Executive Vice Chairman, IBM Common Stock
America's Technological Future with Gary Cohn, Rene Hass and John Stankey at CNBC Invest in America
Questions persist about AI’s return on investment, workforce disruption, supply chain resilience, and the role of policy in unlocking private capital. America’s leading technology and telecom companies are answering the call. The challenges are real, but the pledges to deliver are historic.
Gary Cohn, IBM Vice Chairman
Rene Haas, Arm Chief Executive Officer
John Stankey, AT&T Chairman & CEO
Moderator: Sara Eisen, CNBC
More from the CNBC Events: https://bit.ly/40ZdNd1
Subscribe to the CNBC Events Marketing Newsletter here: https://cnb.cx/3nJqzht
Follow on Instagram: / cnbcevents
Follo...
About Gary Cohn
Gary Cohn, vice chairman of IBM and former director of the National Economic Council under President Trump, has appeared on multiple CNBC programs in recent months to discuss the economy, Federal Reserve policy, and the impact of artificial intelligence. In a June 2026 appearance, Cohn said that without the AI and energy sectors, the stock market would be "floundering" and described the two industries as "intertwined." He stated that he believes computing capacity will be overbuilt and will become a commodity, with companies purchasing it from the lowest-cost provider. Regarding AI's effect on employment, Cohn said he is "in the camp that this time is the same" as past technological advancements, arguing that such innovations historically have not led to the "demise of human capital" but have instead grown GDP and created more jobs. Cohn also commented on economic policy and inflation. In June 2026, he said that if a deal to open the Strait of Hormuz is signed, oil prices would "not... fall like a rock overnight" but that a change in psychology could lead to lower prices over time. On the Federal Reserve, Cohn said that new Chair Kevin Warsh "will remove himself from the political pressure" and "do the right thing economically," adding that the Warsh Fed "will look different than the Powell Fed" with less forward guidance and data releases. In July 2026, when asked about IBM's software business, Cohn affirmed that the company's software is "not being disrupted by AI," and noted that companies are beginning to evaluate the return on investment of their AI spending.
Transcript (10 segments)
U
Unknown0:05
Are we overbuilding? Are we overbuilding data centers? I feel like there's going to be different points of view here.
G
Gary Cohn0:11
So let's break this question apart a little bit. Are we overbuilding data centers that can come online today or next year? No. Are we overbuilding data centers that are coming online in 28, 29, and 30? I happen to think we are for a variety of reasons I can go into, but just start with the most basic one. The AI technology itself is becoming so much faster and more efficient, and it itself can tell a data center how to run more efficiently. Then you throw in there the whole quantum computing piece. When we get to 29, 30, we're talking about machines that run tens of thousands of times faster than the fastest high-end computing machine today. What a quantum machine does that's different is it simultaneously solves multiple equations and it doesn't come up with zeros and ones. It comes up with almost any number you want. So we're going to be able to compute quicker and faster, and we're going to get more efficient in the premise of how we compute and how AI runs itself.
U
Unknown1:15
So they'll be obsolete. Is that what you mean? All the data centers?
G
Gary Cohn1:18
They won't be obsolete. No. Quantum needs data centers. AI needs data centers. But let's take an AI example where if you and I work at the same company and I go do an inquiry today on a document and I ask it 17 questions and I put notes in the document, and then you go run the document tomorrow, you start at ground zero. Why wouldn't you pick up from where I left off? So all that compute power that was used to get me to question 17, you now may not need the compute power because you may have the 17 questions I already asked it. That's what you were looking for. Or you take the knowledge that's in there and you start moving from there. This is a real enterprise solution to AI.
U
Unknown2:03
Why is he wrong, Renee?
R
Renee2:05
I'm going to disagree. I'm not going to say he's wrong. I think when you are in a transformation such as we are in, and I'll give a minor example again, when you looked at Nokia, BlackBerry phones that were basically keyboards, manual, and then the iPhone was introduced, people could not get their heads around what a touchscreen computer could look like and what it could transform. AI is very, very similar in the sense that the industries that have just not even yet been touched, take drug discovery, when you think about drug trials, when you think about inventing medicine, gene therapy, AI has not even touched that industry. World GDP is what, 130, 140 trillion? Could AI be 10% of that market? Of course. That's a 10 trillion dollar market. So no, I don't think we are. I don't think that the data centers we're building right now are going to bust or into too much capacity. I think you may see a shifting in terms of where these AI workloads run, but we are looking at a ramp of which we've never seen in the past, and that's why people are having a hard time getting their heads around. Gravity must take over at some point in time. Of course it will, but I would argue that it's still many, many years away. Back to this world GDP opportunity.
U
Unknown3:20
John, what do you think?
J
John3:21
I don't know. I'll make this observation why I'm comfortable. First of all, in terms of our investment portfolio, the dominant portion of it, 90 plus percent, is going out at the edge. It's going to end users. It's going to businesses. And I don't think their need for data consumption is going to change. I think that trajectory and that need is pretty well known and we can project that, and I feel comfortable that those investments are ultimately going to get utilized. What I will observe is we will not do middle mile and what I would call data center to data center work without permits and actual commitments to construction. There's a big difference between that number and the announced number. So, we will not put fiber in until we know a construction project is underway, permits are granted, work is occurring, and when you kind of temper down the announcements to what's actually happening, there is a difference between the two. And ultimately, may some of that get stranded at some point in time, it's possible, but I do believe that ultimately there's a clustering of data centers that come into particular geographic areas because the conditions are right, power's good, it's the right environment, there's good permitting, that kind of thing. And if data center A doesn't make it, but data center B does, I'm in an okay place as a result of that.
U
Unknown4:39
IBM is benefiting from, I think, the AI book of business. But there are also questions whenever you have a software company about who gets disrupted and what businesses will fall or will not be as profitable. Gary, how do you look at that and how do you look inside IBM and make sure that you are AI-proofed as a company that has gone deep into software?
G
Gary Cohn5:00
So I think I'll take your question with what John just answered. The core of AI and the core of quantum, the core of everything is data. Everything is relying upon data. One of the big breakthroughs that we've had in the last few years is we've been able to take disorganized data and use it as easily as we take organized data. And that's a pretty large breakthrough. And that has changed the way compute works. It's changed the way everything works. We at IBM, we're an enterprise company. We're not trying to sell anyone in here a direct retail product. We are trying to get inside major companies, even small growing companies, and make sure we set up an infrastructure for you that allows you to manage your data. As John said, data today is everywhere. It's in all shapes. It's in all forms. And the more you can put it in your system and the more you can process it, the better your outcomes are going to be. So, we are part of this AI revolution, but we're not the large LLMs. We're doing small LLMs for specific tasks. But we're most importantly sitting in the center of this sort of equation, which is data, data management, and data access.