rival release, detailed in our coverage of the rumored Claude Opus 6 and OpenAI’s Bell model. Days like Tuesday make the market backdrop for those decisions materially less favorable, at least in the short term.

Large infrastructure claims are also facing more scrutiny than they might have a year ago. SpaceX’s reported claim of a $28.5 trillion total addressable market, which we examined in our report on SpaceX’s $28.5 trillion TAM claim topping forecasts by 15 times, is an extreme example of the kind of long-horizon, capital-intensive bet that AI-adjacent investors are now weighing more critically. The common thread across all of these stories is the same question Tuesday’s Kospi drop asked in blunt price terms: how much of today’s spending will actually show up as tomorrow’s earnings, and on what timeline.

Data Table: Tuesday’s Key Market Moves

Market or Asset	Move on October 20, 2026	Level / Detail
Kospi Index	down 3.73%	closed at 5,552.74, intraday low 5,493.24
SK Hynix	down 4.23%	closed at 1.22 million won, lowest close since April
Samsung Electronics	down 2.98%	closed at 211,500 won
Elice Inc. (Seoul debut)	down as much as 33%	new listing, first trading day
MSCI Asia Pacific Index	down 1%	regional benchmark
S&P 500 (early Tuesday, U.S.)	down 0.1%	modest early decline
Nasdaq composite (early Tuesday, U.S.)	down 0.5%	tech-heavy index
Dow Jones Industrial Average	up 0.2%	rose about 80 points
U.S. Dollar Index	down roughly 0.25%	around 101.72 to 101.92
Dollar vs. Japanese yen	weakened	158.14 yen, from 158.46 yen
Dollar vs. euro	weakened	$1.1211, from $1.1227

Data Table: This Week’s Earnings Calendar and AI Model Releases

Event	Date	Relevance to AI Spending Debate
Netflix Q3 earnings	October 20, 2026	first of the week’s major tech reports
Meta Platforms Q3 earnings	October 21, 2026	test of AI infrastructure spending versus revenue growth
Tesla Q3 earnings	October 21, 2026, after market close	tests autonomy and robotics bets beyond vehicle sales
IBM Q3 earnings	October 21, 2026	enterprise AI demand signal
Claude Haiku 5.5 release	October 7, 2026	priced at $0.10 / $0.50 per million tokens for prompts up to 100,000 tokens
Mistral Large 4 public preview	October 6, 2026	discounted pricing, open weights promised by month end
Meta Muse Wonder and Muse Wonder Full	October 12, 2026	Meta’s most capable models to date, tied to Muse Code subscriptions
SK Hynix OCP Global Summit presentation	October 16, 2026	HBM4 and SOCAMM2 showcased for AI server demand

What Comes Next: Five Things to Watch

Whether Meta’s October 21 earnings show AI capital expenditure translating into measurable revenue growth, rather than simply funding the next model release.
Whether Tesla’s Q3 financial results, due after market close on October 21, support the roughly $28.3 billion revenue consensus analysts had modeled ahead of the report.
Whether the Kospi can hold the 5,500 level this week, given it has already tested that threshold twice since October 6.
Whether SK Hynix’s HBM4 and SOCAMM2 announcements at OCP Global Summit 2026 translate into new customer orders that show up in future earnings, easing the gap between infrastructure spending and chip-sector stock performance.
Whether cheaper or open-weight models, such as Mistral’s Large 4, continue to pressure the pricing and margin assumptions behind the biggest AI infrastructure bets.

Frequently Asked Questions

Why did the Kospi fall so sharply on October 20, 2026?
The Kospi dropped 3.73% to 5,552.74 after investors took profits and renewed doubts about the scale, financing and profitability of AI infrastructure spending, with losses concentrated in chip stocks like SK Hynix and Samsung Electronics.

Is SK Hynix’s business actually struggling?
Not based on current disclosures. SK Hynix has approved roughly 54 trillion won in new fab capacity, launched an accelerated share buyback program, and showcased its HBM4 and SOCAMM2 AI memory lineup at OCP Global Summit 2026 just days before the selloff. The stock decline reflects investor sentiment about the broader AI spending cycle, not a confirmed deterioration in SK Hynix’s own results.

What is the connection between the dollar and the AI stock selloff?
The U.S. Dollar Index weakened by roughly a quarter of a percent on October 20 as the tech-led selloff played out in Asia, and the dollar eased against the yen and ticked higher against the euro. Reports noted the move happened alongside the equity selloff but did not establish that AI concerns were the sole cause, since safe-haven flows tied to unrest in France were also cited.

When do Meta and Tesla report earnings?
Both are scheduled to report Q3 2026 results on October 21, 2026, alongside IBM. Tesla’s release is scheduled for after market close, followed by an investor webcast.

Has the Kospi fallen this much before in 2026?
Yes. The index had already dropped a combined 20.7% over seven trading sessions between October 6 and October 15, 2026, before partially recovering. Tuesday’s 3.73% decline is a renewed leg of that same pattern of sharp pullbacks and partial rebounds tied to AI valuation concerns.

What new AI models came out right before this selloff?
In the weeks before October 20, Google released Gemini 4 Argon, OpenAI expanded its GPT-6 family, Anthropic shipped Claude Opus 5.5, Sonnet 5.5 and, on October 7, Claude Haiku 5.5, Mistral opened a public preview of Large 4 on October 6, and Meta introduced Muse Wonder and Muse Wonder Full on October 12.

Did U.S. markets fall as much as Asian markets on October 20?
No. The S&P 500 fell 0.1% and the Nasdaq composite fell 0.5% early Tuesday, while the Dow Jones Industrial Average actually rose about 80 points, or 0.2%. The declines were far milder than the Kospi’s 3.73% drop, though Meta, Micron and Intel shares did fall in early U.S. trading.

Are semiconductor companies still confident about AI demand despite the selloff?
Industry surveys suggest yes. KPMG’s 2025 Global Semiconductor Industry Outlook found 86% of semiconductor executives expected revenue growth over the coming year, and McKinsey’s 2025 workplace survey found 92% of companies plan to increase AI investment over the next three years. Those figures suggest the spending cycle driving Tuesday’s stock jitters is not expected to slow from the supply side, even as public markets reprice the risk.

Tech Stocks Tumble as AI Jitters Sink Kospi 3.73% [2026]