AI narratives are becoming increasingly outrageous

I've been in Japan these past couple of days. I heard about a magical Japanese AI company, alt.ai, which went public with great fanfare last year, boasting the ability to create highly synchronized "digital human clones." But less than a year later, it collapsed. A quick look at the sales figures revealed they were inflated. The founder is elusive, never appearing in person, having only spoken to reporters three times using a digital clone. It seems AI is truly a global issue, even a familiar formula knows no national boundaries. Why is the AI ​​narrative becoming increasingly outrageous? Because there are two plots that are becoming increasingly incomprehensible to the average person: Plot 1: The intensifying data center arms race. On one hand, OpenAI is spending billions of dollars to build a "Stargate"; on the other, Meta wants to build a data center the size of Manhattan. Nvidia, the shovel seller, is investing in mining tunnels for OpenAI. The harder the miners dig, the more shovels they sell. The capital market pours in astronomical sums, seemingly consuming all the electricity on Earth to power its models. But everything comes at a price, and you never know the sacrifices involved. As figures reach new highs, the market begins to wonder: Is this money really worth it? As models pile up, bills pile up, and valuations soar, the air of a bubble grows stronger. With round after round of financing and the emergence of unidentified players, the formula becomes increasingly familiar. According to Bain's estimates, this wave of AI infrastructure investment would have to generate $2 trillion in annual revenue by 2030 to be worthwhile—equivalent to the combined 2024 revenue of the "Big Six" (Apple, Amazon, Google, Microsoft, Meta, and Nvidia). But how many are profitable now? Plot 2: You pay me, I pay you, and we make money together. OpenAI signed $1 trillion in contracts this year, but revenue was only about $12 billion. But that's okay. Everyone can sit down, open their wallets, and give each other money, and it'll look like everyone's making money. Nvidia invested $100 billion in OpenAI to build AI infrastructure, and OpenAI used that money to buy Nvidia chips. xAI received funding from Nvidia and used it to buy Nvidia processors. OpenAI purchased 6 gigawatts of chips from AMD over the next few years and then bought AMD stock at a fraction of the price. If you remove the "you pay me, I pay you" part, where does the real money come from? Currently, it's only subscription fees (theoretically, data center computing power could also be a source of revenue, but it depreciates so quickly, only over 2-3 years). If we rely solely on subscription revenue, we'd need far more people willing to spend money on AI than we do now, reaching a market size more than five times the global subscription software market to break even. This "small goal" is too difficult. Willingness to pay has long been saturated, and token subsidies are no longer sufficient. So what should we do? Why not take a look at the cryptocurrency world? The last episode before the DeFi summer's hot money receded was "Community and Social." After Web3's practicality collapsed, it shifted to "metaverse" marketing. It felt like a good idea—and so, OpenAI's AI-powered TikTok, Sora, arrived, and so did Meta's Vibes. Everything becomes traffic, everything becomes advertising. Perhaps this is the ultimate solution. I remember mentioning a radical theory with Professor Will on a podcast: the cycle of generational bubbles: the gap between two similar generational events is about 25 years. 2025 - 25 = 2000. What happened back then? Of course, how can AI be compared to the internet of its day? I wouldn't dare make that comparison, as I'd be considered a fool (though I don't know why). So, most players right now still have a "better to kill by mistake than to let it slip through" mentality. Who would dare miss this once-in-a-century opportunity?