Has AI capital expenditure peaked? Goldman Sachs: The inflection point is near, and AI is evolving from a "super theme" to a "regular narrative"
The three-year narrative of the AI supercycle may be quietly shifting gears.
Three senior Goldman Sachs figures judge in the latest podcast dialogue that the peak growth rate of AI capital expenditure is just around the corner, and the market's pricing of the AI theme has shifted from the unilateral optimism of "only thinking about the upside" to the two-sided game of "calculating both pros and cons". Meanwhile, the actual magnitude of the Federal Reserve's interest rate hikes is likely to be lower than the market's current pricing expectations.
In this episode of *The Breaks of the Game* recorded on October 29, Tony Pasquariello, Goldman Sachs Global Head of Hedge Fund Coverage, Josh Schiffrin, Global Head of Risk at Goldman Sachs, and Dominic Wilson, Senior Advisor in Goldman Sachs Research, systematically reviewed the evolution of the AI capital expenditure narrative, the Federal Reserve's policy path and the wall of macro risks right after the earnings week of mega-cap tech stocks.
Dominic Wilson stated bluntly, AI has transitioned from a "supercharged theme" to a "more conventional narrative", and the peak growth rate is "very likely quite close".
The background of this judgment is that the S&P 500 has basically moved sideways over the past six months, so has the Nasdaq. Tony Pasquariello pointed out that this earnings season "looks decent on the surface", but the upward revision of forward earnings is far less than that in the Q1 and Q2 periods, "enough to hold the line, but not enough to reignite a breakout rally".
At the same time, Josh Schiffrin lists the escalating situation in the Middle East as the biggest near-term growth risk at present, and names "long global front-end interest rates" as the highest-conviction trade by the end of the year.
01
AI shifts from a "super theme" to a "two-sided story"
When a theme is priced by the market according to the "most optimistic path", any information that falls short of expectations will amplify the negative impact. Goldman Sachs judges that AI infrastructure investment is at a critical tipping point.
The AI trade has reached this point as a result of three concurrent trends. First, market prices have already priced in very optimistic expectations; second, the peak of investment growth is approaching; third, the credit demand to finance this round of investment has begun to leave its mark on the market. The overlap of the three factors has turned this theme from "just believe in it" to "you have to do the math" — faith alone is no longer enough, we need to check whether financing, earnings and interest rates can sustain the momentum.
Dominic Wilson explained on the show that since the "bubble peak" of AI-related sectors last June, the bank has been worried that the market has priced the entire AI space too high — "not that the pricing is absurd, but it is at a level that requires a fairly optimistic path to support current prices, especially in the infrastructure sector".
The data provided during the earnings season of mega-cap tech stocks is no longer one-sided. Wilson's exact words are: "What you see on the earnings side is a scatter plot. Our own forecasts show that the peak of the investment impulse is not too far away. Growth will continue, but the peak growth rate is very likely quite close."
He further listed three specific cracks, including questions about financing and monetization: First, the credit spread pressure brought by the credit financing demand of mega-cap tech companies "comes and goes, but there is indeed a certain degree of underperformance"; Second, there are still question marks on the monetization path of mega-cap tech companies; Third, the performance on the earnings side has diverged, instead of being uniform as before. However, Dominic is talking about the peak of growth rate, not that the absolute scale of capital expenditure will decline.
Wilson's conclusion is: "Once you are in this situation — the growth rate of the main story peaks, interest rates rise, credit spreads widen — it feels like a more conventional narrative. It used to be a supercharged theme that lasted for about three years. Now the risk balance is more two-sided, and the story is more conventional."
02
Earnings propped up the index, but cannot push it higher
In the past six months, the operating logic of US stocks has been a tug of war: interest rates, energy and to a certain extent credit constitute headwinds, and corporate earnings are responsible for underpinning the market. The result is that there is no net displacement of the index. The problem in the earnings season is not poor performance, but "just enough": the performance has passed the threshold, but the magnitude of upward revision of expectations is not enough for the market to break through again.
This is the most anxious point of the market at the moment. During the second quarter earnings season, investors realized that this year might be a year with exceptionally strong earnings growth, and expectations were revised up sharply accordingly. This earnings season did not reproduce this effect. Tony said: "The S&P has basically returned to 7000 points... The Nasdaq has not found a clear direction in the past six months."
"From the perspective of this quarter, earnings look decent on the surface. The threshold is very high... I think the most interesting thing is that forward earnings have not been revised up as they were during the Q1 and Q2 earnings seasons."
In his view, the full-year earnings growth will eventually be considerable, possibly close to 30% or even higher (according to the caliber he mentioned, the text record in the transcript here is not clear enough, it is recommended to check the audio before release). But this year's figures have been basically digested by the market. Tony's summary is: "Good enough to hold the line, but not enough to add fuel to let the market make another breakout."
03
The Fed: The market has priced in too much, and actual rate hikes may stop here
While the AI narrative is cooling down, the uncertainty of the interest rate path constitutes another main line. Josh Schiffrin's interpretation of the Fed's October meeting is generally dovish, and he clearly believes that the market's current pricing of interest rate hikes is overestimated.
Schiffrin's judgment is that the Fed currently believes it has the ability to "act in an orderly manner". Although the inflation situation is still severe, the policy actions in September are already producing effects. Although financial conditions have not tightened to an extreme level, they have tightened to some extent. His baseline scenario is: one more rate hike in December, and possibly another one after that, but "the final number of rate hikes is likely to be less than the market's current expectations".
His exact words are: "You have about 60 to 65 basis points of further rate hikes priced in right now. I do think they will hike rates in December, but that will likely be all. Maybe one more. But my baseline scenario, as we discussed last month, is one more hike in December."
Schiffrin also pointed out that the Federal Reserve will reduce the frequency of policy statements starting from 2027 and 2028, cutting from eight meetings a year to seven, but he does not think this will have a significant impact on the market: "Some meetings are already very close to each other, and not much happens between the two meetings, so I don't think this will have a big impact on the market."
Dominic Wilson highly aligns with Schiffrin's judgment on the interest rate issue. He pointed out that part of the reason for the previous rise in interest rates was that the 10-year forward OIS rate once touched around 5.20% — "this is already very close to the peak of the policy rate cycle in the past 20 years, and if you take this as the central point 10 years from now, you have already priced in quite a lot of risk premium". He believes that the shake-up of the AI narrative is also pushing down yields through the channel of "growth concerns".
04
Wall of macro risks: The biggest threat comes from the Middle East, the most underestimated factor is the decline in oil prices
In terms of macro risk ranking, the three discussants have an interesting divergence and complementarity: Josh Schiffrin lists the escalation of the situation in the Middle East as the biggest near-term risk, while Dominic Wilson believes that the tail risk of "no escalation" in oil prices is instead the most underestimated opportunity by the market at present.
Schiffrin's logic is straightforward: once the conflict escalates and energy prices rise again, this is the biggest threat to growth; conversely, if the situation eases and oil prices fall, "it is good for all assets, whether stocks or bonds".
Wilson cuts in from another angle. He believes that the market has already priced in quite a lot of geopolitical risk premium for Iran, and there is a real risk premium in the oil market. Therefore, "the scenario where we just do not follow up, or the escalation level is less than the market's concern, I think it is underestimated". His specific scenario is: if there is no major escalation two or three weeks after the midterm elections, "the market will find this path better than the path it was worried about, because we have already built this expectation. Even without any substantial progress, just the absence of new bad news is enough".
On the highest-conviction trade by the end of the year, Schiffrin clearly points to "long global front-end interest rates" — covering the US, Europe and the UK, on the grounds that the actual actions of major central banks will be less than what the market has priced in, and once fluctuations occur in the credit market, it will also open a downward channel for front-end interest rates.
Tony Pasquariello made a concluding description of the overall market pattern: In the past six months, macro-level interest rates, energy and credit constituted headwinds, while earnings "held the line". The S&P 500 and the Nasdaq have basically moved sideways. This earnings season "looks decent on the surface", but the upward revision of forward earnings is far less than that in the Q1 and Q2 periods, "enough to hold the line, but not enough to reignite a breakout rally". His conclusion is: if Schiffrin and Wilson's judgments are correct, the stock market may see a "nice glide path" before the end of the year, but it may be "different types of stocks" leading the gains.
This article is from the WeChat official account "Wall Street CN Max", author: Long Yue, published with authorization from 36Kr.