Chips Enter a Bear Market as China's Kimi K3 Rattles the AI Trade
The single most important story of the past week wasn't an earnings miss or a Fed headline — it was a 2.8-trillion-parameter open-weight AI model out of China. Moonshot AI's release of Kimi K3 on July 17 landed like a second DeepSeek moment, and it turned an already-shaky semiconductor trade into a full-blown bear market. By Friday, the Philadelphia Semiconductor Index (SOX) had fallen more than 20% from its June 22 record close, capping its worst week since March 2025 with an 11% single-week drop, as Bloomberg reported.

Photo by Tima Miroshnichenko on Pexels
Why One AI Model Moved Global Markets
Moonshot claimed Kimi K3 performed competitively with top-tier US models while undercutting them sharply on price — $15 per million output tokens versus roughly $50 for comparable output from a leading US model, according to Fortune's coverage. That combination hit the core assumption that has powered the AI trade for two years: that US firms can defend their lead simply by out-spending everyone else on compute. If a Chinese lab can get close to the frontier for a fraction of the cost, the enormous capital expenditure plans behind the chip rally look harder to justify. The reaction was immediate and global — Taiwan Semiconductor fell 7% on Friday even after reporting a 77% jump in quarterly operating profit, and SoftBank dropped 9%, per Bloomberg.
The Chip Selloff by the Numbers
This wasn't a one-day scare. The SOX had rallied 105% from its March low to its June peak, and it gave back a fifth of its value in a matter of weeks. Marvell Technology, ARM Holdings, and Intel have all fallen more than 30% from their recent highs, and memory-chip makers Micron, Samsung, and SK Hynix have separately crossed into their own bear markets, according to Yahoo Finance. South Korea's KOSPI, heavily weighted toward memory chips, fell nearly 20% in just two weeks. Some analysts frame this as a valuation and positioning reset rather than the end of the AI infrastructure buildout — but the magnitude and speed of the drawdown are hard to wave away.

Photo by Darkmoon_Art on Pixabay
Netflix's Earnings Beat That Wasn't
Netflix added its own gut-check to the week. The company reported Q2 revenue of $12.56 billion (just shy of the $12.58 billion consensus) and EPS of 80 cents, a penny ahead of estimates — a fine quarter on paper. The stock still sank as much as 12% after hours, according to The Motley Fool. The trigger wasn't the numbers; it was weaker Q3 guidance and Netflix's decision to stop publishing its twice-yearly "What We Watched" engagement report, shifting to an annual release starting in 2027. Investors read the reduced disclosure as a signal that engagement trends aren't something the company wants under a microscope right now, and the stock touched a 52-week low in the process.
Rotation, Not Capitulation
Money didn't leave the market — it moved. As chips sold off, capital rotated into financials, industrials, and energy, a pattern several outlets flagged as the market broadening out rather than breaking down. Oil also jumped on renewed US-Iran tensions, adding an energy-sector tailwind and a fresh source of macro uncertainty at the same time. The S&P 500 closed the week at 7,457.69, down 1.01% on Friday alone, while the Nasdaq Composite fell 1.4% to 25,520.24 as tech absorbed the brunt of the damage.
What's Actually on the Calendar Next
Earnings season doesn't slow down: roughly 86 S&P 500 companies report this week, with the load concentrated midweek — 73 on Tuesday and 135 on Wednesday, including Danaher, 3M, Northrop Grumman, General Motors, and Capital One. Blended Q2 earnings growth for the S&P 500 stands at 24.7% so far, a strong number that's been overshadowed by the chip story. Looking further out, the two dates that matter most for the broader market are already on the books: the Federal Reserve's rate decision on July 29, and the GDP report on July 30. With chip stocks this volatile and oil moving on geopolitical headlines, either release has more power than usual to swing sentiment in either direction.
This is not financial advice — always do your own research before making investment decisions.
The Takeaway
The AI trade just got its first real stress test in over a year, and it came from an unexpected direction — not a Fed surprise, but a cheaper, competitive model from a Chinese lab. Whether this is a healthy reset in an overextended sector or the start of a deeper repricing of AI capex assumptions will likely become clearer only after the next earnings wave and the July 29 Fed decision. Until then, the rotation into financials, industrials, and energy is worth watching as a signal of where conviction is actually going, not just where it's leaving.
댓글
댓글 쓰기