Welcome to day one of BofA's 2024 Tech Conference. I'm Wamsi Mohan. I cover enterprise hardware and supply chain here for Bank of America. We're delighted to welcome IBM here today to participate at this event. From IBM, we have Red Hat's President and CEO, Matt Hicks. Prior to this role, he was EVP of Products and Technologies, where he was responsible for all product engineering, including products such as Red Hat OpenShift, Red Hat Enterprise Linux, and JBoss Middleware portfolio, OpenStack and Ansible. Pretty much the whole shebang at Red Hat. He's been with Red Hat for 16 years, so brings a wealth of experience. He was also one of the founding members of the OpenShift team and has been at the forefront of cloud computing ever since. Matt, welcome. Thank you. Thanks for having me. Delighted to have you here. From IR, we also have Dina Corbi. In case you have any follow-up questions, please do reach out to her and she'll be able to adequately point you in the right direction. Look, there's been a lot of news at IBM, and it's been an exciting time. I was wondering, to start off maybe a little bit, if you could share about your own career and your journey leading up to this role at Red Hat. Absolutely. So I'm a product. I sort of started my career in the dot-com bust side. A little bit on the boom side, more on the bust side, in consulting, actually in IBM Consulting at the time. But I've always liked that pull of open source technologies and the intuition of finding these durable things in really frothy time periods like the dot-com bust. I went through consulting and software group. I was the Linux guy at IBM, and that drew me to Red Hat, and that held in Red Hat. I started in IT. I moved into engineering with OpenShift, and then I just slowly progressed through that intuition, that building of product sets, until I was running business units, all the business units, my current role now. It's been a fun journey, fun time. I love the space. Excellent. Well, we're glad to have you with all the experience that you bring. I imagine investors are familiar, kind of at a high level, with Red Hat Enterprise Linux. Can you spend a minute maybe talking about some of the other key offerings that round out Red Hat? Absolutely. If you think of Red Hat Enterprise Linux as the core Linux offering that we have. Two other products that I'll talk to, I mean, we do have our middleware technology that tends to be known, JBoss. The two others to talk to are OpenShift, which you can think of RHEL runs something on one server, but then the applications people build these days, they need a lot more than one server. OpenShift was focused on how do you bring together clusters and really large clusters of machines that operate like one thing. That was the OpenShift platform on it. Ansible, it's an automation platform, but it's the bridge between the reality that you live in today and maybe where you want to go with it. As exciting as OpenShift may be, you still have switches to automate. You still have older machines that are there from 10 years ago. Ansible helps you automate all of those things so you can just get the human cost lower, so you can invest in these new platforms. A couple of other things Red Hat does have been really foundational to our growth. It's kind of crazy that Red Hat has only been five years since the acquisition of Red Hat. Yeah. Would love to hear what surprised you about the Red Hat acquisition or what has been some of the outcomes at IBM that you've seen that are surprising? My day one surprise, to be honest, because I had worked at IBM and IBM has a really strong acquisition process and machine, and independence is not one of those playbook terms. On day one of the acquisition, I loved the storyline. I didn't know if IBM was going to actually be able to let Red Hat hold this independence position and keep building the ecosystem that was really core to our success. Five years in, I think two years in, Red Hat really started to believe of like, "No, this is additive to us. IBM continues to let us operate in the way that we do," which was a new playbook for IBM. That was surprise one, on the positive side. Surprise that I wasn't expecting. You can think of IBM and Red Hat as very different on one lens. On another lens, we both know operating systems. We both know middleware. IBM leans more towards the proprietary ecosystem and build. Those similarities actually made some of the cultural aspects of aligning a little bit more difficult in the short term. I think in the long term, we'll talk about some of the things, it's really been additive, but that was a surprise to me, where common knowledge areas could be trickier to align on than the new areas where we didn't have that background in technology. Okay, great. No, that's really good perspective. There's a lot to talk about AI because you guys have made some recent announcements, but there's also another element that I think a lot of investors are struggling a little bit how to think about this is the macro environment around enterprise software. Would love to just get your perspective on what is happening more broadly across enterprise software, where multiples have been de-rating, and it's just been a tough environment in general. Yeah. I'll separate the last couple of weeks on this because my advice is like, how do I separate the short-term froth from the trend on it? I would say, I think the customer pattern trend changed a year or a year plus ago, where really you saw that spike in inflation followed by interest rates and just the fundamentals of getting access to capital was more expensive. That started to change the, well, I have to be more conscious about where I put a dollar of investment, which trickled down into buying patterns, how companies thought about where they were going to invest. I do think there's been an amplification of that to whenever you take a constrained environment, then you add technology disruption to it, you're going to see different outcomes. We've had both the AI technology disruption, where if you're investing a dollar, you better also invest it in a good technology path because it's got to count for you in a few years. We've added a virtualization disruption as well of just a acquisition of a really well-known platform and a change in pricing. Those have been additive, but those are sort of the macro trends I've seen of the one that we all know around capital and then these technology disruptions that it's like anything, it's not good or bad. It changes how you think about your investments and making sure you're placing them right. Okay. No, that's super interesting perspective. Thank you for that. You did make a lot of interesting announcements about Red Hat and AI together with the research team at IBM. Can you just share a little bit about, for those who might not be quite as familiar about what these announcements are, what specifically you invested? Yeah. We talked about the short-term trickiness of having common area, and then how do you get real synergy out of that? Year five has been my favorite year, because we actually, and this was built on years of understanding each other and the advantages we both bring to the table. At Summit, we announced we're going to introduce a new product offering called RHEL AI. You can think of that as a special purpose operating system that as a customer lets you be independent of a GPU choice. The models that you work with will run on Nvidia, they'll run on AMD, and they'll run on Intel. That's sort of a sweet spot of what Red Hat did, is what we did with RHEL in the early days on Linux. We combine this in with what IBM Research does really well, which was creating large language models that can be trained. This is in the smaller models because they're really cost-effective. They're very capable, but you can teach them your business. When we put those two things together, as a customer now, our hope is that you can make a technology choice today that gives you optionality. You have optionality on the hardware vendor. The models themselves are open source, so it's your IP, clearly. What you put into them remains yours. I think the most important part of this is it allows you to teach these models your business. The really big models on the market, they only know your business if your business is in the public domain on it. I think most IP is not in the public domain. That's that combination of RHEL AI. I loved the IBM took a path in an open source ecosystem built with this by open sourcing the Granite families of models and the Instruct Lab tooling. In Red Hat, this really goes back to how we built RHEL itself, was using that ecosystem network to work with OEMs to be able to sell this with hardware, to really have that reach to put it in front of customers. It's been a fun two events on announcements. Yeah. Absolutely. Can you talk a little bit about the pricing tiers and methodology as you think about RHEL AI, OpenShift AI, and watsonx? Yeah. If you look at those three, because they're all additive. If you start with RHEL AI on one server, OpenShift AI is now expanded to multiple servers. watsonx expands that to use cases that you can tackle. It's all progressive. If you look at the pricing of it, RHEL AI will be priced against GPUs, because that's where you're doing training or inference in AI models. OpenShift AI will be priced against servers, because its job is to let you use the server and all the GPUs within it as effectively as you can. watsonx will be priced against servers. They're all infrastructure software coupled on that to really let you use infrastructure capabilities to your advantage as a business. No, that's great. I just wanted to go back, Matt, to your point around using these maybe smaller models that are potentially using enterprise data to train. I presume it's lower cost and faster time to market are the primary advantages of being able to do that. Or maybe there are other factors that you'd like to bring up as well, and what kind of model sizes or parameters are we talking about in these not super large models? Yep, it's a great question. I split the world into two camps. Let's say, there's the omniscient camp. Think of trillion parameter plus. They're very impressive of what they can do, but it's just math. It takes that much to load them into memory and run them. We tend to operate in the, you could say overly simplified, but 10-20 billion parameter range. The reason for that is because those are things that you can run on one GPU card. You can run them on a laptop. The reason that's impactful is one, to your point, the cost of inference for asking them questions is just that much cheaper. We're talking orders of magnitude to ask them questions. If it's a good model, you want your whole company, all your customers asking it questions. Two. The cost of teaching it something also has that same orders of magnitude impact. You probably can't really teach durably those massive models your business, but you can teach a 10 billion parameter model your business, that skill set really well. The other dynamic that I'd say is key here is that when you get into what you're teaching it, you're probably teaching it your intellectual property, and you want really clear framing around that. That if you teach a model something, you don't want it going into the public domain. You want to have control over that, and you have to teach it where your data is. In most cases, I forget the percentage, but we could say 70-80% or so of enterprise data still is on-prem. The ability to run and train on-prem is key. Think of where you're going to ask these models questions going forward. It'll be in the car. It's going to be on the telephone pole. It's going to be in the factory. These are locations where you get a tremendous amount of value out of them, but they have to fit that form factor. It can't be a tether to a 2 trillion parameter model in the cloud that's answering your autonomous driving question. That combination, I think, is why we see so much innovation and draw in this smaller model and why it's such a good fit for enterprises. There are other markets in those where I think the big models could have a great future, but in enterprise use cases, I think smaller trainable models in that space. Yeah. No, that's super interesting. If we were to step back again, just look at the time frame of enterprise adoption in some ways, right? You guys have been talking about a book of business where you've been very actively engaged in many different programs. Just curious, what is, in your opinion, the right time frame to think about enterprise adoption around AI more broadly for the industry? What we've seen so far is very hyperscale driven, very tier 2 CSP driven, but not really a lot of spend at the enterprise itself. Yeah. I would start with, I think, understanding where we are. Yeah in the market. My guess, if we talk about innings on this, we're not through the first inning yet. As a result, you see work that's very heavily skewed to proof of concept. What I've seen in customers, even in the last six months, is the realm of what they're going to tackle with proof of concepts has become much more disciplined than a year ago. A year ago, it was sort of in the AI can do anything, so I can pursue anything. The last six months has been this realm of boring use cases that are very impactful to the bottom line of, can I do support incrementally better by using digital labor to augment and put cases in front of customers better? Can I manage IT knowledge bases better, human resources knowledge bases better, sales operations? If you think of how this changes engineering, I'll give you the simplest example. Whether these can write code better or not, they can most certainly document existing code as well as anyone. That actually makes code easier to maintain. I do think I have been seeing this trend to more real world scenarios that apply to almost every enterprise on the planet, where you're going to get a known return out of it. That's an important phase, I think, to get us from the first inning to where you actually see spend start to accelerate, because you need to have the result of the digital labor cost be lower than the physical labor cost. It doesn't help if you do a POC, but then the cost per question costs you more than the associate you might hire to do it. Right. Still a positive trend, but that's why we're so early. I don't think it's really translated to spend at scale in the industry because we're just where we are in the market. Yeah. It's solving some of the COBOL programmer shortage we have. Yeah. Well, maybe stepping back again. Can you just compare Red Hat software stack versus competitors and where you feel like there's the most differentiation and opportunity for you? Yeah, I think if you look at Red Hat or if we dive into RHEL AI, the most common comparison will be with NVIDIA AI Enterprise. Nvidia is a phenomenal partner on the hardware side for this because we run on Nvidia GPUs as well as Intel and AMD. Nvidia also will have a very capable software stack. I think the goal is achieving many of the same things for customers. The difference will be in the optionality with that. If you're on Nvidia, if you use their software stack, generally, you will stay on Nvidia hardware forever. In the case of Red Hat, it is an open source-based innovation model, and if you build your models to our stack, it's your IP that you own, and you can use it with any hardware in the future. It's sort of the known capabilities versus that optionality and open source innovation. That'll be the on-premise run where your data is. You will also see the hyperscalers, the public clouds provide very vertically integrated stacks on it. I think similar play. They'll be great at the infrastructure side, but that'll be catered to their hardware, their environment, their models on that. Optionality, making it safe for businesses to consume where their data is today is sort of our goal in the market. Okay. That's super helpful. Maybe you can talk for a minute about the competitive landscape. There's been a lot of industry consolidation, and it's created some shakeup at some of the clients. In particular, when you think about where You've had a dominant player in the virtualization market now having changes as pricing that you alluded to earlier. Yeah. What is the impact that your business is seeing, and what is the opportunity set that it unlocks for you? Yeah. In short, I think this is a great opportunity for us. The way I frame it is when you make an infrastructure software platform bet, whether you like it or not, it's probably a decision that's going to stick with you for 10-plus years. In the case of mainframes, it's 50-plus years on this. vSphere has been a great platform for virtualization. When you introduce a compelling event a la pricing or a big acquisition, you make customers reevaluate that. The question they'll ask themselves is this platform, does it serve my use case today and how much into the future? For us, we do virtualization quite well in OpenShift. Our virtualization is based on KVM. You tell people we run several of the major core network, telco network providers in this. We run virtualization at scale, and we surface this in OpenShift. OpenShift does virtualization. What's OpenShift known for? It's known for containers and that hybrid portability for building cloud-native apps. What will that become the foundation of? That will be the foundation of AI model training and inference. When you look at those two, I think the way I tell customers is if virtualization serves your needs for today and tomorrow, you have a known bet with vSphere, and then it's a pricing difference you have to choose. If you're forced to make a platform bet, and you are looking to future scenarios, I'm obviously biased in this, but I don't think virtualization will be the basis of AI training and inference. If your future investments have to pull you in that direction, a platform like OpenShift is very strong there. You go below OpenShift, we talked about RHEL. Everything we build is based on RHEL. RHEL supports the hardware you have in your data center today. It works in all the major cloud providers. It works against the GPU manufacturers today. Those two components is why I think we've seen that strength in hybrid in general, but also why virtualization changes. Any compelling event where you stack up like that creates a good opportunity. You also have a partnership with Nutanix that you mentioned KVM. There is AHV from Nutanix as well. Yeah. Can you talk a little bit about that partnership and sort of where you see the opportunities there? Yeah, absolutely. If you look at OpenShift Virtualization, we have a pretty clear path. It is run this on bare metal. OpenShift has to be the platform you want to train your people around. If you do, you can now do all these things. In some cases, that is too big of a bridge where virtualization still might be the major, and Nutanix, and Acropolis, and AHV can be the better platform for them. They are a great partner in those scenarios because RHEL and OpenShift still run very well on Nutanix. For us, it is finding those platform layers for customers. We have good opportunities partnering with Nutanix. That goes from hardware to their software layer. We have other customers that will make that choice of OpenShift is going to be the center of their platform bet for the next 10 years, and either of those work. Okay, great. You made a pretty big announcement and intention to acquire HashiCorp and called out specific synergies with Ansible. Can you just explain how these solutions complement or are complementary to each other? Yeah, absolutely. If you look at the Ansible comparison, take Hashi. Hashi has an offering called Terraform. We get a question a lot like, they work in automation, you work in automation. Are these things complementary or competitive? The way I describe it is, for my whole time span in IT, there have been two automation technologies. One is procedural. It does exactly what you tell it to do right then, and one is state-based of, I don't care how you get there, but here's how I want you to manage this. The early days for me, it was CFEngine and distributed shell scripts. 10 years ago, maybe it was Puppet and a technology like Func. Today it is Terraform as the state-based management, and it's Ansible as the procedural management. That technology trend has had 20 years of holding that you need both of these pieces. This is why we see them as complementary and why customers use both of those. That creates really good commercial opportunities for how do we make that easier for customers to solidify as a package or combination. We see similarity within Hashi beyond just Terraform, things like Vault as a secrets management for OpenShift, where the security of passwords in distributed systems, it doesn't take long reading The Wall Street Journal to realize how paramount that is to IT operations. Hashi also has solutions like Vault that we see a lot of affinity with in offerings like OpenShift. I see it as complementary. It's really strong technology and product combination for us. Okay, great. Red Hat, just when we think about the revenue growth trajectory, I think last year we exited the year at a slightly lower rate. Sort of re-accelerated from that a little bit now. How do you think about the revenue growth opportunity for Red Hat in the medium and the longer term? Yeah, I think we're in a really strong position, and I'll go through. There are times when Revenue won't sort of show you the full picture. So we started sharing three quarters ago the bookings aspect of this. James Kavanaugh would talk about this as 80% of our business comes from that bookings trajectory. If the revenue, then you have this 20% transactional component to it. Since we started sharing that, this was three quarters ago, we've had three consecutive quarters of mid-teens bookings growth. That really, I think, gets to the fundamentals of demand on the offerings that we have there. That eventually leads into revenue on that. Then we have these new opportunities both in virtualization and AI. That's why we see that Q1 revenue acceleration is based off some of the impact of the bookings results there. That continues as we go through the year, Shantanu. Okay, great. I know that in the past, Jim's called out some project-based spending headwinds as well. Yeah. Are those still persisting? The re-acceleration that you saw, is that more product-based re-acceleration, or how would you characterize the mix between those? Yeah. When we think of project versus product, project, overly simplified, but you can think of as things like professional services where someone can just not choose to spend there, or we talk about our transactional software. Think of this, an example of it would be spinning up RHEL in the public cloud, where you can also just spin it down if you don't need it. Those are some elements that go into the project-based spend. That, for us, really stabilized in Q1. Anything that is volatile with this, though, comes down to week-by-week execution of I think we have great opportunities here in the future. If you look at professional services, the ability to attach that in virtualization and AI offerings. When you look at the transaction businesses, I still think public cloud, as that example, serves a really critical use case in a lot of these areas. I would say we're in the stabilized phase of this, and then continuing just to focus on execution to make sure that we manage the volatility there. Okay, great. I know we don't have a lot of time left, I'll try to sneak in two quick ones. Yeah. Here. Maybe just on the go-to-market, can you talk about any changes that have been made since the acquisition of Red Hat and how you're looking at positioning that on a go-forward basis? Yeah. We talked a lot since the acquisition. We need to strike the balance where Red Hat can invest in our go-to-market model, but we can use IBM's global reach to sort of bring Red Hat technologies across the planet. I think that has yielded pretty well for us, like in new logos and some larger deal structures. That's worked well. In terms of changes, we have to keep that independent ecosystem. There's a ton of opportunity in actually working closely with IBM. Our recent head of sales is actually from IBM to figure out how do we reduce friction in that model? How do we really show up well for customers and maximize that? I think that has been a really good work over the last few years of finding a fit in that model and then hopefully being enhanced by that going forward. That's great. We just have a couple minutes left, so maybe I'll throw the last question out here for you. Yeah. What are you most excited about from a tech or market lens perspective when we look out over the next 2-5 years? You're the perfect person to ask this given your long history of what you have done in product development over the years. If I think about where I started, click your time back, if it goes that far, to 1998 or so, this concept of e-commerce and the internet, it's going to change how we learn and how we program and how we buy stuff. Whether today is too far or too short for you, think of how everything you do today is impacted by this. It's changed the technology landscape. It's created things like public cloud. It has changed how we buy things as consumers. It's changed how we buy things as businesses. It's changed everything. When I look at AI, specifically, the reason I'm passionate about these smaller models in open source innovation, I usually tell people, "I don't bet against the world's creativity." That's what I saw with Linux. It went in that same time period from a hobbyist project from one person to the most impactful operating system on the planet in that span. I think we're going to see exactly the same thing in AI. You don't have to look much further than the traction in Hugging Face, the traction in universities. If you can link into that world innovation model, I think when we look forward, we'll come back and say it has changed every single thing we do, of how we write programs, how we run companies, how we operate as consumers. The only unknown to me is if that's going to be that 20-year lag or if it's going to be accelerated like everything else, and we'll come back and say that in three or four years. It's the Linux moment for me. All of the same intuition around that in this moment for AI right now. Really excited about that. Excellent. On that note, unfortunately, we're going to have to end it over here. Matt, thank you so much for all your insights. Really appreciate you being here. Thanks a lot, Shantanu. Thank you.