Elon Musk spearheaded a warning. Last weekend, several industry professionals published bearish articles on the sector, arguing that AI computing power is transitioning from a phase of frenzied expansion to a shakeout, and that the era of unlimited expansion is over. The reason, they argue, is that most companies currently developing AI computing power borrow heavily to buy expensive GPUs, then build data centers, maximizing leverage, and ultimately profiting by renting out computing power. This business model was very easy to raise funds for last year and this year, and was highly favored by market investors. However, recently, a real situation has emerged in the financial market, prompting many industry professionals to issue warnings: lenders are refusing to provide loans. These companies, aside from relying on future profits from renting out computing power, have almost no other stable revenue streams. If AI demand weakens, they risk losing their money, and lenders, fearing total loss, are simply closing down. Furthermore, building data centers in the real world is not particularly smooth, easily delayed due to insufficient power, land and permit issues, cooling, and network infrastructure. Many projects have already spent money, but by 2027 they may still be only half-built, or completed but idle, failing to recoup interest and costs. This is what Musk refers to as "stranded computing power." Therefore, Musk directly stated that he expects to soon see a large amount of stranded computing power in those half-built data centers, due to overly optimistic expectations regarding the feasibility of data center deployment.