The semiconductor index fell 20% from its June high and entered a bear market on Friday — the first time since the April 2025 tariff meltdown. The worst week for chips in fifteen months. And the catalyst that finally broke it wasn't an earnings miss. It wasn't the blockade. It wasn't the Fed. It was a Chinese startup called Moonshot, which unveiled a 2.8-trillion-parameter open AI model called Kimi K3 and said it rivals the best from OpenAI and Anthropic. The market looked at that and asked the one question it has been avoiding since the AI rally started: what if the moat isn't real? What if the hundreds of billions the hyperscalers spent on Nvidia chips doesn't buy a lead that China can't close? The S&P fell 1%. The Nasdaq dropped 1.4%. The Dow lost 406 points. Netflix crashed 9% on weak guidance. Apple briefly passed Nvidia to become the most valuable company on earth — not because Apple went up, but because Nvidia went down. Consumer sentiment rose to a five-month high. The public feels better. The market doesn't.
Red Friday. The S&P fell 1% to close near 7,458. The Nasdaq dropped 1.4%. The Dow lost 406 points. The semiconductor ETF fell 4% — its seventeenth percent of decline in July alone. The SOX index crossed the 20% threshold from its June peak and entered a bear market. On the weekly scoreboard, the S&P lost 1.5%, the Nasdaq fell 2.9%, and the Dow shed about 1%. After a week that started with the best CPI print in six years and ended with a Chinese AI model shaking the foundations of the trade, the market gave back everything the data gave it.
| The Numbers I Circled | At the close, July 17 · Day change |
|
| SOX Index | bear market | −20% from June |
| S&P 500 | ~7,458 | −1.0% |
| Netflix (NFLX) | Q3 guide miss | −9% |
| S&P 500 Sectors | Day change |
|
| Energy | | +2.0% |
| Real Estate | | +1.0% |
| Consumer Staples | | +0.8% |
| Health Care | | +0.5% |
| Utilities | | +0.4% |
| Materials | | −0.3% |
| Financials | | −0.5% |
| Consumer Disc. | | −1.0% |
| Industrials | | −1.5% |
| Comm. Services | | −2.4% |
| Info. Technology | | −3.0% |
| | Notable Gainers | Day change |
|
| |
| |
| |
| |
| |
Moonshot AI unveiled Kimi K3 on Friday morning — a 2.8 trillion-parameter open-weight model the company says is the world's largest and performs on par with the best from OpenAI and Anthropic. The announcement hit a market that was already selling chips and rotated the question from "how much hardware does AI need?" to "does hardware alone buy you a lead?" Every chip name extended its losses. Nvidia fell 2%. Applied Materials, Lam Research, KLA, Intel, and Arm each dropped about 4%. Micron fell further. The sector closed off its worst levels — some traders bought the dip — but the damage was done.
Netflix crashed 9% after its Q3 revenue and earnings guidance came in below Wall Street estimates. The stock dragged communication services down 2.4%. The consumer sentiment story from PepsiCo last week — North America "softer than anticipated" — just spread to streaming. Coca-Cola slipped on reports of a ransomware attack at its Fairlife milk unit. Uber announced a $14.8 billion deal to acquire Delivery Hero.
Apple briefly passed Nvidia during the session to become the world's most valuable company for the first time since the AI rally began. HSBC upgraded Apple to Buy with a $366 target, calling it an "operational turning point." The rotation from chips to platforms that we tracked on Wednesday just got its starkest data point: the company that buys chips is now worth more than the company that designs them.
Oil rose 2% to above $80 after the U.S. struck bridges and an airport in Iran and Tehran hit a power and desalination plant in Kuwait. Gold topped $4,019. The VIX climbed to 18. But consumer sentiment — the University of Michigan's July preliminary reading — rose to a five-month high. Main Street feels better even as Wall Street sells.
What The Market Is Pricing In
When a sector falls 20% from its recent high, on Wall Street they call it a bear market. The label sounds dramatic, but the mechanics are what matter. Pension funds, insurance companies, and endowments have rules that say they have to reduce their exposure when a sector crosses that line. They don't want to sell. The rules make them. On Wall Street they call this forced liquidation — selling that happens not because someone changed their mind about the business, but because a number on a spreadsheet crossed a threshold. The SOX hitting −20% today triggered some of that. The sector closed off its lows because other traders — the ones who make money buying what the rules force others to sell — stepped in. That tug of war between forced sellers and opportunistic buyers is what the next two weeks look like for chips.
But the bigger story isn't the label. It's Moonshot. A Chinese startup with a fraction of the budget of Google or Microsoft just built a model that — according to its benchmarks — competes with the best American AI has to offer. The market has spent six months paying premium prices for every company in the Nvidia supply chain on the theory that AI requires so much hardware that whoever builds the most wins. Kimi K3 challenges that theory. If China can build competitive AI models with open-weight architectures and cheaper hardware, the moat that justified $60 billion in TSMC capex and $100 billion in hyperscaler spending starts to shrink.
The SOX entered a bear market — down 20% from June — and the market is telling you that the AI trade's biggest risk isn't the blockade, or the Fed, or the earnings, but the possibility that China just demonstrated you don't need a trillion dollars in hardware to compete. Moonshot's Kimi K3 may or may not live up to its benchmarks. Chinese AI models have overpromised before. But the timing matters: the announcement hit a market that was already rotating out of hardware and into platforms, already questioning whether TSMC's $64 billion capex was spending or waste, already watching Apple pass Nvidia in market cap. The question isn't whether Kimi K3 is as good as it claims. The question is whether the market still believes the hardware moat is worth what it paid.
The last time chips fell this fast was April 2025, when Trump's tariff announcement sent the SOX down 20% in two weeks. The recovery took three months but it happened — because the AI demand turned out to be real. If the demand is still real this time — and TSMC's 68% revenue growth says it is — the sell-off is a reset, not a reversal. But if Moonshot proves that competitive AI can be built cheaper, the reset becomes a repricing. Next week the hyperscalers start reporting. Microsoft, Alphabet, Meta, and Amazon will tell you whether the money they spent on hardware is turning into revenue — or just turning into competition.
Three things I'm watching next week and over the weekend:
01 — Does the SOX find a floor Monday July 20?
The semiconductor index just entered a bear market. Korea fell 6% Friday and is instituting new controls on leveraged ETFs. The forced selling from the 20% threshold may not be done. If the SOX stabilizes Monday and Tuesday — holding above the April low — the correction is a reset and the buyers take over. If it breaks below April's floor, the next support is 25% down and the word "crash" enters the conversation. JPMorgan told clients Friday to buy the chip dip, citing strong AI demand through 2028. Watch whether institutional money follows that call.
02 — Hyperscaler earnings begin late next week
Microsoft and Alphabet report on July 29, followed by Meta and Amazon. These four companies have spent a combined $200 billion on AI infrastructure over the past year. The market needs to see that spending converting into revenue — cloud growth, advertising AI, subscription gains. If the hyperscalers guide above consensus and show AI revenue acceleration, the chip sell-off ends because the demand is confirmed. If they guide cautiously — especially on AI monetization — the Moonshot question gets louder and the rotation out of hardware accelerates.
03 — Iran escalation over the weekend
The U.S. struck bridges and an airport in Iran on Friday. Iran hit a power plant in Kuwait. Oil is above $80. The blockade on Iranian shipping is active. If the weekend brings a new round of strikes — or if the Strait functionally closes — oil goes to $85 and the CPI relief from this week gets erased. If diplomats from Qatar or Pakistan announce progress toward talks, oil eases and the market gets room to focus on earnings instead of the war. The geopolitical premium in oil is the swing variable for inflation, the Fed, and the consumer for the rest of the summer.
The SOX is in a bear market. Netflix missed. China showed up. And the best inflation data of the year landed in the same week the market lost faith in the trade that carried it all year. Next week is about whether the hyperscalers can rebuild that faith — or whether Moonshot just changed the math. Have a good weekend. It's going to be a long one.
That's it for today. Have a good weekend. I'll be back on Monday after the close.
— Tom Hartley
Today In Perspective · Published daily, Monday–Friday, after the close
This newsletter is for informational purposes only and does not constitute investment advice. The author is not a registered investment advisor. Past performance does not guarantee future results. Consult a qualified financial professional before making investment decisions.
