Goldman Sachs internal "bull-bear debate": earnings have already peaked vs. clear return on investment

The most brutal week for artificial intelligence led two senior Goldman Sachs strategists to reach diametrically opposite conclusions. This week, Alphabet recorded its largest single-day decline since going public, Intel posted its largest single-day drop in history, Amazon's earnings were equally disappointing, and the Nasdaq 100 index briefly fell into technical correction territory. Although the S&P 500 fell less than 1% for the month, the S&P index excluding AI stocks recorded its strongest monthly performance since 2022the divergence within the market has significantly intensified. Two widely followed strategists within Goldman Sachs published research reports in the same week, each holding their own view. Tony Pasquariello believes that S&P 500 earnings growth has already peaked and that now is not the time to take heavy positions; London-based strategist Mark Wilson insists that this week's earnings from hyperscalers have "quite clearly" answered the question of AI return on investment, and maintains a bullish stance. The Concealed Internal Rift Pasquariello's report opens with a telling paradox: the S&P 500 fell only 0.93% in October, but this calm number masks violent turbulence beneath the surface. He compares the index's monthly performance with Goldman's flagship momentum factor portfolio (GSPRHIMO): over the past four months, the S&P 500's cumulative absolute volatility was about 4 percentage points, while the momentum factor portfolio's cumulative absolute volatility reached about 90 percentage points. Goldman's Bobby Molavi had previously characterized this market state as "unplayable"any investor holding heavy positions during this period suffered substantial losses. Pasquariello's conclusion is: the primary bull market trend remains intact, but the market's upward momentum will weaken somewhat, and volatility beneath the surface will persist. Compared with his stance a week earlier of "licking wounds and swimming toward shore," his position is now more conservativehe explicitly states that he does not intend to make large bets at present. The Earnings Growth Peak Thesis The core judgment of Pasquariello's report this week centers on the interpretation of the tech giants' earnings season. Although some cloud computing businesses continue to compound at high rates on a large base, he still believes the third-quarter earnings "will be remembered as a disappointing reporting season," because several giants showed shortfalls in their core businesses even as hyperscalers' capital expenditure intentions showed no obvious signs of slowing. He extends this to a macro judgment: "S&P 500 earnings themselves have been excellent, without question. But at the same time, earnings growth increasingly appears to have peaked. Moreover, compared with the past several reporting periods, the magnitude of upward revisions to forward EPS estimates has clearly slowed." His core logic is: the second derivative of earnings has turned, and the market has begun to price it in. What is currently benefiting from this rotation is "non-AI" assets. The S&P index excluding AI stocks repeatedly hit new highs as tech giants declined. For the market outlook through year-end, Pasquariello reluctantly cites three supports: sentiment has clearly turned pessimistic; the seasonal shift from the worst period of the year (mid-August to mid-October) to the best period (mid-October to year-end) has begun; and the market has spent a month digesting a large volume of long positions. But he immediately rebuts himself, noting that bears can equally cite the precedent of late 2018. The "Bull Case" of US$1.1 Trillion in Cloud Backlog Wilson's stance stands in sharp contrast. He opens with a self-deprecating footnote in one word"ouch." By his account: Google Cloud revenue grew 103% year-over-year, with backlog up 256% year-over-year to US$561 billion; AWS grew 39%, with backlog up 189% year-over-year to US$578 billion, and management explicitly expressed confidence that AWS revenue growth will exceed 40% in 2027 and 2028; Azure grew "only" 48%, constrained by supply, but committed to accelerating next quarter. The two companies' combined backlog totals approximately US$1.1 trillion. Wilson's conclusion: "These growth numbers are not from startups, but from some of the largest companies in the world. The market may not be in the mood to listen this week, but the ROI (return on investment) question is being answered quite clearly." On the financing side, he remains equally optimistic, citing cases including Meta completing the first two tranches of a record three-tranche equity offering, and Oracle pulling forward its cash flow breakeven timeline and stating it needs only an additional US$20 billion to cover its capital expenditure plans. He believes "the financing gap for AI capital expenditure is far smaller than previously feared." Backlog Does Not Equal Free Cash Flow However, what the market was really asking this week was not revenue scale, but cash. Alphabet is the most typical case: this company sitting on US$561 billion in backlog had negative free cash flow of US$910 million this quarter (market expectation was net inflow of US$3.5 billion), share buybacks have fallen to zero since 2026 (previously US$39.8 billion in buybacks in the first three quarters of 2025), and during the same period it issued US$12.4 billion in new shares through ATM and issued US$28.7 billion in bonds. Goldman's own analysts currently expect its 2027 capital expenditure to be approximately US$350 billion, with 2027 earnings per share below 2026 levels. In addition, the scale of off-balance-sheet liabilities of hyperscalers also deserves attention: according to tracking data in the source material, related off-balance-sheet obligations increased by approximately US$1 trillion within 90 days, roughly equivalent to the combined backlog of the two companies. The positive of Oracle improving its cash flow guidance was also quickly digested after the market studied the earnings report. Goldman strategist Rich Privorotsky on Friday explicitly characterized this issue as a battle over "value capture": there is no suspense on the AI demand side; what the market is really questioning is who can profit along this chain, and at what cost of capital. This is precisely the part that Wilson's backlog argument fails to cover. An Unprecedented Correlation Collapse The two strategists rarely agree on one point: current market behavior is already abnormal. The three-month realized correlation between Apple's stock price and the Nasdaq 100 turned negative for the first time since records began in 1981; the correlation between the S&P 500 and its equal-weighted version has fallen to about 0.62, approaching the 2024 low and the lowest level since 2006. Even rarer, on October 29, about 80% of S&P constituents rose while the index closed down 30 basis pointsplotting all S&P down days against the proportion of stocks that rose that day, this data point sits isolated in the upper-right corner of the sample, far from any similar case in history. At the sector level, this week the S&P excluding AI outperformed the Nasdaq index by more than 3%, and the French CAC index outperformed US tech stocks by 5%. Wilson cites historical market patterns during Fed rate-hiking cycles, arguing that the current situation is merely the stock market digesting a bout of "indigestion," and that the trend has not changed. He frames the core question as "is this 1997 or 1999?" However, 1999 also had impressive revenue growthwhat it lacked was precisely the cheap capital that continued to be supplied after the Fed began its rate-hiking cycle. The Fed held rates steady this week but saw two hawkish dissents; Dallas Fed President Logan said at least two more hikes are needed, while Musalem said the Fed "will soon need to raise rates." Pasquariello said bluntly that the Fed meeting was "almost absent from equity traders' conversations"but that may not last long. This article is reposted from "Wallstreetcn," author: Zhang Yaqi; GMTEight editor: Xu Wenqiang.