The essence of the AI round is that Internet technology companies with the ability to develop large models gain computing power by purchasing NVDA graphics cards, and sell large model tokens to others. Based on this logic, we can make the following judgments: 1. Thanks to the progress of NVDA's graphics card computing power, the cost of token can be reduced, but the pricing power is in the hands of these big model manufacturers. So as long as there are developers, Internet personnel, financial personnel, and these top white-collar workers need AI to help them improve their efficiency, the consumption of token is impossible to end, and it will become more and more. At the same time, there are many API dongles on the market to be second dealers of token, or various micro innovation products to package a layer of big model. This is consumption at the software level, which will bring demand at the hardware level. You can see that various brokerage software (Tiger Futu Long Bridge), various SaaS software (Flying Book Multidimensional Table)). All kinds of video platforms (one click to generate cover) are engaged in AI assistance. 2. The construction of data centers in the United States is not as easy, involving land approval, power installation, rack purchase and installation, refrigeration, and other aspects. It takes several months to open a milk tea shop in the United States for decoration, and it also involves union issues because the people who come to decorate are employed by unions. Therefore, it is very normal for the computing power shortage in the United States. Mining companies have solved their own land and power problems as long as they renovate them. Therefore, in recent months, mining companies have claimed to convert their computing power centers, and many stock prices have skyrocketed 3. The problem with houses is that the government is artificially creating scarcity and deceiving fools into taking over the market. However, when the market is truly unable to buy them, this problem cannot be solved and will enter a period of reverse flow. Therefore, urban investment is artificially building houses, similar to issuing counterfeit cryptocurrency. Unoccupied houses are used as bargaining chips to control the market, increase their premium ability, and deceive fools into taking over the market at high prices. When the market cannot keep up, urban investment ships to the market and pours water on those who take over the market at high prices 4. The problem with AI is that the token itself is getting cheaper with technology iteration, so big model companies are also trying to squeeze out the most powerful graphics cards through technical means to reduce costs. Some companies do not cut prices, but the universality of the token is expanding, the number of users and the frequency of use are increasing, but the cost of use is decreasing, which is very early in the Internet. It is the same as the development process of science and technology. The only scarcity lies in the hands of technology companies with big models. My model is more loser than you, and you must use my code to write. 5. Data centers are actually a tough job, as you can see from the changes in ORCL, so big model companies are not only building their own, but also outsourcing orders and engaging in outsourcing. In this process, expansion also requires money, and the money from expansion will flow into the market, including construction workers, graphics card providers, truck drivers, power equipment providers, and so on. All links will also receive payments from Party A, and the flow of money is normal. 6. To sum up, I think AI is the initial state of the Internet. The price of AI tokens is constantly reduced. However, users can enjoy the ease of use based on tokens. Similar to the popularization of the Internet, large model providers will eventually become similar to telecom operators, selling the premium of large model tokens. Then ordinary users will pay for applications developed using AI functions. It is recommended to vigorously purchase AAPL and GOOG. 7. At the same time, I also make a prediction that China's AI will have multiple directions in the future: a、 It may be like the internet, requiring you to have some means to access, which indicates that the government does not want ordinary people to have the intelligence of AI b、 It may be like applying for a mobile phone number/broadband, which requires you to open an ID card to access. This indicates that the government believes that as long as China's AI big model is managed and dirty things can be eliminated, ordinary people can have the intelligence of AI