AI Capital Expenditure Peaks? Goldman Sachs: The Turning Point Is Near, AI Is Shifting from a "Super Theme" to a "Routine Narrative"

The three-year AI super cycle narrative may be quietly shifting gears. In the latest episode of a podcast, three senior executives from Goldman Sachs believe: AI capital expenditure growth is approaching its peak, and the market’s pricing of the AI theme has shifted from “only thinking positively” unilaterally optimistic, to a “calculating both the pros and cons” two-sided game. At the same time, the actual rate hikes from the Federal Reserve will likely be lower than what the market is currently pricing in.

In this October 29th episode of “The Breaks of the Game,” Tony Pasquariello, Global Head of Hedge Fund Coverage at Goldman Sachs, Josh Schiffrin, Head of Global Risk, and Dominic Wilson, Senior Advisor in Goldman’s Research Department, systematically reviewed the evolution of the AI capital expenditure narrative, the Fed’s path, and the macro risk wall just after the super tech earnings week. Dominic Wilson stated bluntly, AI has shifted from a “supercharged theme” to a “more conventional narrative”, and the peak growth rate is “very likely already close.”

The context for this view is that the S&P 500 Index has been essentially flat for the past six months, as has the Nasdaq. Tony Pasquariello pointed out that although this earning season appears “pretty decent,” the upward revision in forward earnings is far less than in Q1 and Q2, “enough to hold the line, but not enough to ignite another breakout.”

Meanwhile, Josh Schiffrin ranked escalation in the Middle East as the biggest short-term growth risk, and called “going long global front-end rates” the highest conviction trade at year-end.

AI: From a “Super Theme” to a “Two-Sided Story”

When a theme is priced by the market on the “most optimistic path,” any underwhelming news will have its negative impact amplified. Goldman’s view is that investment in AI infrastructure is reaching a critical point.

The AI trade has reached this point because of three concurrent clues. First, market prices have already priced in a very optimistic outlook; second, the peak in investment growth is near; and third, the credit demand to finance this wave of investment is beginning to leave traces in the market. When these three stack up, the theme shifts from “just believe” to “it’s time to do the math”—faith alone is not enough; it’s about financing, profitability, and whether interest rates can hold up.

Dominic Wilson explained on the show that since last June, when AI-related stocks saw that “bubble-like peak,” the bank has been concerned that the pricing for the entire AI space is too high—“not that the pricing is absurd, but it’s at a level that requires a sufficiently optimistic path to support current prices, especially in infrastructure.”

And the data provided by the super tech earnings season is no longer one-sided. Wilson put it this way: “What you see on the profit side is a scatter plot. Our own forecasts show that the peak in investment impulse is probably not far off. There will still be growth, but the peak growth rate is very likely already close.”

He further listed three specific cracks, including questions on financing and monetization: First, the credit spread pressure from super tech companies’ credit financing demands is “intermittent but definitely exists”; second, there are still question marks around super tech companies’ monetization paths; third, profit performance is diverging, unlike the prior uniform performance. However, Dominic was referring to the peak growth rate, not that capital expenditure in absolute terms would decrease.

Wilson’s conclusion is: “Once you are facing this situation—when the main story’s growth rate peaks, rates are rising, and credit spreads are widening—it feels like a more conventional narrative. It was once a supercharged theme that lasted about three years. Now, the risk balance is more two-sided, and the story also more conventional.”

Earnings Hold the Indexes, But Don’t Drive Them Higher

In the past half year, the logic of the US stock market has been a tug-of-war: Rates, energy and, to some extent, credit constitute headwinds, while corporate earnings act as the floor. The result is no net movement in indices. The issue with this earnings season is not weakness, but that they are “just enough”: Results were fine, but the upward revision to expectations was not strong enough to spark a new rally.

This is the market’s biggest source of anxiety. In the Q2 reporting season, investors realized this could be an exceptionally strong year for profit growth, and expectations were sharply revised upward as a result. This time, there was no repeat. Tony said: “The S&P is basically back at 7,000... The Nasdaq hasn’t gone anywhere either in half a year.”

“For this season, earnings look fine on the surface. The bar is very high... I think the most interesting thing is, forward earnings weren’t revised up like they were in Q1, Q2.”

In his view, full-year earnings growth will eventually look impressive—possibly near 30%, or even higher (the recorded language here is not perfectly clear; it is advised to verify with the audio before publication). But this year’s numbers have largely been priced in by the market. Tony summarized: “Good enough to hold the line, but not enough to add fuel and let the market break out again.”

Federal Reserve: Market Pricing is Too Aggressive, Actual Rate Hikes May End Here

As the AI narrative cools, uncertainty around rate paths forms the other main thread. Josh Schiffrin’s interpretation of the Fed’s October meeting is broadly dovish, and he clearly believes the current market pricing for rate hikes is overstated.

Schiffrin’s view is: The Fed currently believes it can “act methodically.” Although inflation remains severe, the September policy move has begun to have an effect. Financial conditions have not tightened to extreme levels, but have tightened somewhat. His base case: Another rate hike in December, maybe one more after, but “the eventual number of hikes will likely be fewer than what the market currently expects.”

His exact words: “You now have about 60 to 65 basis points of further rate hikes priced in. I do think they’ll hike again in December, but that’s likely to be it. Maybe one more. But my base case, as we discussed last month, is one more in December.”

Schiffrin also pointed out, the Federal Reserve will reduce the frequency of policy statements from eight to seven times per year starting in 2027 and 2028, but he does not think this will have a major market impact: “Some meetings are already very close together, and not much happens between them, so I don’t think it will have a big effect on the market.”

Dominic Wilson agrees strongly with Schiffrin on rates. He pointed out that the prior rate rise was partly because the 10-year forward OIS reached around 5.20%—“That’s already very close to the peak of the policy-rate cycle over the past 20 years; if you anchor that as your 10-year forward, you’ve built in a lot of risk premium.” He believes the wavering AI narrative is also pushing yields down through “growth concerns.”

Macro Risk Wall: Biggest Threat from the Middle East, Most Underestimated Is Oil Price Downside

When ordering macro risks, the three podcast participants had an interesting mix of differences and complementarity: Josh Schiffrin sees Middle East escalation as the top near-term risk, while Dominic Wilson thinks the tail risk of oil prices “not escalating” is the market’s most underestimated opportunity.

Schiffrin’s logic is direct: Should conflict escalate and energy prices rise again, this is the biggest growth threat; the opposite, a calming situation and lower oil prices, “bodes well for assets overall, whether stocks or bonds.”

Wilson, on the other hand, approaches it differently. He thinks the market has already built in a fair amount of risk premium for Iranian geopolitical risk, and there is real risk premium in oil. Thus, “We simply don’t follow through, or escalation falls short of the market’s concerns, and I think that’s what is underestimated.” His scenario is: If there’s no significant escalation in the two or three weeks after the midterms, “the market will realize this path is better than feared, because we’ve already built in this expectation. Even without any positive developments, just the absence of new bad news is already enough.”

As for the highest conviction trade through year-end, Schiffrin clearly points to “going long global front-end rates”—covering the US, Europe, and the UK on the view that major central banks will ultimately act less than what the market is pricing in, and that any volatility in the credit market will also provide a downward channel for front-end rates.

Tony Pasquariello gave the overall market picture: For the past six months, rates, energy, and credit have been macro headwinds, while earnings have “held the line.” The S&P 500 and Nasdaq have remained roughly flat. This earnings season “looks pretty decent,” but forward earnings revisions are much less than Q1 and Q2—“enough to hold the line, but not enough to spark another breakout.” His conclusion: If Schiffrin and Wilson are correct, the market may have a "decent glide path" before year-end, but it’s likely to be “a different type of stock” leading the rally.