Original Viewpoint Roundup

⚠️ Market Risk & Pullback Setup

• The market is showing a clear risk-off setup. In the Wednesday discussion with 168X, the focus was already on potential market risks and a correction. The core warning: capital is chasing every Jensen Huang / AI narrative too aggressively—when he mentions AI PC, money floods into the PC / CPU supply chain. That kind of narrative-driven crowded trade can easily unwind. 1

• If this correction becomes sizeable—and the base case is that it very likely could be large—the better move is not to freeze in panic. A deeper flush may become a better entry window rather than a reason to stay sidelined. 2

• Market drawdowns should be measured by percentage, not absolute index points. A 1000-point drop today is not comparable to a 1000-point drop more than a decade ago; back then, that could have meant around a 30% single-day crash. Absolute-point framing creates distorted fear signals. 3

🧠 AI Capex, Data Centers & Monetization Risk

• Microsoft $MSFT is building AI factories at the fastest pace in its history, and the depreciation from those data centers is already showing up in the financials. The key market concern is not demand—AI demand and orders are both strong. The real question is whether AI profitability can outrun the depreciation curve. For now, there is no clear answer. 4

💾 Micron / Memory Trade

• Micron $MU below 900 is not automatically a blind dip-buy. Position management matters. The existing $MU position is 8%, with an average cost of 450美元. Partial profits were already taken twice at 950 and 1000, which pulled the original capital back. The remaining position is still large because the stock has run hard, and the target price remains 1500美元. 5

🤖 Robotics Supply Chain

• In humanoid robotics, the highest-value bottleneck is likely the actuator system. The logic comes from the BOM structure: actuators account for roughly 40%–60% of humanoid robot BOM and directly determine cost, performance, and reliability. Within actuators, harmonic reducers, roller screws, linear guides, and force-control sensors are the segments most likely to face supply constraints. 6

🧰 AI Tools & Personal Productivity

• Heavy AI usage does not automatically convert into high-quality output. Codex usage reached 40 billion tokens over 100 days, with a peak of 1.2 billion tokens in one day and recent weekly usage around 3–4 billion tokens. But much of the early project design and key review work still came from chatbot discussions, with some execution delegated to deepseek and glm; deepseek alone costs nearly 100 RMB per month. The bottleneck may not be token supply—it may be personal output quality. 7

🧑‍💼 Company Quality Signals

• Employee churn is an underrated signal. Bad companies cannot retain the people they hire. Constant turnover resets teams, kills momentum, and makes real progress hard to compound. Watching who leaves can be one of the most useful signals for investors, customers, and job seekers. 8