Let's go. You're listening to Making Data Simple, where we make the world of data effortless, relevant, and yes, even fun. Hey folks, welcome back to Making Data Simple. We're going to talk about IBM today, which is exciting for me. I have Rob Thomas, who's the Senior Vice President of Software and he's the Chief Commercial Officer at IBM. That means he runs development, he runs our sales motion. He's a University of Florida grad, he's a Vanderbilt alum. I think everybody must know you, Rob, by now. I mean, you've been on the podcast like three times, something like that.
Great to be with you again, Al. And I'm impressed with what you're doing. I do follow everything that you're doing. Incredible followership to the podcast that you've built. So congratulations on everything you've done.
I think we got lucky. It's hard work and opportunity, right? But we started at the right time, and yeah, we've got a few listeners now. I'll go around the world and I'll find somebody, I think, hey, maybe I need to take a break, and then I'm in Spain and somebody says, hey, I listen to that podcast, and I'm like, I can't take a break. I'll keep it going. And then I get to talk to people like you, it gives me an excuse, right? So it works out great. So anyway, with Rob here, we were going to start it out at the beginning of the year just so he can set our sights on 2024, but then we thought, hey, we've got IBM Think, which is IBM's flagship conference, that's between May 20 and 23rd in Boston. That's where IBM hosts senior business and technology leaders from across the industries. We also have Tech Exchange, by the way, that's for technologist practitioners and developers. But next month we're going to talk to these technology leaders, and we thought this would be a great time, in preference to that conference, to really get Rob's view on IBM leadership, technology, whatever comes up. So we've got 30 minutes, I want to hit it hard, so I'm going to jump in if you're ready.
Absolutely, always ready.
First off, you still healthy, you still working out every day?
I do most days. It's a good way to kind of keep a routine going. But the problem with getting older is that it's way easier to get hurt. So I have found that mixing up doing different things, a little bike, a little running, a little walking, a little weights, makes a difference. Because if you do one thing over and over again, you start to get hurt, is what I've found.
I think you're probably right. And it gets boring. So you do that in the morning before you start your day, typically?
Yeah, default would be morning. Get up around 5:00 or so and do that right away. Then normally do a bunch of email and reading right after that. That would be the normal. But when you're on the road, sometimes it ends up being early evening, late afternoon kind of thing. So fit it in when I can.
When you can. I know you were pretty religious about that, make sure you have that healthy lifestyle amidst all the craziness. I read a book called Becoming Ageless, which I thought was very good. The thing that stuck out in the book, it was a lot of talking about how do you just keep in shape and up to speed, but it was also about being flexible in your approach. And I actually met the author recently, and I said to him, the thing that surprised me was when you were talking about doing two-a-days, because the last time I did two-a-days was like middle school football. And his point was that even if people get in a routine where they work out in the morning like I do, then they're like, I'm done, which is an acceptable answer. But he's like, then if you just do 20 minutes at the end of the day, that really compounds over time. And it's not about becoming a professional athlete, it's just about being active. I'm not sure I've committed to doing that, but it was an interesting takeaway from that book.
It's smart. Look, sometimes I get in the mode of quantity over quality. I'll go crazy, go on a run or something. You're right, and then you get hurt. So I'm learning myself to tone it down a little bit and just show up.
All right, so let's jump into some tougher questions. So it's been a couple years since you visited the podcast. First question is, where is IBM now? I mean, I don't really want to look back, I'd be remiss if I didn't ask you this question. Where is IBM now versus then, and how would you grade our performance if you put numbers on it?
I think if you go back four years, IBM was 45% to 40% software and consulting. Now it's close to 75%. So the nature of what the company does is completely different than even just a few years ago. Why does that matter? It changes the balance sheet and it changes how you can invest. So I think probably the most under-told story to date is how much we've invested into R&D over the last couple years, primarily in software, but it's also true in parts of hardware like storage. And this is really following through on day one, Arvind said hybrid cloud and AI. You can't really do that if you're not investing aggressively in R&D, engineering capacity, building new products. And I think we've definitely done that, and I think we're just getting started. You mentioned the Think conference before. We've also kind of said we're going to bring new innovation forward twice a year. We'll do the Think conference, which is really the flagship event, and then we have an event called Tech Exchange that runs in the fall, which is really more of a developer conference. And so as I work with the product teams, we're kind of focused on how we're going to show up really well at both of those events. We've got two windows a year to demonstrate where all the R&D investment is going, and I think it's a great way to rally the teams. Everybody always gets stuff done a little faster when you have a deadline, of course, and I think it's a great way for people to demonstrate that this is today's IBM, hybrid cloud and AI.
Right now it sounds like a no-brainer, but at the time when we announced hybrid cloud, I think there were some naysayers that would say, ah, that's because maybe IBM is struggling in cloud. But now it looks like hey, we had the right strategy all along. Can you say a few words about that?
I think both can be true. Let's be honest, IBM was struggling in public cloud. You have to admire what AWS and Microsoft have been able to pull off. But it was pretty clear talking to clients that moving everything to a single public cloud was not going to be the answer. And in comes hybrid. And there was a report from Barclays last week that said their data shows 83% of companies are now repatriating something to run on private cloud. And so the world is going to be multi-public cloud and private cloud. There needs to be a technology answer to that, and I think that answer is IBM. So I think we've actually found a really good position in the market. And at the same time, we've built huge relationships with Microsoft and AWS that's driving a lot of workload for both of those clouds. And I think that is the essence of ecosystem and partnerships: can you do things that are good for both sides? And I think we've found that right balance.
Do you think data centers are coming back in some way? The reason I ask is because in the move to cloud, a lot of people are checking the boxes, these companies are trying to get to cloud, but then they see the cost, then they say, well, now I need to go to multicloud. I see some of them actually saying, look, the data center is working today, why am I moving so quickly? Do you think that's just a blip? I read that projections say $400 billion will be spent on data center refresh in the next five years. That feels about right to me. So does that mean it's coming back or does that mean it's getting modernized?
Probably more of the latter. I think there's always been a place for the data center, but they'd probably gotten a little bit long in the tooth in many cases. Which is why when clients were saying, well, do I want to modernize this or do I just want to move everything to cloud, moving everything to cloud became an easy answer at the moment. But easy is often not sufficient, which is why I think it comes back to we're at the start of a pretty significant modernization of data centers for AI-type workloads, for containers, for multicloud-type workloads. I'm pretty sure that's going to happen.
Back to our question of then versus now, is there anything if you could go back two to three years you would tell yourself and say, hey, look out for this? Any mistakes that you would openly say, hey, we should have got this better? Any blind spots?
The only debate is which mistakes were the biggest, because there's always some. We invested heavily in building out foundation models. It was really second half of 2019, early part of 2020. I think we probably could have gone even more aggressively at that point. We made an investment that was fairly unpopular at the time, which was significant capital investment. I remember that meeting well. And anything in hindsight, I wish I'd actually gone two or three times as aggressive as we did. Just because we've gotten in this world now I'll call them AI years, which seems to happen every week or every day, that's how fast things iterate. And so had we done two or three times the capacity at that point, the cycle times would probably be even faster now. But we're still at the point now where we're iterating weekly. As I tell our sales teams, if you're using a model that we pushed into production 20 days ago, you're hopelessly out of date at this stage. And so I think we've done a good job of coming up the curve on that speed. But if anything, at that time, by the way, 2019, 2020, whether or not generative AI would happen was a complete unknown. So it's actually very easy to say this in retrospect, that I wish we'd done more, because at the time putting $100 million or so into something that you're not sure is going to happen felt aggressive. But now that I play it forward and now that I know it's happened, yeah, of course I wish we'd done two or three times that. But it was hard to predict.
Hindsight's always 20/20. You know what's frustrating for me? I tell this to clients all the time: we've been in the foundational model business for quite a while, and then the Transformer came along. But I guess you got to give a little credit to ChatGPT in that. I do this podcast and I've been doing it for a long time now, we talk about it all the time, and even my friends after ChatGPT, they look at that and then they go, have you seen this? I say, yeah, we've been talking about this. But give them credit for getting it out there and getting the mind share around it. But since you went right into AI, let's talk about this. I agree, I think it's like Moore's law happening every six months. I don't know what the math is on that, but the point is that we're rapidly changing, and that's an understatement. Let me ask you this question, this is a fun question. I banter around with my team, and I ask it because it really outlines the state of AI versus just asking what do you think the future is going to bring with AI. And that is: would you short Nvidia right now?
I'm certainly not going to give any projections on investment markets or stock. You're not going to get stock advice from me. What I would say is, I was an economics major in undergrad, and supply and demand is as true now as it was then. And so I would just look at what's happening in the market for GPUs. Are we supply constrained or do we have a supply glut? I think we're still supply constrained now. I would also say that every big demand moment tends to lead to a supply glut. So the only question is when does that happen? And as Warren Buffett would say, trying to time markets is a fool's errand. So I don't think anybody can predict this. I would just say at the moment, I think there's more demand than supply. At some point, that will probably change.
But let me ask you this in another fashion. The reason I joke around about Nvidia, by the way, I did a presentation the other day and Nvidia was on stage, and I said, well, I can't ask this question right now, that's probably not good. But the point is, it's not against Nvidia, they've been up like 2,000% over the last five years. It's just where the industry is going. And I got to believe, at least my take, and I'd like your feeling on this, is that models are going to get more purposeful, meaning they're going to get smaller. I think there's a lot of companies getting in the game to give competitive options to GPUs. If the models get smaller and you want to reduce inferencing cost, like if I want to put it on my laptop, I put Mistral on my laptop the other day and now I can run it, and it's a CPU and it answers like everything that I've seen out there. And going back to the hype cycle that you just essentially referenced that we've been going through, it's essentially got to correct itself, doesn't it?
I don't know. It's an unbelievable time for technology. I think we have to acknowledge that. If you think of ChatGPT as the Netscape moment for AI, let's follow through that analogy. Did Netscape set off an era around the internet that we later concluded, hey, the internet's going to slow down, the internet's not that interesting? I don't think that actually ever happened. Yes, there were ups and downs post-Netscape, but it's generally been up and to the right. The question is, does AI follow a similar path? And I think I can argue both. I think we'll certainly see some downs, but I feel like this is the direction that the world heads, as opposed to this is a blip.
Fair enough, fair enough. How would you articulate IBM's position on AI when your entire focus is on building AI for other businesses, meaning B2B?
I think the challenge is completely different than if you're focused on B2C. And 90% plus of the people I see in the market right now are going after B2C. And I think it's because the Google search revenue and profit engine is too attractive for most people to ignore. So that puts us in a different camp where we're not necessarily interested in how do you get more and more public available data, how do you get larger parameter models, more tokens. It actually puts us in a place where we need to be way more focused, because we want the data that's going to be interesting for B2B use cases. What will help customer service? What will help digital labor? What will help coding in specific languages, not every language? What will help governance of AI? What will help the next data architect? It's just a different problem set than most people in the market are pursuing right now, which I think is great. Which is why I think a lot of our discussions end up being, can we get you performance and accuracy with smaller models? Because smaller models have lower inference cost, and every business cares about cost ultimately to deliver on AI. So I think we're just playing in a different space than where most of the attention is in the market, and most people don't even make that distinction between B2C and B2B. But I think it's pretty fundamental.
When I do a pitch, I talk to three areas around the enterprise, and it really talks to the areas that maybe I'm most concerned about. I'm interested in your feedback here. That is, look, fake news within AI, I'll call it fake news, governance, being able to know where your data is coming from, make sure it's free of bias, etc. And then I think cost. To me, those are my top three that are impacting or are at the top of the priority list for enterprise AI. And we're addressing that through open systems, targeted, trusted, governance, etc., that I could go into. Do you think those are the three? Is there anything that you would add, or is there anything you think that I'm overstating with one of those three?
I think the real challenge for businesses is more about use cases and the competitive field where they play. And I was with a group of clients this week, and I made the comment that AI is probably not going to disrupt your business, but your competition using AI probably will. And I think every company has to think about that. Which is, what do I need to be doing to stay ahead of the curve? Because on one hand, if everybody does the same use cases, then actually nothing changes. Just everybody probably becomes slightly more productive, slightly more agile, but nobody actually gets an advantage. The advantage is going to come with a proprietary data set that's tuned to solve a problem that your competition is not solving, that you can solve before they do. That's really the kind of discussion that I'm in the most, is trying to help CEOs think through that challenge.
