Morning, everyone. Thanks for joining us here today, day two of BofA's Global Tech Conference. Delighted to have you all over here. I'm Wamsi Mohan. I'm IT Hardware and Supply Chain Analyst here at Bank of America. My pleasure here to welcome IBM. Today, on stage with me, we have Ric Lewis, who is the VP for Infrastructure. Ric is responsible for overseeing IBM Z, Power, storage, and IBM software offerings associated with that, IBM Cloud, and also global infrastructure support. Rick, thank you so much for joining us today. You're welcome. Happy to be here. Ric, I know we've only got 30 minutes, and there's a lot to talk about, especially in your area of expertise, but maybe just start off to level set so the audience can get to know you a little bit. Just talk a little bit about your career, leading up to your current title. Sure. Actually, I worked at Hewlett Packard Enterprise, HP, then HPE for 32 years. Prior to coming to IBM, I was based out here in the Bay Area for the first six of those years, then I moved to Colorado. I would say my time at HPE, I kind of became a little bit of an entrepreneur in a big company. My groups led a lot of things like scalable x86, known as Superdome X, then something called Synergy, which is the industry's first composable infrastructure thing, then the early work on infrastructure as a service, which led to GreenLake. Kind of innovation side of the house. I actually took early retirement, intending to retire, ended up in a conversation with Arvind at IBM, where he wanted to do some entrepreneurialism and innovation inside of his infrastructure group, which had been flat to declining for a while, wondered if I could honestly play around a little bit and see if we could get it to a growth vector. I came into IBM to do that, and it's been four years, and we've done exactly that. We now have a growing, healthy infrastructure business that is an important part of the IBM success story. IBM's had a ton of momentum. I'm happy that my group's been a part of that. It's been a great four years, so way worth the time. I'm glad I unretired and had some benefits along the way, like the resurgence of AI in the industry, or maybe not even resurgence, maybe just the surge of AI in the industry and the resurgence of on-prem infrastructure, of hybrid cloud, of all of that. Really fun. Wonderful business. Yeah. I always say, you timed the transition time pretty well. Correct for 2020. Yeah look at the stock since then, right? Yeah. Well, maybe to talk and segue into sort of the growth of infrastructure. Historically, mainframe has been cyclical, and still is probably to some degree, but the cycles used to be plus minus on alternate years. In a launch year, go up, and the next year go down. Now you're talking about infrastructure, at least under Arvind and sort of this restructuring some of the not restructuring, but just recharacterizing the opportunity for infrastructure. It feels as though now we're talking about low single-digit growth. Yeah. What's driving that change, really? Sure. A lot of different things. I would say, in that journey to get to infrastructure growth, first is around Z and investing for innovation, making sure that we had our A lot of my role is really capital allocation and culture. Capital allocation, getting the right features into Z and making sure that it's healthy, which really meant getting the innovation focused exactly on what clients want and having clients involved in the design process, so that they get exactly what they need. There was a lot of business model we worked on, not so much in Z, but in the rest of the infrastructure, so really healthy product management. Heavy work on strategy on where to play and how to win and how you keep a competitive, differentiated advantage and how you innovate in the right spots. Pivoting a lot of investment to storage software, making sure that we had integrated value propositions in the storage space, making sure that our IBM Power system, IBM Power in those cycles that you were talking about would be down double-digit percentages kind of in year two and three of its cycle. In this latest cycle, through heavy segmentation and optimizing the stack for our IBM Power for purpose computing, we've made that business now where it actually grew in all three years of its cycle in IBM Power. That's really strong. TLS, getting it focused on renew, expand, attach, and be efficient, those key dimensions. All of those things is like IBM Z started to show strong strength in the last decade of its cycles. The rest of the infrastructure business, we've kind of got it now to where it's growing low single digits. It's a very healthy business now. We're really happy about that. Ric, when you step back, right, there was a period between 2010 to 2020 maybe, or close to that, where Yeah there was this worry that Customers were moving off the mainframe, off the IBM Z platform. What has changed from a customer reception standpoint? I hear Arvind talk about sort of investing in innovation on IBM Z. Yeah. Maybe just if you could give us some color on that. I think the industry narrative, and I was working in this in a different place, as I told you, was kind of you're either cloud or you're not cloud. You have to make a choice. Are you cloud or not cloud? I think that was just wrong from the get-go, similar to some other predictions in the industry. There was a time in the industry when people said, "Well, there's only going to be one processor. It's going to be x86, and it'll have all the graphics and all this stuff integrated into it. There'll only be one thing that's offered." That didn't pan out. There's more proliferation of processors today than there was then. GPUs are still their own thing and obviously extremely relevant to AI. I think the same thing happened with cloud. Early on, people predicted it would all go to cloud, there'd be no on-prem infrastructure, but they kind of lost the plot on data. It turns out that this AI era has only exacerbated that. Meaning, AI at its heart is about getting value from the data and using data to get value from the next data, and that data is everywhere, and it's always going to be everywhere. It's going to be in devices, it's going to be in mission-critical hardware. If it's representing 70% of the transaction volume by value, which mainframes do, that's going to stay in somebody's data center. Being able to provide that value and make sure that it exists well in a hybrid cloud environment, meaning there's going to be data in public clouds, data on-prem, data in colos, making sure that your set of tools, your hardware itself, the tools you put on that hardware exist well in that hybrid cloud environment means you're going to be relevant in that spot. I think Arvind called it well, which was one of the reasons I came to IBM, is I knew he had his head in the right spot when he said the strategy of IBM is going to be really simple. It's hybrid cloud and AI. That's how the industry has panned out. We've been investing in making sure our stuff fits well in a hybrid cloud, and that our clients can get value from all of their data in that hybrid cloud, wherever it lives, through the AI capabilities. I think that totally explains the resurgence. There's a lot of mechanics underneath there about working on culture, working on your operational efficiency to be able to fund all the work that you're doing in AI and hybrid cloud and those kind of things. We've been doing that work together, and it's been a very strong success story. Maybe just on AI, what are you seeing in terms of customer use cases and adoption that is driving- Yeah some of the workloads on Z? Z16, our latest cycle, I would say most of the AI, In fact, we announced Z16 and shipped it with AI technology in it before there was a ChatGPT moment. You probably remember, it was a year and a half before ChatGPT and the whole world started talking about AI. In fact, doing some of these sessions three to four years ago when we were launching that round of Z, I would have people in these kind of sessions ask me, "Why did you put AI technology in there? Do you think there's a use for that?" The world has changed a lot in that cycle. The reason we put it in there was clients told us, look, fraud at that time was about a $200 billion problem in the industry, and they needed a way to deal with it. They needed a way to be able to do fraud detection in line in their CPU. They asked us if there was something we could do. We put in some kind of machine learning/AI capabilities in Z16, which was extremely popular. That thing, if you think about it, is a rules-based architecture. It kind of watches for specific types of transaction patterns and then flags if it sees something that looks like it's out of the ordinary. If you imagine now Z17, where we have Spire and we have generative AI added to the mix, what we can do is something called multimodal, meaning not just machine learning, rule-based kind of AI detection, but true gen AI that can take different sources. Instead of just what do the transactions look like, we can say, does the transaction source have a physical address? Do they have a Better Business Bureau report? Do they have a good reputation? What's the sentiment online about this company that's trying to do the transaction? That's a good example of an old use case, Z16, to a new use case in Z17. Still the same thing, it's a lot better fraud detection that's built into the platform. We have insurance companies doing scoring of potential risk associated with given clients. We have cancer research organizations that don't want to move their data off a mainframe because it's medical records and the most sensitive data that there is. They want more AI capability to do cancer research on their population inside of a specific mainframe. Retail, we have marketing campaigns. How did it work? How well did it work? Did it really cause people to transact, or did it just cause sentiment to rise? All of that, think of it as where the data is, should drive your AI strategy and where you want it to be. We're making sure that our platforms deal well with that data in all these distributed locations. That's a lot of kind of what we're seeing in terms of use cases out there in the market. How's IBM monetizing that? As we think about cycle-to-cycle economics. Yeah how should investors think about that? A whole variety of ways. One thing that's unique about IBM is we're a full stack innovator, for lack of a better way of saying. We're not just buying GPU chips and putting those in a cloud and then serving up tokens to different clients. We're trying to provide full workload capability. All the way from, we build the chips that we do our AI on, though we also buy chips from other GPU vendors for specific AI, like huge language models that might work for a specific purpose. We build boxes that optimize that. We have what we call AI in a box with our IBM Fusion solution. We build storage software to allow you to process and pre-process, it's called content-aware storage, the data that we actually feed into our servers or other servers. You kind of build an up-stack. We have software on top of that, data lakehouse software that we can do analytics on for AI. We have consulting services for clients if they don't know how to get started in AI, the consultants in our organization will say, "Here's how you do it. Here's hardware that you can use, either our hardware or other hardware. Here's software." All elements of that stack we can monetize in various different ways. At all elements, we're super careful to say, "You can use all our solution, or you can use pieces and parts of our solution. What we're focused on is what's your outcome that you're trying to achieve on your AI, and we'll help you get there. If you have no preference on what's underneath it, we'll pick the best for you. Often that's our stuff. Sometimes it's not our stuff, depends on your workload. Our goal is to make the best for you." The key thing for our AI monetization is it's about enterprise. Enterprise is the bullseye. There's a lot of work going on in AI out there that has to do with write my essay, tell me answers about how I get a mortgage on my house. We're not doing that. What we're trying to do, in fact, we believe that 99% of enterprise data has never made it into any model, has never done any training, has never gotten the value extracted from it. That specific enterprise problem, that's the Holy Grail for us at IBM, and we're trying to make sure we do that problem extremely well, rather than the generic problem of write better essays for your kid's college app or that kind of thing. Yeah. That was a long one. No, it's important really core to our strategy. We think about that in hardware and software and how the things work together, et cetera. It's about, we know those enterprise customers, especially with the ones that their entire product or their business is that data rather than it's some other thing. It's their lifeblood. We know them better than anybody else, and we want to help them get value from that data. We have a big IT budget, and when I talk to our IT guys, they talk about IBM being the largest chunk of that. Yeah. Good. I'm glad to hear that. Glad to hear that. Maybe just to talk about, and I've covered IBM long enough that I have in my model a breakout of MIPS. Back in 2004, where we had MIPS growth disclosed and price per MIP declines. Yeah Maybe can you just share some flavor around, I know you don't explicitly disclose that, but directionally, how has that translated. Yeah As you added more and more value. Yes into the mainframe? Awesome. MIPS growth. First I'll start with revenue, then I'll kind of get back to MIPS. Revenue-wise, our program-to-program growth has been steady for a decade. In fact, it's accelerated a little bit. I know that Wamsi knows, we're over 120% program to program for z16 versus z15. z15 versus 14 was somewhere in the teens, 115%, something like that. I don't remember the exact number, it was kind of in that range. All that's a good indication of the revenue growth, we've also had MIPS growth in that same time frame. The fascinating thing is most people would think that MIPS growth must be just your core TPS, the old school transactions running on Db2, CICS, IMS, those things. It must be that that's at the core of it. Actually three times that MIPS growth. Those MIPS have been growing due to transaction volume and that kind of capability. We've added new, what we call specialty MIPS, which is the LinuxONE, kind of Linux running on a mainframe where we do either server consolidation or we do container-based workloads, or it runs Java extremely well. We have clients that have taken apps off of x86 and put them in their mainframes because it's so efficient at doing that. It's also really efficient on energy usage. Those MIPS have grown faster than the traditional TPS MIPS that everyone thinks of. In fact, at about 3X the rate over the last decade. That's been a big thing. Maybe the thing I'm the most excited about right now is we think there's yet another new category of MIPS that I think, I'm predicting, will outgrow the specialty MIPS, that's, I'm calling it AI MIPS. Basically has to do with selling Spyre Accelerator, selling software on top of the AI capability that we have inside of those mainframes. If you think of our competitive advantage, if you will, IBM's, is we have in our systems, they're not our systems, they're in the client systems, the most important data in the industry to 45 of the top 50 banks, to nine out of 10 of the top retailers, to four out of the five of the top airlines. That data is absolutely critical. They want value from it. There's a new revenue stream of selling them AI services and software on top of those AI processors inside of those machines. That's yet another category of MIPS. I see our MIPS growth kind of up and to the right for a variety of reasons, and I hope to accelerate it. When you think about that translating into revenues and sort of like the price per MIPS decline, would you say that in aggregate, your MIPS growth is outpacing in a cycle-to-cycle? Yes. Yes. Yeah, it's outpacing for sure. Yeah. If it wasn't, all of those gains would be pricing gains, that is definitely not the case. We have modest price increases based on inflation, value added, and new capabilities that we pass on to our clients. Not only are they involved in the design, they're also involved in our pricing in some sense, by kind of saying, "Hey, here's kind of our sweet spot of what we can do for our capability, price-performance-wise. Yeah. Rick, I think you mentioned Spire cards a couple of times, it might be helpful maybe for those who are- Oh, thank you not quite familiar to just talk about what Spyre is. with IBM z16, the prior version of IBM Z, we introduced that AI capability that I told you about in a processor called Telum, or Telum, sorry. Telum. That was the processor in our IBM z16. When we go to IBM z17, we start shipping in a couple of weeks here, we go to Telum II, which is the next generation of that with AI built in. In addition to the AI built-in to the Telum II IBM Z processor, the core CPU of an IBM Z system, we've added PCIe cards, Spyre cards, which are essentially our own custom-designed GPU-like, except they weren't built to be a GPU. They were specifically built to do enterprise data inferencing capability. We're adding those cards into a mainframe so that you can add increased AI capacity without having to buy new Telum II CPUs. It's a new value capture method for us. More importantly, it's a new capability for our clients to be able to expand the workloads that they can do inside the same secure, trustworthy, high transaction throughput processor that they've built their entire business around in a hybrid cloud environment. Also, I think it's interesting just to, you obviously noted in a couple of weeks, you've got the GA and launch of. Yeah z17. The Spire cards, though, are coming more towards the end of the year. They are. Yes. In some ways, the volatility of sort of maybe some of the mainframe revenues is more muted given the timing of this. I love what you said, Wamsi, because that's my entire intent with that is, we've talked about transitioning. When I came into the infrastructure group, we had two goals on the business. One was get it to growth, the second was dampen the cycle. Dampen the cycle, you could do that by killing the cycle. You don't want to do that. Right. You want to dampen the cycle where it's still growing, but that you have incremental value opportunity for the clients for them to have other ways to consume. Spire on a different schedule. How about Spire that spins before the next Z cycle, so that you have one round now, you turn it again. It's matrix math. We can turn that a lot quicker than we can a from-scratch Z processor kind of thing. Right. If we introduce another Spire, another value capture opportunity mid-cycle, that helps us with flattening out the cycle. That helps with the company and kind of its compares so that it's not, well, the rest of the company's doing this, but infrastructure is in and out. I kind of want to work hard to get away from that. Part of that strategy around that has to do with Z and the thing we just talked about. Part of it, frankly, has to do with shifting more value to our software and the rest of infrastructure, meaning Power storage, and those capabilities, and we're doing that as well. Power was actually, it grew in all three years of the cycle this last time, which is crazy different than the past. It would be down kind of double-digit in the second and third years in prior cycles, but it was flat to growing in all three cycles this time. There's a whole bunch kind of involved in that infrastructure thing, which took a lot of work for us culturally, process optimization, all those things that. Yeah we can talk about. Yeah. No, that's, I think, a really smart way to- Yeah to manage the business. I think one thing that maybe is somewhat lost is just sort of how relevant the mainframe is still today. Yeah. Beyond that, just sort of to the economics of IBM, when you think about all the transaction processing that's going on the platform. Transaction processing in itself has actually gone from a declining business to- Yeah a growth business. Yes. Can you talk a little bit about some of the puts and takes? How much of that is MIPS growth? How much of that is pricing? Yeah How does that shift like from sort of through a cycle when you introduce a new Z platform? Yeah, I love your question, and we didn't rehearse this, but I will tell you, Wamsi wrote a very good report on this topic. Oh, thank you that I read. I thought, I can't say, "Oh, he got the numbers exactly right," because I can't talk about the very specific numbers in it. Directionally, the knobs are correct, meaning there is a healthy component that just has to do with straight MIPS related, meaning transaction volume drives MIPS. MIPS growth, MIPS capability in a Z platform drives software consumption, as you'd imagine. You have it there, and people consume that. It's not just that. It's also incremental workloads, incremental capabilities. I talked about the AI thing driving it, specialty workloads. Our IBM LinuxONE MIPS have grown faster than the other that I mentioned earlier. All those IBM LinuxONE MIPS, those are around container applications. We have people doing native container development, hybrid cloud applications where containers exist in a hybrid cloud environment, Java application server consolidation. I would also say there's a little bit of it that's macro-related, and that is 10 to 20 years ago, you heard people choosing between cloud and not cloud. Now, people have chosen the hybrid cloud environment. If there's no reason to get off of a mainframe, why don't I just maximize the amount of transactions that I'm processing on there anyway? It's economical, it's energy efficient, it's the most safe and secure way to do it. I think just that mental switch for our clients has kind of opened up a willingness to say, "Where's the best place to run something?" A Z mainframe is not the best place to run a mobile app that's globally distributed. We wouldn't recommend you do that. Do that in a public cloud. We have that stuff all over the world. Do that. It is the best place to process your most important transactions and to get value off that data. Do that there. Make sure the two work well together. That's kind of what's driven this TPS thing, is a good fit, a macro environment that's favorable to it, and all of the knobs that I'll refer people to your paper. Well, thank you. I really appreciate that. Yeah. You did a nice job. Well, thank you, Ric. I appreciate that. Actually, one other point on pricing is that it's not just a like-for-like pricing, but the portfolio itself has created to add more value, like watsonx running on Z, for instance. Yeah We hear from clients and talking to them that they're adopting in droves. Yeah because it seems to be very useful, too. Can you talk about what kind of things you're putting into that allow for that excitement? Yeah. I'll even talk a little bit about the pricing thing. We don't price those TPS MIPS the same as we price the LinuxONE MIPS, for example, because it hits a different segment of the market. Similar, our AI MIPS. It kind of depends what you're doing, the value you're getting, and the difficulty for us to provide that value to you. I mentioned the dimensions we worked on in culture in the group around making sure that we have the business model right and the where to play. There's a value capture thing there that clients want you to capture the right value for the given workload because they know that means you'll invest to make that better. watsonx on Z is a great example of that. Clients were demanding easier management and easier skills portability for their Z mainframe environment. In z17, we've built in the capability to run watsonx Assistant for Z that helps you optimize your parameters in a Z system and run it in a more efficient way. We've also built in native support for watsonx Code Assistant for Z, which allows you to translate or comment, document code that you have running inside of your mainframe platform. Those are two software applications built around watsonx that run in your Z that are optimized and tuned for the platform, harmoniously, they make the platform easier for you to use as a specific client. We charge for those things appropriately for that, and the clients are happy to pay it because they know, okay, that means you'll make this part even better in the next generation. It really is like getting to a culture in the group that really focuses IBM's always had a great culture of focusing on the client, sometimes there's got to be the right business approach to say, "How do I do that piece better than the competition can do it? How do I give you the most value and have you pay for it that I can generate the flywheel and give you something you're even more thrilled with and more thrilled with?" I think we've got that magic going really well in the infrastructure group right now. Yeah, I hate to say this. I think you've given me more food for thought on how I got to break up all this into different categories now. Good. I was hoping that. Maybe quickly, I know we've got very little time left over here. I'd love to get your thoughts around Quantum really quick, because last year, I think, when Google announced something around Quantum, their market cap went up by $100 billion, which is half of your market cap. You guys have been doing stuff in Quantum before them. Would love to get your perspective on that. I have a thought. There's a recent article in Fortune that's very good with Arvind kind of talking about this. We've learned our lesson about being too story forward without the content in the early Watson round of AI. Yes. A lot of talk and words, not a lot of meat quite behind that. Sure. We're taking a different approach with Quantum, quite frankly. We've got a lot of meat on Quantum. We have 75 systems that we have everybody from industry people to universities to, and we've got a very strong value proposition around our Qiskit code that runs on top of it. Think of it as kind of the CUDA equivalent to GPUs. We've got that for our quantum architecture. I think we're far ahead in the industry in terms of content, not in terms of hype, and we want it that way this time. I think we're a couple years from quantum utility, which means generally solving problems of interest, and I think we're, by the end of the decade, at a quantum advantage where it's very obvious that there are problems that can only be done by these computers, and I'm really thrilled with the progress associated with it. Anyway. I think we're just about out of time, but maybe I'll give you 30 seconds to just talk about what are you most excited about in the next three to five years? Oh, boy. You can't be in this wave of the industry and not be excited about AI, but I'm sure people are also a little worn out from the term. I'm going to pivot that term a little bit and say data. What I mean by that is that enterprise data, nobody's getting the value out of it yet. It's a huge opportunity for the clients. That means it's a huge opportunity for us, and whether it's processing it, backing it up, managing it, figuring what's in it. Like, we have new content-aware storage offerings. You'd say, "Oh, the industry's been about data." Man, the industry's been about data almost like just deal with that. Make sure I can get at it, but just deal with it. Now it's about, oh, my gosh, all of that is worth gold. How do I get the gold mined out of that data? I think that's what I'm most excited about. I think IBM's really positioned to do that better than anybody else. I really believe that, I'm excited about that. Amazing. Well, Ric, that's all the time we have. Thank you so much. Awesome. Thank you. Really appreciate you being here.