International Business Machines Corporation (IBM) Earnings Call Transcript & Summary

All right. We are super excited to have representing IBM, the CEO of Red Hat, Matt Hicks. And the last time we met, you told me kind of an interesting story about after Red Hat got bought by IBM, you took your team to see the quantum computer. I'm going to try to quote you, which is dangerous in a situation like this but I think what you said was we're joining the company that built this.

Yes.

And so I guess what was your perception of IBM when you joined? And how has it changed since you've been part of the team?

I think starting the whole reason I would take team to Yorktown to put your hands in this machine, and it probably uses parallel universes calculation. It was because there was just a lot of consternation of -- IBM is a big company, and we are open source and we were smaller and nimble. So there, I think there was that fear concern. And so that was the -- build some common empathy of we are fast and we are nimble, but we don't focus on something for 50 years to pull together in an actual working machine and IBM can. So it's straight. Looking forward to today, on one side, like I think the common empathy is established, like we're 5 years in, and we've really kept that independent model, which is let Red Hat thrive. So that's been really impressive. Then 2 of the -- can we work together? I always tell [indiscernible] year 5 is my favorite year because the AI work that IBM has done in Granite and InstructLab, they're not only open source, but we're basically using Red Hat as the go-to-market for those because it's our model. And so this has been a really like serendipitous year of like we're actually combining capabilities that we couldn't do independently on it. So it's been a good 5 years.

Well, and that leads us straight kind of into where I want to go next, where it feels like Red Hat is kind of the central cog to a lot of what IBM does, be involved in consulting, being involved in hardware. Can you just talk about how you think about Red Hat as a part of the broader IBM ecosystem, especially enabling those 2 pillars, the hybrid cloud and AI?

Yes. If we look at Red Hat, if you're not super familiar with Red Hat, we're a bit boring. Like we're an infrastructure software provider, like we boot servers, we boot clusters, we patch them and we keep them running for a long period of time. That's sort of our core. Like we actually wake up and get excited about that. We love that. We've extended that into AI. So like, well, the new hardware we will chase and support will be GPU-based hardware on it. IBM then layers on top of that, all of their strengths in middleware on it. So IBM has like the multi-class SaaS platform. By building that on OpenShift, they can run in all the clouds. They can run on-premise. They can use Red Hat's reach and infrastructure software to deliver on top of that. We have the cloud packs of, well, what if you're a WebSphere customer and you have a foot in that camp but you also want to leverage OpenShift, how do we integrate those technologies? So IBM, I think, has really effectively taken their base to be able to combine it with this. Going forward, if we look at virtualization, it creates really good opportunities for Turbonomics and Instana. So I think it just sort of starts of we try to be as ubiquitous as we can, run on all the hardware, give customers choice, really provide that value. And then IBM is able to bring in their software elements to that to really help you drive efficiencies.

And then when we're thinking more on the economics of it, I guess, you're coming off a really strong quarter for Red Hat, reaccelerating growth up 14%. Can you just talk a little bit more about kind of the strengths and drivers of that? And then how you're thinking about that setting up for the future?

Yes. So yes, I'm not going to complain about a quarter like that. That was -- for those of you that -- for us, it wasn't a huge surprise. Like we started talking about bookings pretty much a year ago to this. And so the momentum that started, that was realized in revenue this quarter started almost a year back. The fundamentals on this motion-wise is if you look at the strength of RHEL, there is just efficiency at this point in standardizing on one Linux stack for your enterprise. If you care about patching, if you're in a regulated industry, if you're dealing with cybersecurity, we patch your stuff today and we patch your stuff from yesterday really well on it. Doing that across multiple vendors tends to be inefficient. RHEL is a great answer. So that's driven some growth there. If you're spanning on-prem and hybrid cloud, multiple hybrid clouds, it just gets more efficient. OpenShift then picks up that same strategy and lets you build a new class of applications to that model. OpenShift growth has sort of not slowed since we started luckily. And Ansible then automates all the human things that stitch those pieces. Going forward, we have some really strong trends around virtualization, which for us, you would see that in OpenShift growth. OpenShift is a really strong platform. It's based on Kubernetes. Last quarter, we disclosed it's a $1.3 billion business at this point. So it is a strong platform. I'm highly biased, but I think it is a great bet for virtualization for customers going forward. We have the AI opportunities like nascent early market for us. But I think the same strength we brought to RHEL with CPUs in the days, we bring to AI with GPUs with RHEL AI. And then if you go like further out from that, we boot servers. The servers 3 years from now will probably be growing more outside of the data center than in the data center. Those servers are going to be in your cars. They're going to be in the manufacturing plants that deliver goods to your house. We bucket that up as Edge. But the same value prop that we provided to CPU-based servers 20 years ago, we provide into cars today because their data centers on wheels basically. So those are sort of the big growth drivers, I think that's why I can be excited about the quarter and then excited about the outlook as well.

Well, and maybe speaking to a value prop you're very familiar with and thinking about AI. So IBM made the decision earlier this year to open up the code base and the granite models. you qualify as an expert on open source, I would say. Can you just talk us through kind of the reasoning behind the decision and why you think that's the right strategy?

Yes. The warning, I'm a shameless supporter of open source. I've done it for 20 years. If you stay in that madness for a long time, you got to believe in it. And so I believe in open source, I believe in this model, and it comes down to like who on the planet drives innovation. So the reason Linux beat all the other major very capable operating system providers is they could get planet -- they could get talent from every corner of the earth to do contributions, make memory management better, make storage throughput faster. It was not limited to one enterprise structure to drive that. AI, I think we're at a very similar moment, which is where are you going to draw your innovation base from? It's either going to be like in the 100 or 1,000 people that I can hire or it is can we find a structure where I can actually draw on innovation from the entire world on this. And I usually, at Summit, I gave the example of like what comes out of MIT today. We work with MIT closely. I believe it's going to come out of the IATs in India tomorrow and in other university systems after that. So I'm a believer of like if you want durable innovation, open source is the best model to be able to give that to you. And then the second piece on this, I think you could care less about that and you are an enterprise like myself and you're going to build the future of your company and put your IP into one of these models, you need to make sure that it's still your IP at the end of the day. There's a lot of different like licensing and contract nuances out there. We try to pick the clearest path possible on open source, which is Apache licensing. Of anything you put in that model is not ours, it's yours. Those are sort of the 2 fundamentals that I really think lets you tap into just like the incredible talent out there to make these models better and better and better and let them put their fingerprints on it. That's why like people at Red Hat, we believe in this. We have the tattoo crowd that we say like that have our logo tattooed on their bodies, and it's because of this. And then the second part is a company of this is probably one of the most important things you're going to pursue, and you don't want to lose your business in the process of it. So those are the 2 big tenets we try to land with that one.

Yes. I've been trying to be a big supporter of being RBC tattoos, but there's been...

It's pretty cool. You got to start. You have to....

There are worst logos. Yes. So, we now know why you want to get to open source. But given Red Hat's success of actually driving that, harnessing the global power, what do you see as like the early steps that you're starting this over again within Gen AI to recreating that same momentum? And what do you look at to know if you are capturing the attention of developers?

Yes. There are 2 pieces we sort of start with to look at. One, you have to be comfortable that you start non-commercial on it. So if you look at -- for us, if you look at the InstructLab project or Granite itself, Granite is available on Hugging Face. You can get it from Ollama, I ran it on my laptop today. InstructLab, I can download and just run on my Mac. And the reason we do that is we always start with -- you can think you're building something of value but it has to deeply resonate with your users. And they have to be able to get it and kick the tires and work with it, see it works for them well before they're going to stroke you a check for it. And so we start there of how accessible is this just for users. The second piece is how many of those users actually send something back to us. I used to say like developers, we're a fickle crowd. We'll use all sorts of stuff. But when I put like my IP into making your project better, I'm pretty close to a believer for life. And so we call that a contributor community. Can we move this from a usage community to a contribution community? That is why we did the work and like you can send us your knowledge, and we will build it into Granite ourselves. We'll pay the training cost to it to make that shift. Then the last aspect we look for on this is when we're starting, who else is going to carry this message like open source is still unique, like open source without friends can be a pretty lonely space in it. So what does your channel look like? Who else believes that you can do this? In our world with RHEL AI, it's -- we start with OEMs. Can we sign up OEMs? Can we get them to actually carry this product? And we had announced we have Dell and Lenovo at this point already signed up. Those are our early signals of you can make open source projects successful, but they can have no commercial value to them. So if you go to make them successful and have a usage and contribution community, what's a good proxy to see if anyone else sees that there's business value and it tends to be the channel. So that's where we start, and then it's a tremendous amount of work in the middle. I'm happy of this Facebook and Insta.

Maybe staying on Gen AI. One thing we talk about within our team sometimes is how different this trend might have looked in 2021 when we were going through like the huge software spending booms. Today, it feels like we're in a much more cost-conscious environment. IBM seems to be taking a very ROI-focused approach to Gen AI. And can you just kind of talk about why you think that's the right model, what you're hearing from customers that lead you towards that? And just kind of the flavors of small versus large models?

Yes. I mean I think it's a financially healthier time than when you say the only thing that matters is growth, and it doesn't matter if I make money or lose money, I just have to grow. It leads to weird behaviors. We're a bit in a balance like the need to consume technology. But on measure, it needs to return more to me than I lose using it on it. And so this has led, I would say the first thing is that like the large model versus small model debate. If we were in 2021, I don't think there would be a debate. It'd be large models all day on it. But there is this practicality of I can use large models to prove the technology, but I cannot run them at volume for my business and still control my margins on it. So that's been one change. That's a pretty pervasive one. Like you will hear this large versus small debate. The other one, I would say, is predictability and time and execution. A pretty large GSI we met with last week actually, had gone through and sort of bake off our offering versus the other areas. And we do what's called synthetic data generation. part of open source is I have to make it approachable to everybody in this room, whether you're a data scientist or not. And if I tell you, you have to come up with like 10,000 super nuanced, well-formed examples to train a model, there aren't going to be many people that have the time and patience to do it. And so we ask you to come up with 3. And then we'll use like MIT math to expand it to the 10,000 to change weights. The GSI's view of that was I have to deliver predictable projects in this environment. All this GPU stuff you tell me is great, but this synthetic data means I can come in and just deliver faster. That is another cost-conscious aspect. I think that's been newer. And then the last one is not just cost. It was actually a state organization where there was a cost element of they have enough money to train a model, but they do not -- they're a state. They don't have the budgets to run this in a cloud for all of the agencies that consume from them to hope that the idea works. So the fact that we could take a model, it still costs money to train it, but let all of the other agencies run it on the laptops that they have today with no commercial contract, just to understand this and understand if it works for them or not was really meaningful. Those, I think, like there was an element of IP for them. They're going to lean towards open source. But there's a practicality of like I have to do this at scale. I have to deliver projects consistently at an aggressive price. I have to build something and let users really see if it's valuable before I sign up for massive consumption plans. Those are just a couple of examples literally in the last 2 weeks that I've been through there.

Yes. No, that's super helpful. In some of my checks, the idea is almost due to query costs, that some projects, the biggest fear is it's too successful because if you're losing money on it every time someone uses it, then that's a rough project.

It is. And in 2021, we weren't asking that question. It was like -- and there are a couple of super popular AI companies that I think can operate in this model better than the typical enterprise. Like they're not going to go raise a multibillion-dollar round to go augment their IT efficiency project. So I've loved it. I think it's a healthy focus that exists now.

Maybe one more on efficient Gen AI. Can you just talk about your proposed acquisition of Neural Magic?

Yes. So RHEL AI, oversimplified, I -- my team's looking angry with me, but I'll put it in 3 pieces. We needed to have a model in there, a model being the thing you could ask questions to and it is a [indiscernible] language stuff. And that's granted for us. Then we sell to enterprises. Those models are all trained on the Internet. They know nothing about your business with it. And so if you're going to use any of them, you have to be able to put your knowledge into it. And data science, like that's a limited skill set that can do this. So we try to democratize that with InstructLab. So you got a model. Now you put your knowledge in it. And then what is the question that anyone asks that's successful with this at all is, if I'm wildly successful with this, do I lose control of my own margins? Do I make more for every invitation of this model than I lose in the GPU costs on it. Neural Magic for us is extending that depth. We call it in the space, we call it surveying or inference, but it's the cost of every you ask a model. Neural Magic has some of the best-in-class CPU optimization. They're deeply involved in vLLM, but it's all just in we're going to take each question you ask down and divide it by 10 and divide it by 10 and divide it by 10, so you can do more and more exciting things without losing your margins, losing your end outcome. So we've been excited about that one, minus the disclaimer, not closed, intent to acquire all this.

So I feel like we've been going to the other pillar of hybrid cloud. We've been in this conversation about the right balance of hybrid over the last decade. I guess post-cloud optimization, how are you seeing conversations change about the idea of the balance between on-prem, cloud, private, Edge, all the flavors?

It's the first wave, we sort of live in hybrid on this was, I'm going to burn down all my data centers, move everything to the public cloud because they're good at running them. That was sort of wave one of I don't need hybrid. I'm going to put everything in one place. And public cloud, don't get me wrong, they're great at what they do, but I haven't met a single enterprise that made it to the other end of that journey. Maybe in the journey, they acquired more companies that were in other clouds. Maybe they got to like core data that they weren't able to move that was too entangled with other parts of their business. So I think there is an acceptance like whether companies like it or not, hybrid is here to stay. Azure and Amazon are not going to make their offerings the same for you ever with it. They're going to differentiate. So you're going to be in multiple camps forever. The changing dynamic, I would actually say AI -- and that can sometimes be a negative conversation like I still wish I could get to one, help me get to one. AI changes that. Think of the best outcome we start with when you're done training this super exciting model that's at the core of your business, where is it going to run? Like where is the best place you want to drive for the non-tech like you generally wanted to run as close to your customer as possible. You're in automotive, you will run it in the car. There is no calling out to the cloud to decide if you're going to move the wheel right or left. If you're in banking, you do retail, you're going to run it in the branch. Worked with enough banks like those branches will run offline from the mother ship, like they -- that is where you are actually touching your customer on it. So this pull, it now has like strategic value to -- whereas it was an annoyance before, I just want to consolidate all things, put them in Amazon, Azure, IBM Cloud, whatever be done with it. Now it's reframing the -- where can I actually drive value from running outside of just one location. So that's been really positive. And then with that becomes where does my data exist that I can actually train that thing with. We talk about like all the landscape shifting. The majority of this data still sits on-prem. And so all of a sudden, you now have a strategic value to on-prem. You can leverage public clouds where they're strong and Edge becomes very real to you because you're going to start like G&A optimization, but not many companies' IP is based on G&A optimization. Like they're going to want to get this to self-driving vehicle if they're in automotive or the tellerless banking if they're in retail banking. So that power, the benefit to companies has shifted. So it makes hybrid -- it's a lever I can pull to help me realize like my business goals, not just an unfortunate consequence that I couldn't get to one place.

And then on virtualization, so we've heard a lot in the market recently about some changes that Broadcom has made post buying VMware. VMware traditionally a provider of virtual machines. How does this opportunity to the extent you feel there is one match up with Red Hat's product set?

Yes. It's a great opportunity. I think one of the reasons that if you look at our quarter and say like, well, can this continue and you'll sense confidence from me, Arvind, this is one of the elements like we're confident in AI. But when we look at virtualization, at its simplest form, customers like in technology, you have to sort of pick a platform at points in time that you're going to invest a lot into. You're going to train staff around them. You're going to build your applications, Jim. You're going to run these applications, whether you like it or not, for decades on that platform. And there was mainframe, then there were sort of like UNIX, x86, and then virtualization came around in containers. This has driven what was a platform choice that's been around 20-plus years. It has changed the economics of that platform trace. So for us, it's created a compelling event of -- at the price increases is still the right platform for me to invest in. I usually tell customers, I'm like, that's where your logic has to start versus us or them. But then the most -- if you say, no, the economics don't work for me anymore, what is the platform that works for you? I'm highly biased on this, but our goal with OpenShift, we talked in the quarter, significant business already. Most of these customers are already running OpenShift. They've already trained staff on OpenShift. So there's a human element. Their careers can benefit from this. They've already designed applications to OpenShift. You just want to add virtualization as another workload to OpenShift itself. Then if customers say like, well, are you credible in virtualization? We've been in this market for 15-plus years. We run some of the largest virtualization deployments in the world on it. So our comfort in saying, yes, we can support KVM really well for the typical enterprise if OpenShift is the right platform for you. My bias pitch is OpenShift lets you take advantage of containers. We do virtualization well, and it sets you up for work that you're going to do in AI because we have RHEL and OpenShift and RHEL AI and OpenShift AI that play together. I think it's a great bet on the technology side of something you're going to get the returns from for 20 years on it. But that's sort of where I start of you got to figure out the right platform strategy that's going to progress your business on it. And then I usually go into my OpenShift sales pitch because like that's what we believe in on our side. But it's a really good opportunity because it hits pretty much every enterprise out there, and it makes you take a decision that no one's had to take for 20-some years.

So we're at about 4.5 minutes. It's a full room. Are there any questions from out in the audience? All right.

Red Hat Under IBM platform helped many, many, many companies improve their operating efficiency. Now we are seeing, or at least that's the plan, the government would like to improve the efficiency of massive government operations. So how can you help them?

I mean one disclaimer, you're asking a highly biased person on this, but I'll go to the -- because -- so one, like we work already significantly within public sector. I'll go to the reasons of why this is so impactful. You're the government serving your populist. Forget any nation in the world, the same thing like that is your job to serve your populist. You probably don't want to be beholden to any one company that's going to choose how much value you can hand to your citizens versus how much value they keep in it. In an open source model, my world in working with governments is pretty clean on it. It's their IP. If they build up a significant enough staff, they can run it themselves. They are never beholden to what I give them and what I enable them with. I just have to say, I think I can do support better than you. I can always make this a positive thing where you are relaying more value to the populist you serve, then I am holding back in it. But you have that choice. This does extend to enterprises. I've always said the open source can drive you crazy some days. It's a tricky model, but companies can make the biggest bets in direction on open source because they know they own it at the end of the day. I think that really extends into government. I think it's the best choice of spreading the economic pie of those things as an offering. So yes, so I'm excited about those opportunities, but I think it's a great parallel.

All right. Maybe a good way to kind of end this is when you're thinking about Red Hat over the next 5 years, what are some of the things that you're most excited for? And then maybe topic du jour, like how do you see maybe Gen AI evolving in the next 5 years?

Yes. I think I'm an old-time Red Hatter. And so when I started and we walked into this sign every day that was, first, they ignore you, then they laugh at you, then they fight you, then you win it. Like every day, this is what you walked into the building. This was our battle with Linux of no one knew what open source was. It was ignored, it was laughed at, it was fought, and now it is the default. When I talk to Red Haters, I'm so excited of like we get to do this again in AI because all the things that drove me to Linux of just the power of I can put my fingerprints on it, I see the world gravitating to this. I know it will win. I have the exact same intuition on -- I'm a techie, I watch Hugging Face. I watch this, I watch academic universities. What are students gravitating to? Where is innovation coming from? How do you unthrottle that? And can you find that fair structure and balance getting to like replay that in AI, which I think will change software, it will change how enterprises run and can be done in a really positive way. It is the part I love about this because it's something we know in a very new and challenging space that just makes you rethink through the models and patterns. But yes, that's probably the thing I love the most right now.

Well, we're staring our red lights. I think we're out of time. But Matt Hicks, thank you so much for spending time with us.

Thanks a lot.