Key Points:
Weekly Drawdown: The Philadelphia Semiconductor Index fell 12.5% last week, its steepest weekly drop since April 2025, with SoftBank down 9% and Nvidia suppliers broadly lower.
Special Report: Forget SpaceX, Elon Is Now Powering the Next Hot IPO (from Brownstone Research)
Forward Demand Threshold: If OpenRouter's top-five ranking stays fully Chinese through Q3, forecast demand for American AI chips into 2026 and 2027 must be revised sharply lower, pressuring the entire Nvidia supplier stack.
Earnings-Price Divergence: Taiwan Semiconductor stock dropped 7% Friday despite the company posting a 77% year-over-year jump in quarterly operating profit. Markets are pricing forward demand, not current results.
Taiwan Semiconductor just posted its best quarter in years. Profit jumped 77%. Their stock fell 7% on Friday anyway. I read that headline three times before it sank in.
A company can break every earnings record and still watch its stock drop the same day. When that happens, the market is telling you something. It's telling you the road ahead looks nothing like the road behind. And if you own an index fund, a 401(k), or almost anything that touches tech, you own a piece of that road.
Here's what worries me. Last week the whole chip index dropped 12.5%. Its worst week since April of last year. SoftBank fell 9%. Nvidia's suppliers took a beating. And the reason isn't what most people think.
Where should you invest $100 right now?
Elon Musk just invented and patented this new AI technology…
And he's predicting it will launch a NEW industry that will grow more than 7 million percent in the coming years.
Even if he's only 10% right, that would still be enough to grow $100 into more than $700,000.
The story on the news is simple. A Chinese company called Moonshot put out a new AI model on July 17. It's called Kimi K3. It has almost three trillion moving parts. That makes it the biggest open AI ever built. Its test scores match the best American ones. So Wall Street panicked, right?
Wrong. That's not what spooked the market.
I dug into this for hours. I still don't think most people realize what's really going on. American tech companies have been quietly using Chinese AI for months. Not testing it. Not experimenting with it in a lab. Using it. Running it inside real products. Products you probably use every week.
DoorDash's technology chief, a man named Andy Fang, said it out loud last week. His company sends work to a Beijing AI model. He called it "lower-level" work. That's the polite version. Cursor, a coding tool a lot of software firms rely on, built its newest system on a Chinese one too. And a brand-new American AI startup used a Chinese model to help train its own product. Then it shipped that product on July 15. Two days before Kimi K3 even came out.
Think about that last one for a second. An American AI company. Building its own American AI. Used a Chinese model to help do it. Then sold the result to American customers. Nobody put out a press release. But it's all public if you look.
I get it. This sounds like inside baseball. But stay with me.
There's a website called OpenRouter. Silicon Valley engineers use it to pick which AI does their work. Every day. It's like a menu. And the top five most-used AI models on that menu this week? All five are Chinese. Tencent. Xiaomi. DeepSeek. MiniMax. And one called z.ai.
All five.
Nobody knows exactly when the switch happened. But it happened well before Kimi K3 came out. K3 didn't cause the rout. K3 just flipped on the lights. Wall Street finally saw what its own customers had been doing for months.
Here's what worries me about our money. The story of the last two years was simple. American companies buy American AI. American AI runs on Nvidia chips. Nvidia goes up. The index goes up. Your 401(k) goes up. Everyone wins.
That story was the bet. Trillions of dollars sit on top of it. Every AI-heavy index fund. Every retirement account tilted toward tech. Every pension fund chasing returns.
And American companies quietly stopped buying only American AI months ago. They didn't tell anyone. Their customers didn't know. Their shareholders didn't know. The chip suppliers didn't know.
Now everyone knows.
Nvidia sells the shovels for the American AI gold rush. That's the whole story of the stock. The story only works if American companies keep digging with American shovels. Last week the market started to wonder if some of them switched shovels.
Taiwan Semiconductor's record earnings don't matter to the market anymore. The market is looking six months ahead. Not six months back. If half of Silicon Valley routes its work through Beijing, what happens to demand for American AI chips next year? What happens in 2027?
Nobody knows. And "nobody knows" is what makes stocks fall the hardest.
I can't stop thinking about one number. The five most-used AI models in Silicon Valley this week are all Chinese. Every single one. If you had told me that a year ago I wouldn't have believed you. Nobody would have.
The trillion-dollar bet on American AI dominance already lost part of its footing. It lost it quietly. Nobody told you. Nobody told me. The market didn't even notice for months.
That's what shook the chip index last week. Not one Chinese model on one day. A hundred quiet choices inside American companies over the last year. All of them added up to the same answer.
The ground already shifted under us. We're just seeing the crack now.
More on this tomorrow.
— Lauren
Editor, American Ledger
