The AI foam will eventually burst, but I think it is far from the real craziness now - tell me my complete judgment on this cycle After reading a conversation between Jim Bianco and David Lin, combined with my own judgment that I have been following, I have compiled it into a complete train of thought. Jim Bianco is the President of Bianco Research LLC. Since 1990, his commentary on the global economy and financial markets has always had a unique perspective, making him a relatively influential voice in the macro circle. Let's talk about Jim Bianco's macro judgment first After Kevin Warsh cancelled the forward guidance, market expectations changed. Bianco's assessment is that there is a high probability that the Federal Reserve will raise interest rates at least once this year, possibly in October. He pointed out that this is not just a matter for the United States. Currently, major central banks around the world have entered a cycle of interest rate hikes or tightening, and the Bank of Japan, the Reserve Bank of Australia, and the European Central Bank have all taken action. The yield on 10-year US Treasury bonds is currently around 4.42%, and he expects it to approach 5% by the end of the year. He believes that short-term interest rates will rise faster than long-term rates, and the yield curve will further flatten. Core PCE inflation is currently at 3.4%, and has been above the 2% target for 63 consecutive months. Bianco's judgment is that the narrative of 'AI will bring long-term deflation' has passed, and we are now in a completely different cycle of high inflation. Speaking of his judgment on AI, this is a framework that I also agree with: the technology maturity curve He believed that AI would eventually experience the bursting of the foam, just as the NASDAQ fell 83% when the Internet foam burst in 2000. However, he believes that we are still in the very early stage, roughly equivalent to the Internet stage from 1997 to 1998, and there are months or even years before the real "peak of craze". The reason he gave was whether overcapacity had occurred. Before the foam burst in 2000, the most typical feature was serious overcapacity, and the number of optical fibers laid in that year was too much for all mankind to use for decades. But he believes that the current AI is completely reversed: there is still a severe shortage of computing power. Alphabet、SpaceX、 Nvidia and other giants are crazily borrowing to expand production capacity, but it will take several years for this money to truly become surplus data centers. He also emphasized that AI is different from short-lived technologies such as metaverse and VR, as it has a clear and disruptive business model that replaces traditional SaaS. In the future, users will only need to use AI prompts for dialogue to process all public and private data across software, which will reshape a software market worth hundreds of billions of dollars. He has a relatively calm judgment on cryptocurrency He believes that cryptocurrency is currently in a "low point of disillusionment" and is not at the same stage as the vibrant AI. Bitcoin has been around for 15 years, while Ethereum has been around for over 10 years. However, apart from its speculative and asset storage functions, its DeFi ecosystem still cannot compete with traditional institutions such as JPMorgan Chase and Jiaxin Wealth Management. The only direction he specifically recognized is stablecoins, especially in high inflation and de dollarization regions like Venezuela, where stablecoins do have a rigid demand as an actual payment method. He suggested that the industry should shift its focus from "selling ETF paper wealth to the baby boomer generation on Wall Street" to truly improving infrastructure such as blockchain underlying trading speed. His final core conclusion As long as the 10-year US Treasury yield remains within a reasonable range of around 5%, he believes that the US stock market, especially the AI sector, still has the momentum to continue its upward push. What investors need to do is to embrace high volatility and be ready to escape alive when the foam really peaks in a few years. Now let's talk about my own judgment I agree that the foam will burst eventually. I don't quite agree with what he said about the interest rate line. The market is now calling for the 10-year term to return to 5% and the Federal Reserve to raise interest rates again, but with such a large scale of US government debt, if the 10-year term really rises to 5% in the long term, fiscal interest rates will rapidly expand, and the pressure on real estate and corporate financing will be even greater. I tend to believe that in the coming years, it is more likely that inflation will be higher than the average level of the past decade, but interest rates will not remain at extremely high levels for a long time. The productivity improvement brought by AI may create a growth environment similar to the technological boom of the 1990s, but today's macro background is different, with higher debt levels, older population structures, and a reversal of globalization. I also agree that the stage of AI foam is more like the Internet stage from 1997 to 1998. I have seen several pieces of evidence myself: firstly, there is currently no situation where supply far exceeds demand, GPUs are still in short supply, electricity is still in short supply, optical modules are still in short supply, and data center land is still in short supply. 800G has not been fully installed yet, 1.6T is just starting, Scale out is just starting, optical engine production capacity is still seriously insufficient, and CoWoS and PIC production capacity are still expanding. The second is that the application layer has not yet truly exploded, and many enterprises are still in the stage of concept verification and small-scale deployment, which will take time to truly change their business models. I would like to add a point about the relationship between AI and SaaS that he mentioned: in the future, the form of software is likely to shift from humans adapting to software to software understanding humans. AI will not eliminate software, but it will restructure the interaction and business model of software. I agree that the foam will burst eventually. But currently, I haven't seen any signs of true irrational frenzy, and I believe the market is still some distance away from the true madness stage. That's why all the short-term technical analysis I do is essentially aimed at finding better positions to add positions and layout for the next two years of financial growth. My basic assumption is that the AI infrastructure investment cycle is still not over, and if there is no fundamental change in supply and demand in the next one or two years, I tend to believe that there is still room for this cycle to continue rising. A sentence for those who are eager to short the US stock market: the market can remain irrational longer than you can remain the solution. The time the market remains irrational may last longer than your money. The current strategy should be to embrace volatility, ride the wave， Leaving with dignity before the real carnival ends, it's not about getting off the car early now and missing the fattest part in the middle. Rational judgment, keep enough money for daily needs, try not to use leverage, and steadily earn money during this period. The market is always there, and opportunities will always exist. Living is more important than making money quickly. Drawing a longer timeline for the next few years is my own extended judgment If the foam is really about to burst, I tend to think that it will peak and fall from the end of 27 to the first half of 28, and will experience a relatively obvious decline in 29 and 30 years, and the entire industry chain of AI hardware will fall again. To add a more rigorous statement: this decline is not due to the industry's demise, but rather a decrease in the return on capital, compounded by overvaluation and changes in the financing environment. The same was true of the Internet in those days. Not only were there too many optical fibers, but also the valuation was extremely crazy, a large number of companies with poor quality went public, interest rates rose, investors began to focus on profitability, and the financing environment suddenly tightened. These factors combined to form the rupture. The rapid expansion of supply and the decreasing marginal return on new investments have led to a slowdown in revenue growth, inventory accumulation, and valuation compression. But this decline is not the end, it paves the way for the next stage. The collapse of infrastructure means that the underlying costs of computing power, storage, and networking will become very cheap. Time may be earlier than many people imagine, but I believe that when the infrastructure costs are low enough, the AI application layer will usher in a real explosion. I tend to believe that the biggest opportunity in the next five or ten years will gradually shift from selling shovels to selling applications. In this AI cycle, infrastructure is the first to be priced by the market. Although the application layer has begun to emerge, the overall commercial scale is still far from reaching a mature stage. This is my complete assessment of the entire AI cycle: infrastructure is priced first, then undergoes a round of valuation clearing, and finally the application layer truly explodes on cheap infrastructure. We are still in the first stage. One thing should be added to this framework: the difference between AI and the Internet is that the infrastructure and application layers are likely to overlap for a long time this time, and the application relay will not wait until the hardware is down. The Internet era is a linear process of laying the Internet, surfing the Internet, and e-commerce step by step. However, the current situation of AI is that GPU is still rising. At the same time, AI search, AI programming, and AI customer service are already making money. AI Agents are also entering enterprises, and robots are slowly landing. That is to say, the hardware has not yet reached its peak, and the applications have already been released. These two lines are likely to be parallel, not sequential. My judgment is that the technological revolution will go through a fixed script: infrastructure boom, capital return rate decline, infrastructure valuation clearing, and application layer harvesting. This is the case in the railway era, the power era, the Internet and, to some extent, cloud computing. History always repeats itself, and this time is likely no exception.