Goldman Sachs: AI Capex Growth Nearing Peak, Theme Shifting from One-Sided Optimism to Two-Way Trading
nashnova research
Three senior Goldman Sachs figures say AI investment growth is approaching its peak, and the market is shifting from one-way optimism to a two-way trade — earnings can hold the index but not push a breakout, rate pricing may be too hawkish, and three threads will determine year-end direction.
What's happening with AI capex — still growing, but the fastest growth is almost over?
Goldman senior advisor Dominic Wilson said plainly: AI has moved from a "supercharged theme" to a "more conventional narrative," and the peak in investment growth is very likely close.
This means → companies aren't stopping their AI spending. The acceleration is peaking. In plain terms = the money keeps rising, but the quarter-over-quarter increase is shrinking.
Wilson pointed to three converging drivers: market prices already embed considerable optimism; the growth-rate peak is near; and credit demand to finance this investment round is leaving marks in the market.
Are Big Tech earnings still fine — the surface looks okay, but cracks are showing?
Wilson flagged three specific cracks: credit-spread pressure from mega-cap tech financing is intermittent but real; the monetization path still has question marks; and earnings results are diverging rather than marching in lockstep.
This means → the era of "every big tech stock rallies together, every report beats" is over. The market is now separating who is actually earning from AI and who is still burning cash.
This reflects a shift from a "faith trade" to a "show-me phase" — prices need earnings data to back them, not just narrative momentum.
Earnings growth near 30% — so why isn't the index moving?
Tony Pasquariello noted that over the past six months the S&P 500 and Nasdaq have essentially gone nowhere. Macro headwinds — rates, energy, credit — offset earnings, which were just enough to hold the floor.
This earnings season is "decent on the surface," but forward-earnings revisions are far weaker than in Q1 or Q2. His summary: "Good enough to hold the line, but not enough to re-ignite a breakout."
In plain terms = full-year earnings growth may reach 30% or higher, but the market has already priced that in. A new leg up requires upside surprises — and right now, the surprises aren't big enough.
How many more Fed hikes — is the market pricing in too many?
Goldman's global head of risk, Josh Schiffrin, clearly views current hike pricing as overstated. His base case: one more hike in December, perhaps one after that, but the final count will likely be fewer than the market expects.
Schiffrin's words: "You've got roughly 60 to 65 basis points of further hikes priced in. I do think they'll hike in December, but that's probably it. Maybe one more."
Wilson added: the 10-year forward OIS rate — a gauge of where the market sees policy rates heading — touched around 5.20%, very close to the peak of the past 20 years of rate cycles. This means → if Goldman is right, there is room to go long bonds at current levels.
How to rank geopolitical risk — the Middle East is the biggest threat, but an oil-price drop is underpriced?
Schiffrin ranked a Middle East escalation as the biggest near-term growth risk.
Wilson's angle is the opposite — and more interesting: the tail risk of oil prices not escalating is the most underpriced opportunity right now. His logic: the market has already built in a sizable risk premium for Iranian geopolitical risk, and "even without any real progress, just the absence of new bad news is enough."
In plain terms = everyone is worried oil spikes on a Middle East conflict. But if nothing happens, oil could actually fall as the premium unwinds — and almost nobody is positioned for that.
What's the top trade into year-end — and how do the three threads get tested?
Schiffrin's highest-conviction trade: go long global front-end rates — betting short-term rates will fall — across the U.S., Europe, and the U.K. His reasoning: major central banks will do less than the market prices, and any credit-market volatility would push front-end rates lower still.
This means → Goldman's best year-end idea is not a stock call but a rates call — betting central banks are less hawkish than the market imagines.
Three verification points: does AI growth confirm its peak + does the rate path tilt dovish as Goldman expects + does the geopolitical risk premium fade. Whether all three play out together will determine if the market can produce a new leadership rotation before year-end.
市场有风险，内容仅供研究参考，不构成投资建议。