Great. Good morning, everyone. Thank you for joining the UBS Global Tech Conference. I am David Vogt. I'm the hardware and networking analyst here, we're excited to have with us IBM today. From the company, we have Dinesh Nirmal, SVP of products and IBM Software, the audience is the IBM investor relations team, so you can hit them with all the deferral questions later that we don't cover. Dinesh, welcome. Thank you, David. Great to be here. We thought it would be great to have Dinesh here, given the theme for most of the meetings this week has been around data center AI software. Given you're uniquely, I think, qualified to speak to it, we wanted to start with your perspective of IBM Software portfolio, how we got to where we are today, either through organic growth, inorganic growth. We could, I guess, maybe go back five, six, seven years, start with Red Hat, maybe let's level set where we are today from a portfolio perspective that falls under your purview. If you go about five or so years back, we said our strategy is hybrid cloud and AI. If you look at five years back, right? It was all about going to a public cloud. Every customer was racing to a public cloud. Now you fast-forward five years, every customer is looking at to say, "How do I stay on multiple clouds? How do I keep some of my data behind the firewall?" The hybrid cloud strategy that we put forward about five years is coming real, partly because of regulations, right? If you look at what EMEA is doing with DORA, it's all about resiliency and risk. Resiliency is about, is your data on a single hyperscaler, or is your data can be recovered in multiple places, right? That, along with cost, no customer want to be boxed in. You don't want to go stay in a single cloud. All those things are really driving the hybrid cloud strategy. You talked about five, six years ago, we acquired Red Hat. If you are on multiple clouds or hybrid cloud, you want to develop an application once and deploy it anywhere. How do I do deployment I mean, development once and deployment in any place I want? OpenShift can enable it. The growth that you're seeing in Red Hat, for example, is partly because customers are really taking their applications and deploying in multiple places. That's one. Two, we have really brought every service that we have, whether it's Apptio, whether it's Concert, whether it's Databand, every service that we have on every single hyperscalers, including IBM Cloud and behind the firewall. I think we're the only vendor who really have 6 or 7 form factors that really delivers, right? That's the second on the hybrid cloud. The hybrid cloud strategy for us has really panned well, and even if you look at the Hashi acquisitions we are doing, right? It enables you to do infrastructure as code. That's all Terraform is about. It enables you to do secrets management using Vault. All those things that we have done, all the acquisitions we have done, the organic development we have done, has always been focused on hybrid cloud. How do we bring on the multiple clouds? The second piece of our strategy has been around AI. Even before generative AI took off, right? There was the predictive AI, deep learning, all those things. In the generative AI side, there's 3 core pillars. One is data, mainly unstructured data. How can we really take advantage of unstructured data? We are playing with watsonx to say, can you really bring your unstructured data, govern that data, really make that available for fine-tuning or prompt engineering or prompt tuning? That's one. The two is models. We have our set of models, smaller models called Granite models. Initially, it was all about the bigger the model, the better, whether it's 30 billion parameter model, 100 billion parameter model. We think smaller the model that's domain-specific is where the future will go. Because cost also becomes an angle for customers to say, "Do I need 100 billion parameter model?" Right? That's the second one. The third one is applications, which is, how do I generate more code, or how do I build more applications using the same resources I have? That's what generative AI will enable. These applications have to do some inferencing, use the models, all those things. If you look at generative AI, those are the core things. The outcome is automation. We are playing heavily on the automation side, whether it's IT automation or line of business automation. Our strategy that we put forward five years ago focused on hybrid cloud and AI has really enabled us to be here and really be in the forefront, partly because we are consistent in the strategy. We haven't changed it. We have always said it's all about hybrid cloud. We will bring every service on every single hyperscalers, also behind the firewall, and a customer can choose where they want to deploy. On the AI side, we are playing with models, unstructured data, and applications. How can we enable our customers to really drive more application development? One more thing. Then I'll pause. We are not just delivering, we are becoming client zero. If you take my own example, right? I run about 15,000 developers worldwide. About 6%-7% of all new code generated today is through generative AI. We would like to get to a 25%-30%. It's really helping enable us to deliver more for our customers through generative AI. It's really enabling us to be more productive, us to be more optimized. It's not that we are just going and selling our product. We are saying we are a client zero. We are using it. Another example, watsonx Orchestrate that we are selling, which is all about digital labor, automating your line of business. We are using internally. Our HR team is using watsonx Orchestrate to really automate HR processes. That is also very powerful where you as a vendor can say, not only we are delivering a service or a product to you, but we are internally using to optimize ourselves. That's a great overview, and it opens up a lot of different vectors here. Let me just go back to the software, pure software side first. If you look at the transactions over the last five years, you've obviously, Red Hat was foundational to the business today. Recently, you've done Apptio, HashiCorp is pending. As you look at the portfolio today, maybe what's the common thread in addition to AI that all of these assets sort of pull together this multi-cloud strategy? When you look at, for example, going forward, obviously, I think Arvind's talked about doing incremental software deals that bring multi-cloud, more functionality to customers or clients together. Where do you see the business going from here from the software side? Presumably HashiCorp closes in Q1 of this year. What's the next sort of initiative, the next step within software? Yeah. The common thread is hybrid cloud, because any new offering that we are bringing forward or any acquisition we are doing, we always look back to say, how is that going to enable us to deploy it in multiple clouds? If I take HashiCorp as an example, right? HashiCorp has two core things. One is the Terraform, which is infrastructure as code, and there is the Vault, which is secrets management. If me, as a developer or as a DevOps persona, I can now really take and deploy my code on any hyperscalers or behind the firewall using Terraform. I can use secrets management or Vault and really use, as a developer, I can use the secrets management to go deploy my application on any cloud. I don't have to worry about secrets management of a specific hyperscaler or behind the firewall. That's all taken care of by the Vault. Ansible from Red Hat, it's all about automation. How can I really automate my software? We have really stayed core, David, to our fundamental strategic principle, which is, can we really accelerate hybrid cloud, multi-cloud journey for our customers through the acquisitions and organic investment we are doing? Another example, organically, we announced Concert, which is a product that we announced. What is the biggest challenge customers face as they bring generative AI? I think it's three Rs, I call it. One is responsible AI. You got to make sure that it's responsible. Two is risk in AI, because as these developers use open source framework, whether it's LangChain, or any of the open source frameworks, there's a lot of open source packages that comes with it. What is the vulnerability that's associated with it? There's a risk in AI. The last piece is resilience. How do I make sure my infrastructure, my systems are resilient to do it? All these things that we are developing organically and buying inorganically has been focused on one thing, which is hybrid cloud. Can we enable our customers to develop once, deploy anywhere, and then can we really accelerate their journey on the generative AI side of things? You mentioned resiliency. Can you expand upon that? Today, as the portfolio is currently configured, what does IBM bring to bear to a corporate customer to help them along their resiliency journey, given all these complexities and risks and challenges that you just sort of articulated? Yeah. If you look at resiliency, right? There's maybe four or five core elements to it. One is recoverability. If your application goes down, how soon can you recover it? Or your data gets corrupted, how soon can you recover it, right? There's observability. How well is your estate observed? There is maintenance, right? Which is a very critical piece. I mean, look at CrowdStrike. If they would have deployed that application on a canary method, it wouldn't have happened because they would have caught in Australia or an early phase. Those are very critical for maintenance perspective, right? There is scalability. How can you scale? There's four or five core elements to resilience to make sure that how resilient is my system. Recoverability, right? People say, "How soon can I recover? What is my RTO, RPO? If my data get corrupted, I have a time limit, 30 minutes to recover my data." In its honesty, how many enterprises have really tested it to say, "I can recover my data in 30 minutes"? Those are things that we as IBM looking at to say, how can we really make sure that we can bring a product or a service to our customers that will give you a complete score, take all these things into account, and give you a complete score for an enterprise to say, "Your resiliency score is 85. Take these five actions, and that will get you to a 95 score." Now me as a CIO or a CRO, now I know I'm resilient in these areas. I'm not resilient in these areas. For auditing purposes, right? It's so critical that me as a customer can stand up and say, "My resiliency is 85. These are the actions that I need to take." How do I keep the actions that I take in an evidence locker for future audit? There's a lot of things from a resiliency perspective, and we as a vendor, I mean, having worked with enterprises for the last 10 decades, we know the enterprises so well. We can authoritatively come and help them to really make them more resilient. Can you take all of those offerings and that view that IBM has and talk about sort of specifically within AI from the software portfolio to date? Really, it seems like the consulting business has been leading the initiative, right? If I just look at the metrics that the company's reported, 80% of your AI-focused orders to date have been really on the consulting side. Maybe can you expand upon how that helps lead the software business, and what are the initial feedback or learnings that you're getting from large corporate enterprises on AI, particularly around your software portfolio, and how do you think that sort of evolves over the next couple of years with consulting obviously taking the lead? Yeah. Obviously, anytime a new project comes in, a customer wants to look at what are the areas. There's some study that needs to be done to say, "Okay, what are the areas that I really need to bring generative AI into it?" If I simplify it, I think there's two core areas where I see generative AI disrupting. One is application development. Look, every enterprise or a business is going to live and die with two things, applications and data. If your applications go down, it makes headlines, as you know. If your data gets corrupted or is not available, it will make headlines. Those are the two core things. If I'm an enterprise, I really want my applications to be stable, but I also want to create more code or more applications to drive more business. Generative AI is really enabling customers to create more code. Code is one area that is really getting disrupted by generative AI. Like I said, I myself is using it in my development organization, every customer will use it because that enables them to write the 10,000 lines of code they couldn't write last year or create the five applications they couldn't do last year. The second area that's getting disrupted is the line of business, because that's very process-heavy. Whether it's the credit approval process or whether it's the loan approval process. The line of business side of things is getting disrupted through a couple of ways. One, a customer agent, you get a call. Before they have to read through a document, understand all those things. Now, with the click of a button, you can summarize the 10-page document in no time. You could get a sentiment of a particular call based on the dialogue that happened right after the call. End of the day, that is not just a process transformation that's happening, but it's also every business' end product is a happy customer. That's also happening through generative AI. Those are the two areas that we see where generative AI mainly is disrupting. On the code side, we came up with a product called watsonx Code Assistant, which is, one aspect of it is taking our strength, like COBOL to Java, right? Nobody else can do better than us. Right. We figured we will train the model and bring it. The second is generic purpose coding, which is, you are a developer, can you have a VS Code plugin that you can enable, call the Granite models that we have, all those things. That is the disruption that's happening there. On the line of business side, there is four or five core things that you need to really do. It's not just generative AI. You need these tools. One is workflow. You need a very strong workflow for to-dos. Two is RPA. Some level of robotic process automation you need, because some of them are not just a process, it's tasks. Three is process mining. You need to mine a process to understand where are the inefficiencies existing in a process. Four is a rules engine, because not everything can be AI. There has to be a human in the loop in many cases. Last but not least, is document processing. Meaning, how do I extract the fields in a document and put it somewhere to be? Those are the five core things. Right. Unless you have those tools, you really cannot bring AI to really automate it. To me, I might be generalizing it, but there's only two vendors who can really do it. One is Power Automate with Microsoft, and two is IBM with the tools that we have. We have all five of them, so that they'll do this. They only work on Azure, we work on any cloud, so it gives us the benefit. Generative AI can come in there and really enable it. Look at the process, right. That's an area that could easily get disrupted. Add a rules engine into it. You could completely automate that using an agentic framework. You want a human in the loop, you add a rules engine to it. Those are the two areas we are really focused on. On the bottom layer, David, is the data aspect of it, and it's mostly unstructured data. We announced watsonx.data for that to really say, "Look, you can bring tremendous amount of data. We will manage it, we will catalog it, we will copy the metadata, we will govern the data, we will securely place the data in a persistent place." Those are the three core software areas that we are betting. I just want to make sure I understand. The way I thought about Watson, what you've announced over the last couple of years, it's more of a platform of technologies and tools and applications that can help enterprises move down their journey, but also help automate and improve your existing functional software technologies. Is that the right way to think about how Watson will get developed over the next couple of years? Watson, all this capability you're building helps the security side of your business. It helps the data, the automation side. Is that the right way to frame it? It's not necessarily a standalone offering. It's integrated in everything under your purview that you're thinking about developing over the next couple of years. Yeah. The watsonx brand that we created is specifically for generative AI. We said there's three core pieces to it. One is watsonx.data. That really looks at the unstructured set of data. Majority of the growth in data is happening on the unstructured side. There is the watsonx.ai, which is how can I develop an application used for inferencing purposes? The three is watsonx.governance, which is once you deploy the model, a lot of times the model hallucinates. That's the reality. Governor. It could drift, right? How do I govern the set of models, right? With the right set of guardrails. Those are the 3 core elements of the platform that is called watsonx. Now, on top of it, we have added watsonx Code Assistant, which is to help developers code. We announced watsonx Orchestrate, which is about the digital labor, the 5 line of business thing. We are bringing new things depending on where we see the market and where things are going. In its entirety, the core is 3 things, which is the data, the governance, and the model/the inferencing engine. Does watsonx, you mentioned COBOL to Java, that would help assist on the mainframe transaction processing side as well, I would imagine. Right. It permeates other sort of sectors within IBM. Right. Exactly. watsonx Code Assistant is a product. We took our Granite model, trained the COBOL code. Right. That can convert that into Java. When you look at code development, right, people think it's just writing code. It's not. There's test case development. There's refactoring of code that needs to be done. There's documentation of code that needs to be done. A customer could buy watsonx Code Assistant and say, "Well, I don't want to do any code development, but I want to do test case generation. Okay. A code could have been written in the 1980s. Those people are not there. I want to create documentation for that code, right? I myself was a developer. The last thing I want to do is write documentation. Everybody wants to write code. Whether it works or not, I want to write code. If generative AI can help me write documentation, all of a sudden, I have 15% more productivity because I can use that time to write more code. WCA has really focused on the developer, multiple aspects of development. The model itself, we have multiple models. One of the models is about COBOL to Java conversion. Got it. I wanted to maybe take a step back and talk about the common thread for a second. Red Hat, you talked about all these other companies that you've acquired, Apptio, HashiCorp's part of it. We get a lot of questions on Red Hat in terms of hybrid cloud, multi-cloud versus public cloud. Red Hat has had fairly recently good acceleration and growth. Can you maybe talk to what the markets look like that are driving that underlying demand for Red Hat? Primarily, I think you talked about OpenShift bookings being incredibly strong, teams fairly recently, Ansible's strong. Is that standalone pre-gen AI demand or is there some sort of element to customers looking for multi-cloud, hybrid cloud solutions because they're going to embark on a gen AI journey effectively? Yeah. I think you hit the nail on the head. Like I said, every customer, especially large enterprise, is going to look at running applications on a hybrid cloud. I think that's the reality, because no one will want to get boxed in into a single cloud. If that premise holds true, then a customer has to develop an application in a way that is developed once and deployed in multiple places. That's one reason why Red Hat is getting the momentum, because there is no other container group that can do the kind of things it can do like Red Hat. The second one is Ansible, because you want to do automation, right? As you do deployment of this application and software, you want to make sure that these applications are seamlessly automated. The third area of growth that we are seeing is the whole VMware set of things. When customers get a bill that they used to pay $10, I'm just making that up, and it becomes $50, their immediate thought process is that I want to get out, right? If I have to pay five times more for something I used to pay one day before, they are thinking about alternatives. KubeVirt is one of the tools that Red Hat has that enables you to embark on that journey. That's the third aspect of it, of Red Hat growth. The fourth one is, as Gen AI is driving, it's creating more applications. That means you have to run it somewhere, and that obviously becomes, whether you take ROSA, ARO, any of those flavors that exist on hybrid cloud from Red Hat, that enables even more growth. Those are three or four things I would say. Did you see that VMware impact in the business over the last couple of quarters? Or is that more of a recent phenomenon given sort of the pricing dynamic? It's coming, right? It's coming. I was in Europe about a month ago, a public services company said to me, "Dinesh, I used to pay a couple of million. It has gone up to $20 million, $25 million." It's starting to kick in customers' head that it's time to look at alternatives. I think that, and I said that in my roundtable this morning. Once you get into an enterprise, any vendor, any application, it's hard to get in, but it's hard to get replaced. That's why software that has been around for 30, 40 years is still running because it's hard. It's not just replacing it, there is multitude of tentacles that is calling that application that's being attached to it. There's data, there's security, all those things. It's not just going to be just go rip and replace. It's going to be a progressive process whereby which pieces, the low-hanging fruits will get moved. Understood Then the medium, then the big. To me, David, I think it's a journey, but we are starting to see the uptick of the journey as part of that. Got it. In the interest of time, we get a lot of questions on your transaction processing business. I think that business over the last 5-10 years has been somewhat misunderstood, right? In the sense that it's been tied inextricably to the mainframe business for all intents and purposes, and had been sort of a source of headwind from a growth perspective. It appears to have changed over the last couple of years, obviously during COVID, now post-COVID. Can you walk through and articulate what the growth trajectory of that business looks like? What are some of the underpinnings to support that growth? Why that business now going forward is actually on much steadier footing than it was two, three, four, five years ago. If you look at mainframe, David, it has one purpose. It's the best transactional engine that has been ever built, will ever be built. Partly because it's machine-level code, meaning you can execute transactions in milliseconds, right? It's purpose-built for one thing, which is the transactions. As you create new applications using generative AI and all those things, what is it doing? It's driving more transactions on the back end, which in most cases is mainframe, right? You're getting some tailwind from that aspect. The second is some of the new offerings that we are bringing on mainframe itself, like watsonx Code Assistant or COBOL to Java. Now customers are saying, "Okay," like I said, it's not just about taking a COBOL application and converting it to Java. This COBOL application was built in the '70s or '80s. I want to create documentation for it. Can I use generative AI to do that? Or I want to refactor parts of it. Can I use generative AI? We are seeing that momentum of bringing new offerings on mainframe. Another one that we brought forward is watsonx Assistant for Z, which is, what is the biggest challenge customers have on mainframe is somewhat skills. Because a lot of the colleges or schools might not be training the new youngsters coming out of college on mainframe, right? We said, "Okay, let's give us a very UI-based simple tool where you can ask simple questions like, tell me my certificate expiry on this particular LPAR," right? It posts because that model has been trained through the documentation, everything that's related to mainframe. It's two or three things. One, generative AI driving more transactions on mainframe. Two, the new offerings that we are bringing on mainframe, three, new workloads that's coming on mainframe. Is the business at all tied to the somewhat cyclicality around the launches of new iterations of the mainframe? Is it just strictly transactional intensity effectively has improved dramatically? We do have a new mainframe cycle later next year. At some point, I would imagine, I don't think it's been announced, but it's been in your turn queue. Do we see an acceleration in transaction processing with a new generation of mainframe? Or is it really the installed base of transactional capacity is what really drives So- that business? I think every time we announce a new mainframe, obviously there is a tailwind that carries because customers want the newest and the best to run more transactions on it in a more optimized way. That's one. We always have seen uptick when we announce a new mainframe. If you look at the last two, three years, right? We have kind of broken the cycle whereby which, usually it goes up for the first three quarters, and then it kind of dips down. We have seen a steady, actually an increase in growth and transaction growth and the mix growth if you look at it. Partly because, I think COVID drove a lot of transactions. Generative AI now is driving a lot of transactions. We are seeing a steady flow of growth. I think what we are coming up in Z17 will enable us to have a bit more growth. The other aspect, David, that people don't understand is, when you look at an application calling a data on mainframe or an application on mainframe, there is a set of other elements that also benefit from it. For example, everything goes through an API call. The API traffic drives more. A lot of times it's message queue, so a lot of the MQ transactions go up. There is other peripheral benefits that we get, not just the mainframe, but also the distributed set of benefits. Perfect. Maybe just in the interest of time, I know you've been incredibly generous with your time. I want to bring it back to your perspective of where IBM sits today with the portfolio under your purview. What do you think is maybe misunderstood or from a market or investor perception that really hasn't resonated yet that you think IBM is not getting credit for? How you think about the portfolio as we move forward over the next couple of years that really gets you the most excited about IBM today versus IBM five years ago? I think the one thing I'm most excited about is the opportunity. The opportunity that generative AI brings, for IBM, we understand enterprises like nobody. Partly because we have been in those enterprises for decades. What I'm most excited about is that generative AI is at three or four core things, the outcome is automation, who can help them automate like IBM? I don't think there's no other vendor because you have to understand the intricacies of an enterprise to go help them automate. The most I'm interested and excited about the opportunity is the opportunity that exists and how IBM can go enable customers to automate and take advantage of generative AI. Got it. All right. I think we're out of time. Dinesh, thank you for your time. Thank you everyone for joining. We'll speak soon. Again, Olympics is in the audience if you want to hit her with the hard-hitting questions. Thank you again, everyone.