I'm Patrick O'Shaughnessy, the capital markets technology analyst here at Raymond James. Up next we have S&P Global. On their behalf, we have CEO Martina Cheung and CFO Eric Aboaf. Thank you guys very much for joining us today.
Thank you.
To kick it off, and as you might expect, I have several questions that pertain to AI to get us started. I think almost everybody would agree that your benchmark businesses, S&P Ratings, S&P Dow Jones Indices, Platts, they're not really at risk definitionally from competitive disruption. I do think a very active debate currently does exist around some of your workflow tools, with some arguing that AI can cheaply replicate those tools that sit on top of client data or non-proprietary third-party data. How do you respond to those concerns as they would apply to S&P solutions?
Yeah. Thanks, Patrick. It is great to be here. As you said, we have an incredible breadth and depth of our proprietary content, whether it's benchmarks, unique IP and workflows. If you take just the benchmark businesses that we have, so index ratings or price assessments, and you add in the revenue that we get from the redistribution of our ratings IP that's in market intelligence, just that alone makes up about 3/4 of our operating income at S&P Global. Really the vast majority of what we do very tied to these areas. In terms of the workflows that we have, and we mentioned during IR Day that about 13% of our revenue comes from these workflows, and many of them are incredibly sticky and very unique and perform critical tasks for our customers.
I would differentiate what we do in a couple different ways. The first I would say is, when we're actually running these workflows, many of them are what we would characterize as enterprise-grade critical systems of record. Think WSO, Debtdomain, ClearPar, iLEVEL, Capital IQ Pro, Platts Connect, et cetera. These services, in many ways are actually moving our clients' decision-making. Oftentimes they're actually facilitating the flow of capital and supporting critical decision-making on asset allocation, for example. We also see that many of these workflows are actually, I would say, in many ways, performing regulatory tasks for our customers.
For example, we will, as part of some of our managed services contracts, we will also make sure that our customers are in position to comply with very complex regulations like the DORA regulation in Europe, for example. I think that's the way to sort of think about these particular platforms. We also have a couple of things that are quite unique amongst these platforms. The first is that many of them are part of ecosystems, so they're not just supporting one client and one client's systems. They're also supporting an entire industry sector. Think WSO, for example, or iLEVEL. Many times you can't really realize the full benefit of the actual product without our unique data that comes with it.
Sometimes that data is generated by the flow of work in the product, and other times it is the data that actually triggers the value of the product for our customers. Think loan reference data in WSO. You know, these things are, I think a little bit. You know, we'll put it this way, not all software is the same in terms of the level of criticality that it represents for our customers. You know, when I think about what our customers are telling us, because at the end of the day, that is the best barometer for the need and demand and value that these products will create over a period of time.
They're essentially saying, "Look, we want fewer vendors, not more." They're saying that our unique IP is incredibly important to them, including the software that we run as part of that value proposition. They're asking us to co-create with them in the case of our larger clients with Kensho, for example, to create additional value. You know that we've been integrating generative AI across these platforms, including using generative AI ourselves, and so we have a very informed view around how we can continue to create value for our customers through that. Maybe just as We'll throw that out as a starting point. I know you have a lot more questions on AI.
Yeah. I think, you know, at the same time as, you know, okay, your workflow tools, they're more mody than people might expect. You've also have argued in the past that, you know, you think AI will only increase the value of your proprietary data. What are some of the early examples of that playing out?
Yeah. Well, we're very excited about it, and I would say that we've seen this play out in a few different areas. Think about, perhaps the ways in which we've deployed the actual capabilities ourselves. We've been in the market now for around about a year with Cloud for Financial Services. If you remember last year, we were one of the first to announce our MCP connector in Cloud for Financial Services. We also were very early in the market, again around about a year ago, with providing our energy content through Microsoft Copilot. Between those two, we see about, we have about 140 customers, between those two. Around 80 of them are actively using the MCP connector, largely, market intelligence.
We've got a chunk of those who are either paying for it, some of them who are on pilot timelines and actually converting to paying as well. About 80 customers using our MCP connector, largely MI. With the energy division where we were very early in the partnership with Microsoft Copilot, we have around 60 customers who are using Microsoft Copilot and have paid to use our content in Microsoft Copilot as an add-on. This begins to give you a sense for the value that we're generating for our customers by actually providing these different channels.
When we talk about, you know, the value of our data, every time our customers use these products, whether it is a partner product or it is our product specifically, like a Document Intelligence or a Chat AI on Platts Connect, they're essentially pulling in more of the data that they're licensed to every time they use it. We track all that telemetry because the usage of our content, the frequency and the volume and coverage of usage, that all forms part of the conversation around value creation for the customer when we're renegotiating. It also gives us an opportunity to position new data sales.
Our CCO is really the tip of the spear with some of our more sophisticated clients who recognize that they can actually scale these capabilities to use more data and get more signals, get more real-time insights, and we're having a lot of conversations with them around new potential data sales as well. You know, the other point that I would make is, I made the point earlier that we're making our products smarter and integrating generative AI into our own solutions. You know, we know the clients are willing to pay for that as well. We mentioned during Investor Day that almost 20% of our clients have signed up for automated data ingestion at iLEVEL. That is an additional entitlement that they're paying for. Again, speaks to the value that we have here.
Maybe, you know, one other point that I would make is if you look at the energy customers who have signed up for receiving our energy content through Microsoft Copilot, the growth rate for ARR for those customers about 2x the rest of our energy base, and the retention for those customers is about 4% better than the rest of our energy base. We're really seeing a lot of value and the beginnings of it, particularly where we have about a year now of experience with customers going through these channels and using our content more and more.
As your content gets distributed through LLMs like Claude and Anthropic, like you mentioned, how confident are you that you're going to be able to defend your intellectual property indefinitely?
Patrick, let me take that because it's an area that, you know, we've always worked on, right? We've had proprietary data and benchmarks and price assessments and so and so forth for decades and decades. So this is just a new area to expand our capabilities. First, I'd remind you and everyone that our relationships are intrinsically with the client, the asset manager, the bank, the corporate, the energy company. We have a direct relationship with them in terms of what's the data that they're procuring from us, what platform are they using, what are the reuse rights, what are the pricing schedules, and so on and so forth. That's a bilateral relationship, and that one is one that we retain under all circumstances.
When a client goes and says, "Look, we're gonna start to take advantages on the feature functionality, the AI feature functionality that Martin, you know, described within our ecosystem," then they continue with those relationships. When in addition, they may go to, you know, one of the vertical AI providers or one of the LLMs, right? Those end users need to actually before they can access S&P data, actually get us to turn on, right? To basically permission them. That's done through the MCP connector, through the grounding agent, and so forth. There's a whole set of technical tools for permissioning, accessing, and then for us to monitor, right, the volume of data, what they're using, how they're using it, who's using it on the other side.
In fact, that's actually the core, you know, protective feature. Finally, we, you know, our data is for our clients to use, sometimes directly with us, sometimes with the LLM. Our data is not available for training, right? That's not the purpose of our data. Our data is there as a proprietary data set that clients can use for their purposes. You know, training is in a different place. Training is done by the LLMs, and, you know, that's up to them. We're not going to, you know, let the unknown be behind our paywall.
I would maybe add that, obviously we have clients whose appetite for data has been increasing, and the combination of our capabilities, whether it's the MCP connector, our grounding agent, and our AI-Ready Data, we actually have the ability to essentially throttle the number of rows that actually cross over between our databases and that show up in the client's queries. In other words, you know, if you're querying multiple of our datasets, you don't get the whole dataset back, right? You get, our data continues to sit on our servers, and you get through the grounding agent just the data needed to answer the query that you're getting. That throttling really helps with the IP protection as well.
Going back to your November Investor Day presentation, there was a slide that kind of categorized your different revenue buckets. You know, there was a, you know, a top bar that said, like, less than 5% of your revenues was undifferentiated, non-proprietary data. It's not all that material, but I'm just curious, like, how do you go about protecting even that sort of content?
Yeah, I mean, look, this is, this is sort of table stakes expected as part of the customer experience. Our customers really do just expect to get that. We don't charge separately for it. I mean, I'm not aware of any customer that pays separately for any of these more commoditized datasets. Essentially, they're priced as part of the enterprise license. The only point I would make in addition to that is we did say that we'd expect the 95% to grow over time relative to the rest. Over time, as proportioned, you'd see that 5% decrease.
Very helpful. Thank you. Another angle to the discussion is the impact that AI can potentially have on your profit margins. I think the bear case is that maybe new competition is gonna force increased investment. The bull case is that AI structurally improves your incremental margin profile of many businesses. How do you think it all nets out?
You know, in our minds, and we described this at Investor Day, this is a net positive for us in a, in a number of ways. You know, first you've seen us invest early in AI, so this is not new, right? We bought Kensho back in 2018 and been systematically building out their capability. What's important, we have always been in the AI ecosystem, which is what makes our, you know, partnerships with Microsoft Copilot or OpenAI or Anthropic natural for us to secure and to secure it in in-depth ways. Secondly, we've been rolling out AI area by area and not just from a, like, a generalized usage perspective 'cause that's not enough, right?
What we've done is gone deep, you know, we've tackled four or five areas where there are large swaths of activity and staff in our ecosystem and said, "How do you do those fundamentally differently?" You know, a good example that we shared at the last earnings was around our Enterprise Data Office or EDO, covers about 9,000 of our 40,000 people. It's about a half billion dollar of expense spend on a seven and a half billion dollar base. We see by the end of 2027 a 20% save in expenses there. It's really a combination of end-to-end process reengineering, right? Because for 9,000 folks, it's not about giving them an LLM.
It's actually dividing up what their processes, sub-processes are and so forth, thinking about how to simplify, standardize, tool them differently. We've been doing that for years with a set of analytic models and machine learning models and neural networks, and now we have a new set of large language models that we can leverage as well. That's an example, but that's one of several examples. It's probably the one that's furthest along. You know, we're working with our software developers, another pool of about 7,000 staff. In truth, you've all seen the benefits you can get from software coding. You know, we're already at benefits of 10%+ in terms of productivity.
We see ways to double, triple that amount of productivity, just as we think about what a Scrum team looks like, how agents can factor into that and be part of that team, how the teams don't need to be the same size, how the software development life cycle actually becomes integrated with the product development life cycle, and so on and so forth. There are areas that we are just marching across the company on to really think about how do we transform the way we work. It's a mix of new tools and end-to-end process reengineering. There's researchers, there's analysts, there's sales operations. There's area after area, which in our mind, you know, easily comprise about half of our staff.
You think about comp and benefits, it's closer to two-thirds of our cost base, right, out of $7.5 billion. That means there's a $5 billion pool of expenses that we can tackle and think about how do we systematically change and transform and make more efficient and more effective at the same time. I think that's why, in effect, we were very clear at Investor Day, and we're happy to reaffirm. We see it that we can comfortably deliver on the 50 to 75 basis point margin expansion each year, and we think there's even more upside in years of cyclical outperformance. That's because we have these programs in place.
We see the next three or four programs in front of us, they cover so much of our staff in really significant ways. In a way, I think it's not just the bull or the bear case. The actual case is AI pays for itself, and it pays for itself multiple times over and delivers that margin expansion. You know, margin expansion doesn't come, you know, on a perfectly smooth line, right? You know, we had nice margin expansion in fourth quarter. First quarter, we expect margin expansion again, but it'll vary. For example, in first quarter, we expect margin expansion of about 50 basis points, both in ratings and in market intelligence.
Then, relatively muted margin expansion in some of the other divisions as we go through some seasonal patterns, productivity, and investments. But we feel comfortable with margin expansion on a full year basis, you know, 50-75 basis points, margin expansion in that range for every business, because that's, you know, that's what we've been driving towards, and those are the plans that are in place and are, you know, we already see coming through with action and results.
Very helpful. With the ratings business and kind of maybe wrapping up the discussion around AI, how do you see AI spend broadly in data center build out specifically have an impact on ratings this year and beyond?
Yeah, I can take, maybe start, and Eric, you can add in as well. It has been quite positive for us. I mean, particularly we saw in Q4 some outsized issuance around the data center spend and some of the announcements from the hyperscalers around their intentions of CapEx. I think what's important for us is that we're in the market. We have very well-developed and robust methodologies, and we have deep expertise that we've developed over several years on this. We find ourselves to be quite well-positioned, I would say, in terms of being able to rate the issuance when it comes. The only, you know, point that I would make, and you know that we are always thoughtful about issuance. We've seen about $700 billion worth of CapEx announced.
The truth is, we don't know how much of that will materialize in 2026, and we don't specifically know for a fact how much of that will actually be funded by debt in 2026 or under what structures. You know, we are very prudent in terms of how we think about it. I would say over the medium term, having outsized growth in data center issuance is not required for us to develop our medium-term plan for ratings.
I'd just add the, you know, the data center hyperscale or project finance because it's a mix of investment grade and project finance. We'll have to see how it plays out. You know, it's worth about 2 points of revenue growth for us in the ratings business over the last year. It could easily be 1 point, 2 points, 3 points, 4 points, just depending on how that market evolves, what we see. We've been prudent, as we said, at earnings around describing our issuance activity. We're careful, and we really wanna just see how it plays out.
We just don't know if it's gonna zigzag a little bit or not, and we'd rather see it come through, see it in first quarter then second quarter, and then, you know, appropriately adjust our estimates if if it's another strong year.
I think, Eric, one of the things that I think we mentioned at IR Day on this one, as well, is that it is possible, depending on how much of that CapEx comes to market, that we could see build issuance improve in the, you know, kind of 2-ish%, 3% range, depending again on how much of that CapEx comes to market that's debt funded.
A lot of upside.
Sticking with ratings, you know, another current event is private credit markets are in the news. Just broadly speaking about private credit markets, where do you think we stand in terms of their evolution relative to the public markets and then S&P's ability to build out a franchise in the private market space?
I can start, and Eric, please chime in. I think maybe start with ratings, which I think is where you're going with your original question, and I can kind of expand around that. We've obviously been working a lot, and you all have asked us many, many questions on private markets and how we think about that for ratings. It has been really a phenomenal growth driver for us, and we'd expect that to continue to grow. It comes not just in the form of, you know, private credit specifically where it originated around direct lending, but also, of course, in asset-backed financing, energy infrastructure, data center deals, alternative investment fund ratings, and things of that nature. We see really a very, I would say, diversified business that has developed within ratings.
As you know, we are very, very specific about using the same ratings methodology, whether something gets rated in the public market or the private market. We've always viewed that as incredibly important. I would say that it is even more important as we see these very large issuances that could have a public piece and a private piece, for example, in the hyperscale area. That's been really important for us, and we'd expect that to continue. I would say outside of ratings, if you want to sort of like take a bigger step back, I mean, the health of the private credit and more broadly private market space is strong. You know, we saw about an 11% increase year-over-year in fundraising in 2025 versus 2024.
One of the things that is very interesting to us about that increase is that we're seeing a bit of a rotation so that it's not as much U.S., but we're seeing quite a bit of interest in Europe and across Asia. As a global player, we are very well positioned to take advantage of that. Just to give you a sense, almost 70% of the fundraising in 2025 went to Europe funds and multi-region funds. This is where the strength of the global franchise that we have is very important, whether it comes to actually rating that debt or even working with sponsors and investors who want information around those funds and those markets. This is where the acquisition of With Intelligence comes in as being incredibly valuable for how we serve the overall market with this as well.
There's plenty that we will do there around generating those synergies based on that very unique set of content that With Intelligence brings to us on top of the great private company data, BigDough data, and other data that we have also. Perhaps the last point I would make is, we do see an increasing opportunity around indices, whether it's public-private, private stock indices, private credit indices, and we're continuing to innovate around that. You'll see us launch new products. You've seen that already, and the team is working very quickly already with the With Intelligence team to look at the data and see what they can launch there by way of new innovative indices also.
A follow-up on private credit specifically. How are the competitive dynamics within private credit, and how do you think about your pricing power there as opposed to the public credit markets?
I'd say it's relatively advanced, and part of the reason is that, you know, the private credit and private credit area are deeply part of that 95% of data that is proprietary, hard to assemble, contributory nature, and so forth. You know, as we did diligence, for example, and With Intelligence, it was quite clear how much clients were willing to pay for it, how that merged with other data sets that we already had, they value and how the, I'll call it the value equation of data that comes only through contributory sources, right? Is incredibly valuable. Now we have it there. We also have it in iLEVEL and WSO, et cetera, because each one of those are pools of data.
It's really the, it's almost the network effect, right? The more pools of data of that sort that you have, the more clients feel like S&P is the place to operate, to do their workflow with, to secure data because it makes it so easy for them. Once you have that kind of tight client relationship and demand, then pricing and value comes naturally to us.
Maybe just on the, on the ratings piece, I think we've mentioned perhaps even you may have even asked me this question last year. We don't use different pricing methodology between public and private in ratings specifically. It's essentially the same pricing approach. For us, really, it's the mix shift in terms of investment-grade high yield that might lead to, you know, different monetization within ratings. We don't charge differently for public versus private.
All right. Perfect. Maybe time for a couple more questions here as we wrap up. Maybe a higher-level question to tie things together as we approach the end of half hour here. The IHS Markit acquisition closed about four years ago. Obviously, still some moving parts. Mobility spin is to be determined. Second quarter, I think is expected timing of that. Big picture, do you feel like S&P Global is a better business model today than it was four years ago?
I would say we don't Candidly, we don't spend a huge amount of time looking in the rearview mirror. I think we have been extraordinarily pleased with how the merger has gone, whether it's the revenue synergies, the cost synergies, and just the strength of the overall franchise. That's one of the reasons why we were so incredibly excited to speak with you all at Investor Day about our mission of advancing essential intelligence with this incredibly unique set of assets that we have.
You know, I would say that we're more probably concerned with what happens, you know, four years from now as opposed to, you know, four years ago and figuring out how to really optimize in the sense of like monetization, how to get the best monetization outcome from the products that we have by creating the greatest value that we can for our customers. We believe that we have the most unique. Continuing to create these horizontal capabilities, whether it's Enterprise Data Office or Chief Client Office today, moving on to making sure that we're creating horizontal capabilities and technology in other areas tomorrow. You know, it's exciting time for us. We're very much looking forward.
All right. Perfect. Then maybe as we wrap up, just what are some of the key messages that you wanna make sure that people walk away with today?
Yeah, look, I think that for many of our key investors, stakeholders, all of you, the question I think has been sort of how do you disprove the negative around AI. We don't look at AI as a negative. We look at it as an incredible opportunity for S&P Global. We're extremely excited about it. Look, the base case is AI more than pays for itself. Beyond that, we see additional opportunities for growth. We see additional opportunities for operating at a really high level in terms of productivity. You know, for our perspective, we see that we can change our total addressable market. There's a lot more that we can do in different client sectors and different regions because the technology makes it easier for us to do it.
In the same way that you can vibe coding, we can also do that. We also have the ability to really take enterprise-grade scaled views on how we use the technology internally and how we bring it to bear to create value for our customers. We see that our data is much more valuable with AI than it is without AI. The number of our clients who are interested in consuming more data from their existing licenses or getting more data licenses to new data that they don't have today is really increasing by the day, and we see that in some of the metrics, for example, that we've quoted today. We're incredibly excited.
We think that it's a wonderful opportunity, and we're gonna continue to engage at the highest level with our customers on this and really extract those opportunities for increased value creation for our customers and for all of you.
All right. Terrific. Good note to end on. Thank you very much, everybody, for joining us.