Korea was closed on Friday. The chip selloff carried on without KOSPI.
Taiwan fell 6.47%, Nikkei 225 lost 4.03%, and China’s capitalization-weighted market return was negative 3.4%. The broad European and U.S. indexes had a much easier day: STOXX 600 fell 0.3%, the S&P 500 lost 1.0%, and the Nasdaq Composite declined 1.4%. Chip stocks were another matter. Taipei, Tokyo, Shanghai, Amsterdam, and New York all got hit.
That makes some of Thursday’s convenient explanations less useful. Korean rates, leverage, and the KOSPI crash certainly amplified the move. But not a single Korean share traded on Friday, and the selling still crossed five markets. Korea was an amplifier, not the whole story.
Earnings Could Not Hold the Price
TSMC reported record quarterly profit on July 16, up 77% from a year earlier and above expectations. The report had one soft spot: the cost of ramping 2-nanometer production pulled third-quarter gross-margin guidance below some investors’ expectations. TSMC fell 7.29% the next day.
It was a very strong, imperfect report landing on a stock that was no longer allowed to make mistakes.
ASML offered an even cleaner test. It beat expectations and raised its full-year outlook. The shares still lost about 3.8%; ASM International and BESI fell roughly 4.0% and 4.5%. The exact decline in any one stock is not the point. What caught my attention was the absence of buyers willing to lean against the move using TSMC’s 77% profit growth or ASML’s upgrade. Buyers had spent months testing demand. Sellers now wanted smaller positions.
Asian index construction added force. Taiwan’s benchmark suffered its largest one-day point loss on turnover of roughly NT$1.21 trillion. Nikkei fell as much as 6.18% intraday. Kioxia closed down 16.1%, Advantest 7.2%, Tokyo Electron 8.2%, and SoftBank Group about 9%. The heaviest index weights were also among the hottest AI-hardware assets going into July. Selling the theme quickly meant selling the index.
China’s tape looked worse. Across Shanghai, Shenzhen, and Beijing, 5,001 stocks fell and only 482 rose. Turnover increased to RMB2.67 trillion from RMB2.42 trillion. Eastmoney’s semiconductor basket lost 8.1%; only three of its 181 constituents rose, and the group finished the week down 19.9%.
High-volume declines are harder to dismiss than quiet pullbacks. Lower prices did not exhaust the selling; more stock came out. Margin data push the start date back as well. By July 16, China’s combined margin balance had fallen for 11 consecutive trading sessions, from roughly RMB3.03 trillion on July 1 to RMB2.86 trillion, a decline of about RMB172.3 billion.
That does not turn every sale into a forced liquidation. It does tell us leverage was already coming down before Kimi K3 appeared. Friday accelerated an existing move.
Kimi Was Not the Starting Gun
Blaming the selloff on Kimi K3 makes for an easy headline. The timeline does not cooperate.
Moonshot AI released Kimi K3 on July 16. Investors immediately connected it to the question that followed DeepSeek: if frontier models keep getting cheaper, will the money poured into data centers earn the return cloud companies once expected? It is a real question, and it can lower the multiple paid for hardware.
Chip stocks, however, had been falling for weeks. Friday was the Philadelphia Semiconductor Index’s third straight decline. It traded as much as 23.5% below its June 22 record close before recovering to end roughly 20% below it. Kimi gave sellers another reason. It did not press the button.
The orders are still there. The patience is not. Investors are asking when application revenue will catch up with capital spending, and how much return the next dollar of compute can produce. Until that ledger looks better, the most expensive, liquid, and heavily owned hardware stocks are the easiest places to reduce risk.
Oil made buyers more hesitant. Renewed escalation between the United States and Iran pushed Brent up 4.6% to $88.10 a barrel on Friday, from roughly $76 one week earlier. Higher oil raises inflation and corporate costs. Yet long Treasuries, gold, and utilities held up relatively well, so this did not look like a pure inflation panic. Chip positions were already loosening; oil gave buyers another reason to wait.
There is a plainer explanation. These stocks had run too far, earnings day became profit-taking day, and Kimi and oil were stories attached afterward. I would not dismiss it. If semiconductor volume subsides next week, breadth improves, and TSMC and ASML recover quickly, July 17 will look mostly like a positioning accident. If the selling spreads into credit, banks, and industrials while long bonds stop providing protection, the macro explanation deserves more weight.
At Friday’s close, this still looked like a chip-positioning problem, not a full macro breakdown.
Sources: Eastmoney Choice market data, Shanghai and Shenzhen exchange margin-financing aggregates, Yahoo Finance, Taiwan Stock Exchange, Central News Agency, TSMC, ASML, KRX, Reuters, AP, and Kyodo News. Chinese, Japanese, Taiwanese, and European figures refer to the July 17, 2026 close; U.S. equity and energy figures refer to the New York close later that day. Chinese margin data run through July 16. Korean securities markets were closed for Constitution Day. Sector returns, breadth, and fund flow use Eastmoney classifications and may differ across vendors. The relationship among shared risk reduction, Kimi K3, and the oil shock is a research judgment based on chronology, market breadth, and public information, not a proven single cause. This piece is a personal observation and does not constitute investment advice.