AI and The Future of Quantum: A Conversation with IBM CEO Arvind Krishna at SXSW 2025

What could the future of enterprise AI and Quantum look like?

🔮 Daniel Newman caught up with Arvind Krishna, Chairman and CEO at IBM, at SXSW 2025, for a captivating conversation on the current and future landscape of artificial intelligence, its economic impact, the integration of AI with quantum computing, and the transformative potential of these technologies across industries.

‍

‍Key takeaways include:

‍

🔹AI’s Economic Acceleration & Enterprise ROAI: The remarkable growth in AI’s projected economic impact, with insights into enterprise AI implementation and outcomes, reflecting on scale, workforce productivity, and return on AI investment (ROAI).

🔹Strategic Model Deployment: The strategic approach to System Lifecycle Management (SLM) through the deployment of smaller, purpose-built enterprise models, their performance, and the anticipated evolution of this trend.

🔹Practical Application Timelines: Progress on the quantum roadmap, and the expected timeline for quantum computing to gain practical application.

🔹Quantum-AI Synergy: The synergy between quantum computing and AI, highlighting the most promising real-world applications and their potential to revolutionize various industries by addressing complex challenges that were previously deemed unsolvable.

‍

Learn more at IBM. Watch the video above, and be sure to subscribe to our YouTube channel, so you never miss an episode.

Daniel Newman:
Hey, everyone. The Six Five is On The Road. We are here in beautiful Austin, Texas, South by Southwest 2025. I’m Daniel Newman, CEO of the Futurum Group and co-founder of the Six Five. Excited to have this conversation, this exclusive conversation here at South by Southwest with a guest from the past and someone I’ve always enjoyed very much talking to Arvind Krishna, CEO, IBM. Arvind, welcome back.

Arvind Krishna:
Thank you, Daniel. It’s always great to be with you and in your hometown.

Daniel Newman:
It is my hometown. It is great to be here. It’s a beautiful day. You just came off stage. You delivered a great keynote. This place is flowing, it’s packed, it’s really back. And AI, of course, is one of these kinds of overwhelming themes. But I want to start, Arvin, I want to go into history a little bit. You and I, not in such a warm climate. What about 14, 15 months ago, stood on the rooftop in Davos at your IBM location and we were talking about $4 trillion of AI economic opportunity. Now, over the last year, we’ve heard that number go from 4 depending on the source, 10, 20 depending on the horizon. Let’s just start there and talk a little bit about what you’ve seen and how quickly this has moved over the last year.

Arvind Krishna:
Yeah, look, AI is moving really fast. It is going to be one of the technological revolutions. There have only been seven or eight in all of history. Steam engine, electricity, kind of take your pick when you think about transformational technologies. And in the end, the benefit comes from everybody using it to become more productive. That’s the 4 trillion number. Whether it’s 4 or 8 or 10, it doesn’t matter. Four is big enough. If you think about getting 4% more productive, that’s really hard to get done. Could it be 8 or could it be 10? Sure it could, but let’s get to 4. And so I urge everybody to get going on the journey. Start with the lower risk use cases, get your company, your government, your country used to them and then you’ll get the full benefit.

Daniel Newman:
And let’s talk about that a little bit. I was one of the analysts that kind of early came out and talked about IBM having this fully governed data AI solution for enterprise. And that’s I think really important for enterprise. You’ve been focused from the very beginning. How is the journey progressing within IBM in terms of building that, getting those customers on those early use cases and heck, even you guys being customer zero, which I know you’ve talked a little bit about over the last few months.

Arvind Krishna:
So a lot of questions in the head. Daniel, let me unpack it a little bit. So if I look at how it’s going, look as a commercial company, you get measured on revenue and signings. We announced in early January a $5 billion, inception to date book of business in AI. I’d say that’s probably the best signal of it’s going well. But then when I begin to unpack from beyond that, I think that if I look at our customer zero, how do we use it to improve customer service for all of the 3 to 7 million calls a year we get? How do we give more self service? How do we get problem triage done quicker so when people call us with a problem on their computing systems, we can help them get to an answer in minutes, not hours? How do we get going on enterprise processes? You talked about client zero. Applying it to ourselves, we have now generated over $3 billion of productivity inside the company over the last two and a half years. Leveraging AI and automation, I’d say that’s a pretty strong signal. What’s an example? Can we do an HR chatbot so you don’t need to know the intricacies of your HR system? How do you promote people, how do you move people, how do you verify employment? Can all these things be done using AI? And we had a great proof point that that is true. And now I can imagine procurement, supply chain, cyber, so many more use cases where AI can go.

Daniel Newman:
Yeah. And in tracking the journey, it’s been really good to see how you’ve been focused on showing that ROAI, I call it return on AI because you know, we’ve seen this kind of early wave of the market and the market got very obsessed with the infrastructure. We’ve seen the build out and what we don’t want to end up having is shadow cities. We need this infrastructure to be put to use. And being a company focused on consumption, one of the things you talked about at Investor Day, new facility, by the way, One Madison. Beautiful. And it was great to have the opportunity to spend some time with you and your team there. But you talked a lot about kind of, you know, models being more and more commoditized. I think you could say there’s a lot of people, a lot of your peer group, including you, are kind of looking at how this is evolving, size of model, energy, accuracy. And in the enterprise it’s going to be all those things because they have a lot of technical debt and infrastructure. Talk a little bit about how a smaller purpose built enterprise approach to AI is going to be so important going forward.

Arvind Krishna:
So first, on a side note, since you mentioned One Madison, where you and I met last time, One Madison is our new flagship location in New York City. Why I think it’s so important and you mentioned it, but I can’t help but talk about it is culture. The culture of people, having that has brought people back in. So they’re there three, four days a week, Tuesday through Thursday. You can’t find a place to sit. It is that packed. And you can see the buzz all the way from 7, 7:
30 in the morning to about 6, 7 in the evening. I think that that piece about culture and having people engaged is quite important to your main question you’re asking. Look, we’ve always been a believer that smaller fit for purpose models is what the enterprise needs to answer your question about enterprise.

Daniel Newman:
Yep.

Arvind Krishna:
Enterprises are not trying to get a one model. I kind of say it somewhat jokingly but somewhat seriously. If you are having one model that has to answer all questions that a consumer might ask, you might ask it to help you write an email. You might ask it to summarize a document for you. You could ask it to translate Finnish to French. You could say write me a haiku. You can say compose something in the voice of Steinbeck. Okay, that’s not our world. Our world in the enterprise is you might say summarize this legal document and tell me what’s the difference between these two versions? I don’t need that bigger model. Actually our proof point, a 6, 7, maybe 10 billion parameter model can be more accurate than a 100 billion parameter model. And what we showed at our investor day is that these models can run with 97% less energy than the big model at slightly higher accuracies that I think is remarkable. But that sort of proves the thesis that smaller fit for purpose models are going to have a huge role to play in the enterprise. By the way, 97% energy also means 97% less GPU cost.

Daniel Ne wman:
Oh yeah. And I think for everyone out there, it’s sort of important to double click on that, that level of efficiency. We heard about the reopening of Three Mile Islands and we heard about, you know, nuclear. And part of the solution is efficiency. I think whether it was the deep sea moment, like we can argue about all the semantics of it, but figuring out more efficient ways to do better AI with less parameters, less energy and less infrastructure has to be the answer if we’re going to get AI to where we want it to go, yeah?

Arvind Krishna:
Also just the physics of it. When people talk about Manhattan sized data centers, we’re talking about a 20 year journey. We don’t want to wait 20 years. So people forget that it takes that long to get physical infrastructure done. And so put aside the implications on energy usage, electricity prices, water consumption, amount of semiconductors. So that’s why efficiency is really, really important. And that lets you really, really win, but also deploy faster, which is what we all need for quality of life and productivity.

Daniel Newman:
Yeah. And it’s worth noting, you know, our performance team at Signal 65, we actually evaluated Granite, we tested your claims and we found them to be absolutely accurate. And just anyone out there that kind of wants to know more about it. But the point is like some of that stuff was almost hard to believe. 97% less energy and more accurate, that’s a big thing. Let’s move on to Quantum for a moment. Arvind, I know Quantum sort of had its moment, then the moment goes away and then it, you know, then a bunch of companies that used to sell energy drinks become quantum companies and they become worth $2 billion. And then, but full circle IBM, you know, jokes aside, have been in this market, committed to this market. Talk a little about your progress and your view on Quantum and its intersection with AI.

Arvind Krishna:
Yeah, so first when all these other companies talk about it, it actually excites me as opposed to thinking about it as competition. Because in the early, early stages of a market, it’s really hard to go out there and explain to people what is this? Why is it good for you? So if everybody else is also doing it, I actually call it, we are making a market. Now, once the market is made, we don’t have to walk in and explain why it’s useful. We can then talk about why ours is better. So when we think about it first, to really win in quantum computing, you actually need a quantum computer. I kind of say that tongue in cheek because many of the others talking about it don’t have a quantum computer. Ask them a simple question, can I access it in reality or over a cloud or over a network or physically? And your answer is not really. We’ll get to you when that’s possible. We have built 75 of them, many of them still functioning. We retire them when the next generation gets better. We have 13 of them right now. You can access over a cloud, each one at 100 qubits or better. So that’s kind of where we are. We believe over the next three or four years you’re going to start solving problems that will surprise you. We need to get a bit better on errors and a bit better on coherence. Sort of coherence is a term for how long a quantum computer works and does computations before it kind of becomes just noise. And so those are the two, but we’re talking like we are within a factor of 10 on each and that we’ve been improving at three times a year on both of those. So that tells you we’re not very far away.

Daniel Newman:
Yeah, it’s super interesting because you really are getting a continuum of answers. Of course there are business leaders that are in the AI space that are sort of downplaying quantum, but of course it would potentially be in their best interest to say less. And then of course you kind of mentioned the companies that don’t actually even have a quantum computer, but are kind of hyping up the idea of quantum, but can’t put two or ten or twenty logical qubits together. So how fast though do you sort of see this acceleration? Because I think that is the big mystery. What’s the IBM viewpoint there?

Arvind Krishna:
So we’ve got a sort of baseline. Where are we? Yeah, so we are at about a little over 100 cubits, 127 if you want to be more accurate. And we have to do error correction. I’ll say that’s step one. I think that’ll take about two years. When you do error correction, it effectively decreases the size of your machine by 10 times. It used to be a million times smaller. And that’s when people used to talk in those ludicrous numbers that you need 100 million qubits to be real…no longer. So maybe you need 1000 qubits to have 100 completely error free qubits. That sounds pretty exciting. We know how to do that. We publish papers on it. So that is now just getting it built, probably two years. Then the next part is can we increase the how long or coherence time. We think that that journey is probably a two to three year journey. You put the two together and I think we’ll solve really interesting problems in 28 or 29.

Daniel Newman:
So what about the intersection though with AI? Because we do hear a lot about it, even an example is could quantum networking drastically decrease the amount of energy required for GPUs and quantum computing? Like what are some of the use cases and intersections with AI that you’re really paying attention to?

Arvind Krishna:
So I think that some of the optimization techniques and all those will come early on. To have quantum computers really build a model for you which is kind of what excites people in the medium and long term. I do think that that’s a bit further out. So I would not say that that is going to be happening this decade, but maybe in the next decade. The reason you’re not hearing me jump up and down on AI models. How about lightweight alloys so we can build stronger but lighter weight vehicles? How about better batteries so that electrification can go faster and easier. We know there’s 100 times gap between the chemistry of what these batteries can do and what we get. That’s exciting. How about fertilizers using less energy so we have a more stable food supply. How about carbon sequestration so we don’t warm up the atmosphere. I got all these problems that I’m super excited about, that can be done actually in the early part. We don’t have to wait for these computers to get that big to do the others.

Daniel Newman:
Just a little bit more on that and really appreciate the time. And I know you, you’ve got a lot going on here at SXSW. Everybody sort of gets lost in this quantum conversation about what we should do with it. Everything you just said feels so practical to me. Maybe I’d add one more like fraud detection. I’ve heard that one a lot of times. Are you saying though that you think like between say now and maybe the turn of the next decade, so 2030 we’re going to start to see quantum with classical computing working together to develop things like materials innovations in a way that people can wrap their arms and commercialize. Because that’s what’s been in my head is like how do you buy the stock? Because IBM has a real business. But some of these companies like how do you buy stock when I can’t figure out what’s commercialized? I’m hearing you’re saying there’s a commercialization that’s coming towards the end of the decade.

Arvind Krishna:
There’s a commercialization that’s coming, I believe around the end of the decade, maybe one or two years earlier, maybe one or two years later. You always get it when something is still three, four years out, give yourself a little bit of a window. And you mentioned some of the fraud. I think another great example, you know, right now, this last week, markets have been somewhat turbulent. I’ll use that word. Is that fair?

Daniel Newman:
Just a little.

Arvind Krishna:
So when you do that pricing on financial instruments, which by the way flows into the economy, it flows into mortgage rates, flows into what rates you pay on bonds and all those things. If you can figure that out better using quantum, that would be quite an advantage for everybody, wouldn’t it?

Daniel Newman:
Oh yeah, no, absolutely. I’m really excited about it. And of course we’ve been tracking your journey for quite a long time and what you’ve been doing at IBM, I think the company is just about at that turn where maybe some of their financial model will start crediting you for all the innovation because the great work, the fast growth in AI the market is seeing it, the great operational excellence, the way you’re able to return value to your shareholders has been very impressive and the company’s been very stable through what I would call really a turbulent 4, 5 years ups and downs. So I want to congratulate you on all that. Fun question to take us home. You’re here in Austin. I know you’re not here for very long. I think you were just at the White House. So you’ve been busy doing things. Do you ever get barbecue when you come to Austin?

Arvind Krishna:
I do sometimes. It’s always when I have the pre thought to think that I need to go at nine in the morning and go get it. Because as you know, the best barbecue in Austin is in the morning, not in the night, as people know.

Daniel Newman:
Well, thanks for all the work you’re doing. Thanks for, you know, going to DC, having those conversations, coming here to SXSW helping the world understand the innovation and the work that’s being done by IBM. We really appreciate having you regularly here on the Six Five Arvind. Hope to have you again soon and have a great rest of your South by Southwest.

Arvind Krishna:
Thank you. And I always enjoy spending time with you Daniel.

Daniel Newman:
Thank you so much. And thank you everybody for tuning to this exclusive conversation with Arvind Krishna, CEO at IBM. We are here at South by Southwest 2025 in beautiful Austin, Texas. Maybe, just maybe, I can get him back here for the F1 GP. I know he’s a fan, but if not, we’ll be talking more IBM, we’ll be talking more AI, more quantum here on the show. Subscribe, stick with us. We’ll see you all later. Bye.