Goldman Sachs warns: AI earnings surprises "retreat" are imminent, and the US stock earnings season may enter a new phase of "expectation verification."

According to the Zhitong Finance APP, Christian Mueller-Glissmann, head of asset allocation research at Goldman Sachs, stated that the wave of earnings surprises driven by artificial intelligence (AI) in the last earnings season significantly boosted U.S. stocks, but this phenomenon is unlikely to be repeated this season. Simply relying on corporate performance alone is insufficient to trigger a new round of significant market rebound.

Although publicly listed companies are still likely to exceed expectations, "the benchmark threshold for this earnings season has clearly been raised." Mueller-Glissmann pointed out that investors will pay more attention to companies' forward guidance and management commentary to capture signals on whether stock indices can continue to rise slowly from the current level.

Goldman Sachs predicts that the earnings growth rate for S&P 500 constituent companies in the second quarter will reach 22%, a figure that aligns closely with compiled market consensus expectations. However, the issue is not whether expectations can be exceeded, but rather "by how much" and "whether the market will respond positively after exceeding expectations."

The surprises from the previous quarter mainly came from the "non-linear expansion" of the AI-related industrial chain, particularly the explosive demand for chips, servers, and cloud infrastructure. Current sell-side expectations have fully priced in these benefits, and the valuations of some leading stocks already imply high growth assumptions. Even if there are surprises this season, the marginal elasticity will significantly weaken, and the market's reaction to "better than expected" results may become muted.

Mueller-Glissmann further explained that typically, in the later stages of a cycle, the upward momentum of earnings revisions tends to last longer, but "this round of large-scale earnings surprises linked to AI capital expenditures is likely nearing its end."

This year, leading U.S. technology companies plan to spend a total of $725 billion on data centers, dedicated chips, and network equipment. Mueller-Glissmann believes that hyperscalers, due to their substantial AI infrastructure assets, are still in a favorable competitive position, but they should focus more on improving operational efficiency and accelerating the commercialization of AI.

The focus for the second half of the year should shift to: improving asset utilization—avoiding redundant construction and idle computing power; strengthening pricing power—transforming computing power into billable cloud service revenue; optimizing capital allocation—seeking a balance between investment and shareholder returns.

"The overall structural trend of AI has not been disrupted," emphasized Mueller-Glissmann, "but the market will no longer generously reward any story related to AI; instead, it will more critically assess the ROI of each business."