One benefit of Twitter is that you can see what celebrities living in the news are actually posting. Many media reported that Michael Burry, the prototype of "Big Short", the investor who had shorted subprime loans in 2008, again and again warned AI that there was a big foam. I have reviewed his arguments one by one and summarized them as follows: 1. The growth of cloud services by Amazon, Google, and Microsoft (2023-2025 vs 2018-2022) is slowing down 2. Capital investment in infrastructure of the US technology industry reached the level of the Internet foam in 2000 3. NVIDIA OpenAI、 Complex cross investment relationships such as Oracle, Intel, etc. Actually, Chart 1 is not very related to bearish AI; I admit that the mutual investment between these institutions in Chart 3 is quite chaotic, but it is also difficult to constitute a direct bearish argument. The more direct evidence is actually shown in Figure 2, where the Capex growth index measures the speed at which American technology companies have been investing in infrastructure expansion in recent years. Specifically, it belongs to the "big money" expenses that enterprises spend for long-term future profits: building data centers, buying machinery and equipment, repairing factories, laying networks, buying servers, chips, etc., not operating expenses such as daily wages and rent. From a common sense point of view, only when people are floating away, they will spend money and engage in the illusory infrastructure. Therefore, a higher value means a foam. To deepen our understanding, Burry also thoughtfully posted a piece of investment pain literature from the past: By 2002, it was widely reported that the utilization rate of telecommunications capacity in the United States was less than 5%. Thousands of miles of expensive fiber optic networks remained unlit and buried underground. Fifteen continental level telecommunications networks, each providing basic and indiscriminate services, were struggling for survival PS: How do I feel that this text doesn't seem to be criticizing AI, but rather... the unused chains in our blockchain industry, as well as the DeFi and GameFi that nobody uses on them? But the problem is, regarding Chart 2, Burry and I have completely opposite views. Most of the infrastructure built for AI is data centers and graphics card computing power. Currently, these things are not expensive fiber optic cables that have not been lit up, and computing power is very scarce. Just tell me, these chips are just enough for text-based GPT now. How much gap is there before everyone can use Sora2? And this gap hasn't been filled yet, the humanoid robot has come again. Nowadays, everyone is even discussing investing in power plants because the electricity consumption of AI cannot keep up. Do you think this gap is significant? AI must have a foam, but we can't take the old almanac 20 years ago as an example. Burry is 53 years old this year, and overall, he is definitely a strong investor. According to GPT's analysis, his fund and personal wealth should outperform the US stock market in all aspects. Unfortunately, after making a huge profit in 2008 (approximately $100 million for himself and $700 million for his clients), he has issued 12 crash warnings in the past 20 years, most of which have not been proven, such as calling Tesla empty. This time, he returned to Twitter (yes, he was missing for a while), and someone found him to have placed quite a few Nvda put options. Of course, the integration of knowledge and action is normal. Once it's empty, there will be pyramid schemes. However, without presenting more compelling evidence, I feel that the current logic cannot convince me. Considering the huge gap in computing power for humans and the constantly creating new demands, we will continue to invest in chip ETFs, buy small drops and buy big drops.