市场版：人工智能对股市的微妙影响

Recorded: Aug. 4, 2026 Published: Aug. 5, 2026 Host: Scott T Chronert, Head of U.S. Equity Strategy, Citi Guest: Patrick Galvin, U.S. Equity Strategy, Citi Transcript: Opening Teaser: (00:00) Research @ Citi Markets Edition. Scott Chronert (00:03) Hi, I'm Scott Chronert, Head of U.S. Equity Strategy at Citi Research. Welcome toResearch @ Citi Markets Edition, covering various topics at work within the U .S. equitymarkets. With me today is Pat Galvin, my colleague in our U.S. Equity Strategy team here at CitiResearch, and who I often refer to as the brains of the operation. I've invited Pat heretoday to talk about our cluster approach and our differentiated view of the impact on AIon the equity stock indexes. Welcome, Pat. Thanks for joining. Patrick Galvin (00:31) Thanks for having me, Scott. Scott Chronert (00:33) There's an old adage in life that what you see is what you get. But is that true when itcomes to equity investing? The simple answer is yes and no. Many financial advisors and market pundits may say that for U.S. equity exposure, justown the S&P 500. There's a lot to this, and we don't disagree. But what is critical to thisview is that you have to know what it is you're buying. The focus of today's podcast is to dive into this and explain just how the S&P 500 hasevolved over time and what that means for broader exposure to U.S. equities and the AItheme in general. Quite simply, the high-level takeaway is that the index has undergone a major evolutionover the past 30 years. We address this by dividing the index into what we call clusters.Pat, can you step in here and give us a quick explanation of this approach and withsome immediate observations from this? Patrick Galvin (01:26) Yeah, we've been using this cluster approach now for over three years and have foundit to be a very helpful mechanism to break down the market. More helpful thana traditional Value vs. Growth style breakout. And we can go into that. But the beauty of the industry-group cluster approach is the simplicity. We just assigneach S&P 500 industry group into one of three buckets: either Cyclicals, Defensives, orGrowth. And we do this based on the relative macro sensitivity, as well as exposure tosecular or thematic fundamental growth drivers. Somewhat subjective in the bucketing there, but for example, we'd put semis in theGrowth cluster, we'd put utilities in Defensives, we'd put energy in Cyclicals, just to givean example there. And what this gives us is three market-cap-weighted slices of the index that togetheradd back-up to the index. And why that's important is it really helps us do attributionwork, decomposing both index-level performance as well as index earnings growth onestep below the index. Helps us look at what's happening under the index surface. It's also helpful to understand valuations relative to history using the industry-groupindices to build this approach. We can create very long histories of valuations,fundamental series, which just gives us a great perspective to decompose the marketand find out what's happening under the index surface. Scott Chronert (02:43) Great. So the high-level takeaway of this work is that when we look at this Growthcluster — which we would say, generally speaking, is going to be that component of theindex that is AI-influenced — it comprises roughly 55% of the index. Now, this is up fromless than 20% 30 years ago. So the high-level takeaway here is that as the market has evolved over the past 30years, the Growth influence on the index has significantly changed. And so, what youget from an incremental and higher Growth component is a smaller influence of bothCyclicals and Defensives. And this is where for us it begins to get pretty interesting. So Pat, as we go throughthis, let's start adding a little bit of perspective here on this Growth cluster, howit's evolved and how you see it and we see it shaking through in terms of the indexbehavior. Patrick Galvin (3:35) Yeah, sure thing. I mean, something we've covered at length in our work is really, as you mentioned, thechanging index composition through time. I think the cluster has given you a really greatlook at this. As you mentioned, 30 years ago, that Growth cluster cohort was about 17% of the index market cap. Now it's like 55%. So we've seen that Growth weight in theindex triple over 30 years and nearly double in just the last 10 years. And with that comes a number of implications. I would say first, it's reduced cyclicality ofthe index. Those Growth names, they carry a little bit less macro sensitivity when we'reeither looking at macro sensitivity of returns or earnings growth. And really this 20-percentage-point decline in the Cyclicals weight that's moved over to Growth, just reallysetting up for a little bit less macro-sensitive index. The second part I would say is index valuations certainly need context given thiscomposition shift, something we've talked about at length, but really, it's not