华尔街有多悲观？高盛直接把“软件”类比“报纸”

Core Viewpoint - Goldman Sachs compares the current software industry to the newspaper industry disrupted by the internet in the early 2000s and the tobacco industry facing regulatory challenges in the late 1990s, indicating a fundamental skepticism about the long-term growth and profitability of the software sector [1][2][3] Group 1: Market Sentiment and Valuation - The software sector has seen a significant decline, with a 15% drop in one week and a cumulative retreat of 29% from its September 2025 peak, driven by fears of AI impacting software business models [5] - The forward P/E ratio for the software sector has decreased from approximately 35 times at the end of 2025 to around 20 times, marking the lowest level since 2014 [7] - Despite the apparent return to rational valuation, the underlying assumptions driving these valuations are collapsing, as the market anticipates significant downgrades in future growth and profitability [8][10] Group 2: Earnings Stability and Historical Comparisons - Historical cases, such as the newspaper industry from 2002 to 2009, show that stock prices did not bottom out until earnings expectations stabilized, not merely when valuations appeared cheap [11] - Similar patterns were observed in the tobacco industry during the late 1990s, where stock prices remained under pressure despite significant valuation compression until regulatory uncertainties were resolved [12] - Goldman Sachs concludes that even if short-term earnings reports show resilience, they do not negate the long-term risks posed by AI [13] Group 3: Investment Shifts and Defensive Sectors - There is a noticeable shift in market preference away from "AI risk" towards sectors perceived as more stable, such as industrials, energy, chemicals, transportation, and banking [14][15] - Hedge funds have significantly reduced their exposure to the software sector, while large mutual funds began systematically underweighting software stocks since mid-last year [15] - Certain sub-sectors, such as vertical software and companies with proprietary data and clear industry barriers, may still offer defensive characteristics against AI disruption [16]