Recently, Li Bei's letter "To H&B Investors" went viral online:

There was a passage I read several times:
"Taking Anthropic's ARR as an example, revenue growth for the most downstream model companies has significantly slowed, and by the end of the year it will likely be significantly below the market's previously optimistic expectations. A subsequent decline in capital expenditure expectations is probable.
Moreover, based on market performance over the past month, downstream large tech companies and core GPU companies are already declining, and token prices have been falling for three consecutive weeks after peaking."
Undeniably, she missed the AI surge this year and made major mistakes in choosing investment targets. However, the metric she mentioned is something I believe is worth paying attention to.
Lately, I've been thinking about which stage of this AI frenzy we are in.
It's definitely not the early stage; I think we are in the mid-to-late stage of accelerating investment. The current Capex has already consumed the cash that the Meg7 accumulated over many years, moving into the mid-to-late stage of continuing to raise funds through debt and equity issuance.
But is it the end? Looking at the current fundraising situation, investment has not slowed, and there is still enough capital.
However, from the recent volatile market, we can see that investor optimism is fading, and people are beginning to ask: "Do model companies/end users make money? Especially after memory has eaten up a large portion of profits."
Of course, "model profitability" can be seen as the ultimate goal, but before reaching profitability, maintaining high growth is more important.
But we need to understand that the simplest "dumb mode" of investing is over. Capital is continuously narrowing and converging upstream, further upstream. All companies that spend money, issue debt, or have tight cash flows are being sold off, and the Meg7 are falling relentlessly. Materials, semiconductor equipment, and memory are where capital flows.
Key indicators to watch for investment in the second half of the year:
1️⃣ Observe whether the growth rate of AI models can be sustained;
2️⃣ Whether each company's Capex will be lowered, raised, or maintained;
3️⃣ Last is interest rate hikes.
As I said long ago, market sentiment always swings between extreme optimism and extreme pessimism, but the fundamentals have not actually changed much.
Remove this "layer of emotional filter," and look for real companies with demand, execution ability, and that are bottlenecks. Do not let emotions dictate your trades. Hold positions you can "hold on to."