Alexander LorenzoCoinPicks, CoinPicks · Founder & Chief Investment Officer CoinPicks Capital 👇🏾This strategy made me 8 million dollars

The Market Wiped Out $3.3 Trillion on a Price Nobody Checked

4 Aug 2026 · 2 min

Watch on YouTube

2:23

In this video, the narrator explains how a misperception about the pricing of a new Chinese AI model, Kimi‑3 (also called DeepSeek 2.0), triggered a $3.3 trillion sell‑off in semiconductor stocks. Wall Street believed China would offer a cheap AI service, undermining U.S. providers, and investors dumped chip‑related equities. The speaker pulls the actual price list: Moonshot charges $3 per million input tokens and $15 per million output tokens – exactly the same rates as the American model Claude Sonnet 5 and five times higher than Moonshot’s previous offering. Moreover, Kimi‑3 is a massive 2.8‑trillion‑parameter model, which raises electricity consumption and operational cost. This contradicts the industry’s long‑standing assumption that AI models become smarter and cheaper over time, meaning the market’s narrative was not just factually wrong but directionally opposite. The narrator urges viewers to always verify primary sources and suggests that such misreadings create undervalued opportunities. He concludes by promoting his own automated trading bot, offered for $1, that capitalizes on these insights.

Chapters

So last month, the market wiped out $3.3 trillion in chip stocks because they thought that China was about to make AI cheap. I went and pulled the actual price list and it says the exact opposite. July 17th, a Chinese lab called Moonshot releases a new AI model called Kimi-3. Wall Street called it DeepSeek 2.0, and the story was simple: China builds something great, gives it away for pennies, American AI companies lose their pricing power. Semiconductor stocks get destroyed. $3.3 trillion of market value gone in weeks because of perception. Let's check the actual truth in the receipt. AI companies charge by the token. A token is basically a chunk of text the model reads or writes. They price it per million. Moonshot charged $3 per million tokens in and $15 per million out. Now here's the part that nobody's reporting. That is the exact same price as Claude Sonnet 5, an American model, and that's roughly 5 times what Moonshot charged for their own previous version. I need you to listen to that again. The Chinese company did not undercut anybody. They raised their own prices to pretty much match the West. Trillions of dollars moved on a narrative that the published price list flatly disagrees with. And almost nobody is checking. But here's why the price went up on the actual model and why this is actually bigger than the sell-off themselves. The model actually got bigger. It got massive— 2.8 trillion parameters— and those are adjustable settings inside the model's brain. More settings equal more electricity and more cost. For 3 years, every forecast in the industry assumed that models get smarter and cheaper forever. This one got smarter and it got way more expensive, which means the assumption underneath one of the most bearish math on AI just completely broke quietly in a press release that nobody read. The market didn't get a fact wrong, it got the entire direction wrong. Check the primary source, always check the primary source, and when you see things like this happen, that is an undervalued opportunity. If you wanna know how I make money off of all of this, I have an automatic bot that trades on my behalf. It buys, sells, and does the research. Go to the link in my bio. We give it away for $1.