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, after four years of grinding, is growing impatient to look ahead.

In a recently recorded episode of Goldman Sachs' "The Macro Call" podcast, Chief U.S. Equity Strategist Ben Schneider and TMT sector expert Pete Callahan engaged in an in-depth discussion covering the evolution of AI trade structure, market breadth, key takeaways for the Q3 earnings season, and the year-end outlook.

During the program, both Goldman Sachs representatives noted that the resilience the stock market has shown amid elevated interest rates is fundamentally supported by strong earnings led by large-cap tech stocks—and the AI theme itself is undergoing a profound "narrative migration."

AI Trade Moving 'Up the Stack': From Selling Shovels to Selling Water

Pete Callahan directly pointed out the core logic shift in the current market. He stated that around 2027 capital expenditure, the market has already formed a healthy preliminary consensus, which provides roughly 12 to 15 months of visibility.

But the question follows: what happens after 2027?

"That's why you're seeing multiple compression in the semiconductor sector—the market is asking whether this level of high profitability is sustainable. Are companies over-earning or under-earning? And what does the roadmap look like beyond 2027?" Callahan said.

Against this backdrop, the market's attention is beginning to migrate toward the "application layer." Callahan noted that the "inference economy" and tool companies that help enterprises and consumers deploy AI are drawing interest, specifically including areas such as cybersecurity and data infrastructure.

"I think the trade has moved up the stack to some extent," he said, "but I prefer to see it as an 'and' rather than an 'or.' Semiconductors are still within striking distance of all-time highs, and the market has found room for both directions."

'Coding AI Is No Longer a Secret': Are Personal Agents the Next Big Thing?

In the Q3 earnings season outlook segment, Callahan offered a widely noted observation:

"The market is very hungry for the next story. Coding AI is no longer a secret in the market—it's already well known as a major AI application. So how big can the personal agent story get, and how fast can it arrive? And the ripple effects across the entire consumer supply chain that come with it—I think that will be very much worth watching."

He added that during earnings season, commentary from S&P 500 companies on cost savings, new revenue, new products, and new businesses will be key signals the market uses to sustain the AI narrative.

On the question of return on invested capital (ROIC) for large hyperscalers, Callahan said the market felt good about it after Q2 earnings and will seek positive validation again in Q3. Additionally, if semiconductor companies can provide visibility extending to 2028, it would help alleviate the biggest overhang currently suppressing valuations.

Market Breadth Narrowest Since the Dot-Com Bubble, But Year-End Still Fundamentally Bullish

Moderator Mark Wilson mentioned that market breadth is currently extremely narrow. Chief U.S. Equity Strategist Ben Schneider responded candidly:

"The median S&P stock is more than 15% below its high. Our preferred measure of market breadth is already at its narrowest level since the dot-com bubble."

However, he believes that based on historical patterns, the repair of narrow market breadth tends not to come through "catching down" but rather through "catching up." "Most investors think back to 2000 and assume it must end with a decline, but in reality, history more often shows convergence through chasing rallies."

On the pace of AI capital expenditure growth, Schneider offered a key forecast: current market consensus expects earnings growth of about 30%, but this will slow as we enter 2027—"not only because of macroeconomic pressures, but also because while AI capital expenditure will continue to grow, it won't maintain the same rate of growth."

On the election factor, Schneider noted that compared to previous election years, investor attention this cycle is noticeably lower, primarily due to the AI narrative, the macroeconomic backdrop, and the consensus that "major policy changes are unlikely in the next year or two." Combining three factors—strong earnings, light positioning, and reasonable valuations—his base case is: the market moves higher through year-end.

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 delighted to have with us our Chief U.S. Equity Strategist from our research division, Ben Schneider, and our TMT sector expert from our markets business, Pete Callahan. Gentlemen, thank you very much for joining.

Ben Schneider: Thank you, Mark.

Mark Wilson: So over the past few weeks, a lot of the equity story has really been a rates story. The path of rates has been front and center. And yet, equities have actually traded really well. Ben, what's going on here? How sustainable is this equity strength or resilience in the face of this rates move?

Ben Schneider: First of all, we should acknowledge that while the S&P 500 is near all-time highs, it has to some extent already reflected the impact of higher yields, with valuation multiples coming down quite a bit over the past year. If you look beneath the surface, the more cyclical stocks, the more rate-sensitive areas of the equity market, have clearly declined as yields have risen. That said, I think the simple reason for this resilience is strong earnings, particularly earnings related to the largest tech stocks tied to the AI trade. It's not that these stocks are completely immune to higher yields, but to a large extent, they are less sensitive than most of the rest of the market.

Mark Wilson: So I've been saying for some time that micro has been driving macro. Ben mentioned this. There's still genuine AI enthusiasm, Pete, but the nature of interest and price action has changed. Some recent product successes, especially when you think about products like Muse, have genuinely reignited some new enthusiasm. The story has changed from the first half of the year, when price action was really all about the beneficiaries and recipients of that massive capital expenditure. Right. What are you looking at, and which themes within AI are you most focused on right now?

Pete Callahan: Yeah, Mark. So we're definitely starting to see, I think, some healthy consensus forming around calendar 2027 from a capital expenditure perspective. That's good because it means there's visibility for roughly the next 12 to 15 months. It also means the market has slightly less visibility beyond that, and I think that's why you're seeing multiple compression in the semiconductor sector—because the market is saying, how durable is this? Are companies over-earning? Or under-earning?

You could even make that argument, if you will. And how do you think about the roadmap beyond calendar 2027? So that's driving some of the multiple compression in semis. As that happens, you're starting to see the market get excited about things higher up the stack. So think about the inference economy, companies and tools that help deploy AI on both the enterprise and consumer side. So things like cybersecurity or data infrastructure.

So I think the trade has moved up the stack to some extent. I do think it's more of an "and" than an "or." Semiconductors are still in. Just a step away from all-time highs, so I think the market has found room for both in this trade and this theme, but you're definitely maturing to some extent—you're in year four of the AI theme, and the market continues to look forward and search for clues about calendar 2028.

Mark Wilson: And Ben, the equity market has been trading with what people keep calling very poor breadth. The narrowness is striking. How nervous are you about this, and how nervous should we be? When you look under the hood of the equity market, what themes are you most focused on, and what are you most excited about?

Ben Schneider: It's striking—the median S&P stock is more than 15% below its high. Yes, our preferred measure of market breadth is the narrowest since the dot-com bubble. That's quite extreme. I would say the easy bet there is that at some point, there will be reconvergence. The hard part is that timing it is very difficult. And judging the direction of that reconvergence is also very difficult. Most investors look back at 2000 and think it must be resolved through downward catch-up.

But in reality, narrow breadth is more often resolved through upward catch-up. So I think the takeaway for investors is that you want to own a bit of everything, you want to stick with some of the AI themes Pete discussed. You might also want to look at some parts of the market that have actually been quite beaten down by macro pressures recently.

For example, we've been highlighting consumer experience stocks. They're trading at fairly low multiples. If you get a little relief on rates or on oil, I think they can catch up nicely. And we're in the first week of October, with a big wave of earnings starting next week, which will be the market's focus over the coming weeks. As we head into Q3 earnings season, what are you most focused on? And then obviously, the calendar year will turn, and people will start talking about 2028 valuation multiples. As those key themes, those conversations begin, what are you thinking about?

Ben Schneider: Well, going back to the earlier point about rates and valuations, what has really been supporting the market is earnings. We need to see signs that earnings strength will continue. Consensus expectations, which I think are correct, are for earnings growth of about 30%. That's pretty good. But it looks like as we head into 2027, we'll see some deceleration.

That's true both because of the macro forces we've been discussing and because AI capital expenditure, while it 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 going back to the earlier discussion, as investors focus not just on AI capital expenditure but on the rest of the AI economy, getting more data there will be very critical.

Mark Wilson: And Pete, as you think about earnings and the big themes we'll be hearing from tech companies, what are you most focused on? And again, how does that tie into the themes you're watching for next year?

Pete Callahan: Yeah, a few things that align with what Ben said, but I think the first focus again is probably the return on invested capital for these large hyperscalers, right? That's still the core engine of this market. I felt like the market felt good about it coming out of Q2, so we'll be looking again for positive reinforcement from the large hyperscalers.

I think the other thing to consider is visibility—you're starting to hear some semiconductor companies talk about visibility into calendar 2028. That's the bridge the market is, I think, trying to resolve. And that's a big overhang on semiconductor multiples broadly. So companies that have 12, 15 months of visibility and can say something about what might be coming in 2028—that will be a focus. I think the market is very hungry for the next thing, right? Everyone knows coding. As a major AI application, it's no longer a secret in the market.

So, how big and how fast can this personal agent story get, and then its impact up and down the entire consumer stack—I think that will be very interesting to watch. And then finally, of course, related to AI, just trying to think about any tidbits you get from S&P 500 companies about cost savings, about new revenue, about new products, about business formation. I think those are the things the market is really hungry for to continue supporting the AI trade.

Mark Wilson: Finally, Ben, we're less than a month away from the midterm elections. I guess the real question is, do elections matter? More substantively, as you think about the period from now to year-end, the historical analogy that XYZ actually trades quite well after midterms is deeply ingrained and well covered. How much should investors be thinking about that playbook?

Ben Schneider: Again, I think it speaks to the backdrop that you're only asking about it now. Normally in an election year, a month out, we'd be talking about it constantly. Yes, investor attention this cycle is much lower. I think it's both because of AI, because of the macro backdrop, and because of the consensus view that we're unlikely to see major policy changes over the next year or two. That said, as you mentioned, the clearest historical dynamic is that a risk premium builds ahead of elections, and that risk premium actually starts to ease around now, in October. I think there are obviously a lot of moving parts at the moment. But if I look at the underlying earnings strength we've been talking about and combine that with fairly light investor positioning and fairly reasonable valuations, I do think the base case should be for the market to be higher by year-end.

Mark Wilson: Ben, Pete, thank you very much. Thank you for joining the latest episode of "The Macro Call," and remember to tune in to the weekly edition, available to everyone on the Goldman Sachs Marquee platform.