I mentioned AI at halftime, while Howard Marks said AI is only in the third inning of baseball. Here are the key points from his latest CNBC interview: 1. The best way to prevent irrational exuberance is for the market to experience occasional setbacks. Markets always tend to amplify everything. Over the past year to year and a half, the market has experienced several significant fluctuations, including the recent correction in AI and tech stocks. In his view, these setbacks are not bad, but rather healthy. The truly dangerous time is when everyone believes that nothing will fail. When investors begin to believe that no investment will go wrong, the most serious bubbles and irrational exuberance often form. Again, this stems from reflexivity. 2. Overall, he predicts that interest rates will largely remain around current levels for the next few years. The ultra-low interest rate period from 2009 to 2021 was itself an anomaly, and it's unlikely we'll return to that era. Considering that inflation remains a concern, he's quite certain that future interest rates won't be much lower than they are now. This places higher demands on investment; 3. Even so, he still emphasizes risk control in his investments: 1) He doesn't believe that the bonds of these companies are currently so cheap that they're worth buying indiscriminately; 2) AI is the most unpredictable thing he's encountered in his entire investment career, but he doesn't advocate completely avoiding it. He prefers to participate through relatively safe methods. To elaborate further, Oaktree Capital, founded by Marks, primarily deals in lending and distressed assets. He says that a safer approach is to participate in building data centers in less crowded areas. Why do we believe AI is in its mid-stage, not its end? The core reason is that the large-scale commercialization of AI has only just begun. Looking at the tweet that was quoted and forwarded, the cloud businesses of major cloud vendors only started to accelerate three quarters ago, and the next year or two will be a phase of gradually deploying computing power (depending on the pace of data center construction). Therefore, the growth rate of cloud business performance will continue to accelerate. Consider what the scenario would be if AWS reached a scale of $360 billion in 2028, and Microsoft and Google Cloud reached scales of over $300 billion. Anthropic's revenue reached $4.5 billion in 2025, and $33.2 billion in the third quarter of this year, with a full-year target exceeding $50 billion. While it's unlikely to maintain a year-on-year growth rate exceeding 1000% in the future, a relatively high growth rate is highly probable for the next few years. Those familiar with the commercialization of the mobile internet know that once a technology's penetration rate crosses the 10% threshold, it spreads rapidly, and commercialization accelerates until saturation and a plateau. Clearly, after coding, enterprise adoption of AI is deepening and spreading, and new scenarios such as personal agents and AI scientific discoveries are constantly emerging. This indicates that large-scale commercialization of AI is accelerating, not reaching its peak. The real risk to the future market is a slowdown in AI commercialization, but not now. Of course, problems exist: huge capital expenditures and negative free cash flow are not to be ignored, contributing to market volatility and adjustments. However, this does not mean it has reached its limit. As mentioned in the tweet "Ice and Fire in the Midfield of AI": Trend investing isn't about stubbornly sticking to the existing winners, but about constantly identifying where the next profit pool is shifting to and which sectors are the new beneficiaries and new main themes. Of course, I'm not a die-hard bull; in fact, I was probably one of the first to predict significant volatility and correction in the US stock market before the midterm elections. 1) As early as late August, I discussed this here: https://(((x.com)))/qinbafrank/status/2091511833140363356?a=_1C75sTz2787bXpIVDE&s=46, that mid-September to October might be a period of significant pressure; the major tech earnings reports in late October would determine whether the fundamentals could provide support; and after November 3rd, with the election uncertainty fading, there would be an opportunity for a recovery. The core logic is also about oil prices, yields, and debt issues; 2) Then, on September 1st, I participated in an interview with 168X (https://(((x.com)))/qinbafrank/status/2094643546334916871?a=9IUWhi8Rs-Uz3ffr3B0&s=46), where we discussed: the uncertainty of the midterm elections from mid-September to mid-October, the long-term bond yield game, and the impact of Anthropic's IPO on the market. Increased volatility; After November, with reduced political uncertainty, Q3 earnings reports confirming the trend, and valuations becoming cheaper (SOX's current dynamic valuation is around 19.2, only about 10% lower than the low point during last year's tariff war), the market will truly recover; 3) Then on October 9th, https://(((x.com)))/qinbafrank/status/2109645190432772561?a=f9qTh4-wX9zz4ssikP8&s=46 it was discussed again that the current market will be very unstable. Risk aversion before the CPI, the election, and concerns about industries will all lead to adjustments at the slightest sign of trouble. High beta and momentum stocks are prone to decline. From a personal perspective, we need to look at the risks and opportunities, acknowledging the problems, their magnitude, and their impact on the market. But we also need to see the aspects where the fundamentals are improving. It's not that prices will skyrocket or plummet, and many people's opinions and judgments are driven by price movements. Of course, from an investment perspective, the real test is position management and portfolio adjustment capabilities (requiring a shift towards new winners and main themes). Looking at this earnings season again, commercialization is indeed accelerating, but debt and cash flow issues are also amplifying and accelerating, and the macroeconomy remains unstable, especially with oil prices still high. How far this recovery can go depends on several factors: 1) Whether oil prices can fall will determine the inflation trend. Inflation, in turn, determines the Fed's subsequent interest rate direction; the market can accept up to two rate hikes, but more than that will be unbearable. 2) In terms of industry, commercialization is accelerating further. As for capital expenditures reaching sufficient returns, it may go through three stages: First step: converting production capacity into revenue, looking at revenue and growth rate; Second step: looking at operating profit after depreciation; Third step: the operating cash flow/cash Capex changing from less than 1 to equal to 1 and then to greater than 1. 3) Regarding the midterm elections, my personal view is that the conclusion is more important than the result. The conclusion means reduced political uncertainty, and the market will infer policy directions based on the new congressional power structure. As discussed last weekend, even if the Democrats win both houses of Congress, it doesn't necessarily mean they'll curb AI and crypto development; it just means their efforts will be less forceful than the Republicans'. It's not a matter of direction, but rather the intensity. 4) Of course, we also need to watch Anthropic's IPO roadshow soon. From a personal perspective, the overall market is gradually shifting its logic: 1) Over the past year, the market has rewarded scarcity, expansion, and imagination; 2) Going forward, the market will place more emphasis on utilization, profit margins, free cash flow, and return on capital. This is just my personal opinion. This post is sponsored by @bitget_zh, "Bitget Buy US Stocks: Instant Entry, Smooth Trading"