Goldman Sachs Discussion: AI trading is shifting from hardware to applications, and the market is eagerly anticipating the "next AI story."

Author: Bu Shuqing, Wall Street Watch

The AI investment narrative is quietly shifting gears------from the chip arms race to the "inference economy" and personal agents, Goldman Sachs believes the market has been honing its edge for four years and is eager to look ahead.

In a recent episode of the podcast "The Macro Call," Chief U.S. Equity Strategist Ben Schneider and TMT sector expert Pete Callahan from Goldman Sachs engaged in an in-depth discussion about the evolution of AI trading structures, market breadth, highlights of the Q3 earnings season, and year-end market trends.

During the program, both Goldman Sachs representatives pointed out that the resilience shown by the current stock market amid high interest rates is primarily supported by strong earnings, particularly from large tech stocks------and the AI theme itself is undergoing a profound "narrative shift."

AI Trading "Shifting Up": From Selling Shovels to Selling Water

Pete Callahan directly pointed out the core logic transformation in the current market. He stated that there is a healthy consensus forming around capital expenditures for 2027, providing visibility for about 12 to 15 months.

But the question arises: What happens after 2027?

"That's why you see the valuation compression in the semiconductor sector------the market is asking whether this high profitability can be sustained. Are companies over-earning or underestimating their profit potential? What does the roadmap look like after 2027?" Callahan said.

In this context, the market's focus is beginning to shift towards the "application layer." Callahan noted that companies and tools that help businesses and consumers deploy AI, particularly in areas like cybersecurity and data infrastructure, are gaining attention.

"I think trading has shifted up to some extent," he said, "but I prefer to see it as a 'both-and' rather than an 'either-or.' Semiconductors are still within reach of their historical highs, and the market has found space for both directions."

"Programming AI is No Longer a Secret": Are Personal Agents the Next Big Thing?

In the outlook for the Q3 earnings season, Callahan made a widely noted judgment:

"The market is very eager for the next story. Programming AI is no longer a secret in the market; it is already well-known as a significant AI application. So, how big and how fast can the personal agent story be? And the chain reaction it brings to the entire consumer-side industry chain will be very worth watching."

He also added that during the earnings season, statements from S&P 500 companies regarding cost savings, new revenues, new products, and new businesses will be key signals for the market to support the sustainability of the AI narrative.

Regarding the return on investment (ROIC) for large hyperscale cloud service providers, Callahan stated that the market felt good about this after the Q2 earnings reports, and it will seek positive validation again in Q3. Additionally, if semiconductor companies can provide visibility extending to 2028, it will help alleviate the biggest suspense currently suppressing valuations.

Market Breadth is the Narrowest Since the Internet Bubble, but the Year-End Outlook Remains Bullish

Host Mark Wilson mentioned that the current market breadth is extremely narrow. Chief U.S. Equity Strategist Ben Schneider candidly stated:

"The median S&P stock is over 15% below its peak. Our preferred indicator for measuring market breadth is at its narrowest level since the internet bubble."

However, he believes that historically, the direction of recovery from narrow market breadth is often not "catching down," but rather "catching up." "Most investors looking back at 2000 think it must end with a decline, but in reality, history shows that it often converges through chasing up."

Regarding the growth rate of AI capital expenditures, Schneider provided a key prediction: the current market consensus expects earnings growth of about 30%, but there will be a slowdown entering 2027------"not only due to macro pressures but also because while AI capital expenditures will continue to grow, the growth rate will not maintain the same pace."

On the election factors, Schneider believes that compared to previous election years, investor attention is significantly lower this time, mainly due to the AI narrative, macro background, and the consensus that "there is unlikely to be significant policy changes in the next year or two." Considering the three factors of strong earnings, light positioning, and reasonable valuations, his basic scenario is: the market will rise before the end of the year.

The full transcript of this episode of "The Macro Call" is as follows (translated with AI assistance):

Mark Wilson: Hello everyone, welcome to the latest episode of "The Macro Call." I’m Mark Wilson. I’m pleased to invite our Chief U.S. Equity Strategist Ben Schneider from our research department, and TMT industry expert Pete Callahan from our market business. Thank you both for joining.

Ben Schneider: Thank you, Mark.

Mark Wilson: So, in the past few weeks, many stock stories have really been interest rate stories. Interest rate trends have been at the core. Meanwhile, the stock market has actually been performing very well. Ben, what’s going on? How sustainable is this strength or resilience in the stock market in the face of this round of interest rate trends?

Ben Schneider: First, we should acknowledge that although the S&P 500 index is near historical highs, it has somewhat reflected the impact of higher yields, with valuation multiples declining significantly over the past year. If you look internally at the market, those more cyclical stocks and sectors that are more sensitive to interest rates have clearly declined as yields have risen. That said, I think the simple reason for this resilience is strong earnings, particularly from the largest tech stocks related to AI trading. It’s not that these stocks are completely unaffected by higher yields, but to a large extent, they are less sensitive than most other stocks in the market.

Mark Wilson: So I’ve been saying for a while that the micro has been driving the macro. Ben pointed this out. There is still real enthusiasm for AI, Pete, but the nature of interest and stock price movements has changed. Some recent product successes, especially when you think of products like Muse, have indeed reignited some new enthusiasm. The story has changed from the first half of the year, when stock price movements were really all about the beneficiaries and recipients of that massive capital expenditure. What are you looking at, and what themes in current AI are you most focused on?

Pete Callahan: Yes, Mark. So we are definitely starting to see, I think, a healthy consensus forming around capital expenditures for calendar year 2027. That’s good because it means there is visibility for about 12 to 15 months ahead. It also means that the visibility for the market after that is somewhat lower, and I think that’s why you see the multiples in the semiconductor sector compressing, because the market is asking how sustainable this is. Are companies over-earning? Or are they underestimating their earnings?

If you will, you could even make that kind of argument. And how do you think about the roadmap after calendar year 2027? So that has driven some multiple compression in the semiconductor sector. As this happens, you start to see the market getting excited about higher things in the stack. So think about the inference economy, companies and tools that help deploy AI and artificial intelligence on the enterprise and consumer side. So these are things like cybersecurity or data infrastructure.

So I think this trading has shifted up to some extent. I do think it’s more of an "and" rather than an "or." Semiconductors are still in the game. They are just a step away from historical highs, so I think the market has found space for both in this trading and theme, but you are definitely maturing to some extent. You are in the fourth year of the AI theme, and the market continues to look forward and seek clues for calendar year 2028.

Mark Wilson: And Ben, the stock market has been trading with what people keep mentioning as very poor breadth. This narrowness is astonishing. How concerned are you about this, and how concerned should we be? When you look under the hood of the stock market, what themes are you most focused on, and what excites you the most?

Ben Schneider: It’s astonishing, the median S&P stock is over 15% below its peak. Yes, our preferred market breadth indicator is at its narrowest level since the internet bubble. That’s quite extreme. I would say the easy bet there is that at some point, there will be a re-convergence. The hard part is timing that very difficult. And judging the direction of that re-convergence is also very difficult. Most investors looking back at 2000 think it must resolve by catching down.

But in reality, narrow breadth more often resolves by catching up. So I think the conclusion for investors is that you want to have a bit of everything, want to stick with some of the AI themes Pete discussed. You might also want to look for some parts of the market that have indeed been heavily suppressed by macro pressures recently.

For example, we have been emphasizing consumer experience stocks. They trade at quite low multiples. If you get a little relief on interest rates or oil, I think they will catch up well. And we are in the first week of October, with a lot of earnings reports starting next week, which will be the focus of the market in the coming weeks. As we enter the Q3 earnings season, what are you most focused on? And obviously, as the calendar year changes, people will start talking about 2028 valuation multiples. When these key themes and conversations start, what are you thinking?

Ben Schneider: Well, going back to the earlier point about interest rates and valuations, what has really supported the market has always been earnings. We need to see signs that strong earnings will continue. The consensus expectation, I think, is correct, is earnings growth of about 30%. That’s quite good. But it looks like as we enter 2027, we will see some deceleration.

This is true, both because of the macro forces we have been discussing and because AI capital expenditures, while they should continue to grow, may not continue to grow at the same pace. So the question is how mild or steep that deceleration will be? And then back to the earlier discussion, as investors focus not only on AI capital expenditures but also on the rest of the AI economy, getting more data there will be very key.

Mark Wilson: And Pete, when you think about earnings and the major themes we will see from tech companies, what are you most focused on? Furthermore, how does this tie into the themes you will be watching next year?

Pete Callahan: Yes, there are a few things that align with what Ben said, but I think the first focus again may be the return on investment capital for these large hyperscale cloud providers, right? That’s still the core engine of this market. I feel like the market felt good about this after the second quarter, so we will be looking for positive reinforcement from the large hyperscale cloud providers again.

I think other things to consider are visibility; you start to hear some semiconductor companies talking about visibility into the calendar year 2028. That is, that is the bridge that the market is trying to solve, in my opinion. And this is a big overhang on semiconductor multiples broadly. So those companies that have 12 to 15 months of visibility and can articulate what might happen in 2028 will be the focus. I think the market is very eager for the next thing, right? Everyone knows about programming. As a large AI application, it is no longer a secret in the market.

So, how big and how fast this personal agent story can grow, and its impact on the entire consumer stack, I think will be interesting to observe. And finally, of course, related to AI, just trying to think about any snippets you get from S&P 500 companies regarding cost savings, new revenues, new products, and corporate formations. I think these are the things the market is truly eager for to continue supporting AI trades.

Mark Wilson: Finally, Ben, we are less than a month away from the midterm elections. I think the question really is, are elections important? More substantively, when you think about the time from now until the end of the year, XYZ has indeed traded quite well historically after midterm elections, and that analogy is deeply rooted and well-covered. To what extent should investors think about that script?

Ben Schneider: Furthermore, I think this illustrates the context; you only just asked it now. Typically, in an election year, the month before, we would be talking about it constantly. Yes, investor focus is much less in this cycle. I think it’s both because of AI, due to the macro backdrop, and because of the consensus view that we are unlikely to see significant policy changes in the next year or two. That said, as you mentioned, the clearest historical dynamic is the accumulation of risk premium before elections, and that risk premium actually starts to ease around now in October. I think there are clearly a lot of moving parts at the moment. But if I look at the potential earnings strength we’ve been discussing and combine it with relatively light investor positioning and fairly reasonable valuations, I do believe the baseline scenario should be that the market is higher by the end of the year.

Mark Wilson: Ben, Pete, thank you very much. Thank you for joining the latest edition of the Macro Call, and remember to tune in to the weekly version, which everyone on the Goldman Sachs marquee platform can listen to.