$NVDA 2X Short ETF(NVD.US)Made a shallow profit of 12% and cleared, will buy back if the stock reaches 185. The next generation of chips will always be stronger than the current one, computer and mobile phone performance will be excessive, AI cloud service performance will also be excessive, and there is a limit to power load. Eternal high growth is impossible, but despite foreseeing the surplus of computing power after 2026, companies are still forced to participate in this 'prisoner's dilemma' style arms race—stopping procurement means being out, while continuous investment faces the risk of asset impairment. The current strategy is essentially using short-term financial pressure to exchange for survival rights.

This phenomenon indeed reflects the cruel competitive logic of the tech industry. The choice companies face is not 'whether to profit', but 'whether to survive'—just like smartphone manufacturers knowing there will be better chips next year, still must release new models every year. The key lies in: 1. The valuation system of the capital market requires a growth narrative; 2. The generational gap in technology directly translates into market share differences; 3. The migration cost of the ecosystem far exceeds the hardware procurement cost.

Take Meta as an example: In 2024, its AI computing power investment will reach $18 billion, leading to negative free cash flow. But if it stops investing, its ad recommendation system accuracy will lag behind Google's by 2.3 percentage points (according to internal estimates), equivalent to an annual revenue loss of $15 billion. This 'bleeding for survival' strategy is essentially the inevitable pain during the transition from barbaric growth to rational competition in the industry. The sustainability of this model depends on two points: 1. Whether absolute technological monopoly can be achieved before computing power surplus (such as the CUDA ecosystem); 2. Whether hardware costs can be converted into software service revenue (such as Microsoft's Copilot annualized revenue exceeding $10 billion). But after 2027, the industry may face large-scale consolidation, and companies that fail to transform will face real financial liquidation.