Flex CEO on AI, manufacturing and jobs of the future

So the idea of having seamless AI integration that is going to generate information that's going to help make decisions I think is going to make be a game-changer for manufacturing.

Where is the world going in the next 5 years? And we have to have a systematic way of really reskilling the American workforce.

There's something magical about being on a factory floor and watching something being made. And nothing can replace that.

We make things for other people and we make everything. We'll make everything from consumer products like vacuum cleaners to hair straighteners to very complicated things like autonomous car compute, you know, to healthcare devices to kind of, you know, data center products.

Yeah, first, thanks for having me, David. And you and I first spoke in 2020, I think, just at the height of the COVID pandemic, and lots of different things since then. I would say, I remember my life then, you know, I would drive into shift change in the morning, leave my home at 4:30, get there by 5:30 for shift change at 6, and I had this notebook and pencil that I would kind of take the shift change notes from the previous supervisor. And then I'd translate all of that into this big whiteboard, you know, to make sure that everybody who came into the shift knew their task, because I had like 50 machinists, you know, big machine shop that we had there making hydraulic pumps and motors for the kind of construction industry. And it was all very manual. Every day we would manually exchange notes. We would manually give directions. I would take notes from the shop floor guys and then translate it back to the next shift. Nothing got stored. You know, we didn't know the issues that happened in the last shift the next day. And that was life then. That was in 1995. And I mean, you think about it now, right? And I'm sure our factories and those factories run similarly. A shop floor supervisor before they leave home probably knows how their day looks. What equipment is down? How many people are showing up? What setup changes? Utilization is going to run at 30% or 80%. Make all the tweaks probably in their bed in their house before they even come in, and probably driving an EV car or something like that. I was driving a Honda Civic. But the beautiful part I'm sure that's happening today in every factory is that all this information is getting stored, and next day when you have a machine down, you have all this intelligence that's making those decisions for you. You don't have to think about, oh, what happened 5 days ago and what did I do when that big machining equipment was down. So I'm sure it's a whole different world. At least I hope it's a whole different world. If it's not, Eaton's got a problem.

Correct.

Yeah, so I came into Flex 7 years ago and my whole career was with industrial companies like Eaton or Honeywell, working in aerospace, working in oil and gas, working in hydraulics, and then most of my career in the energy space. So when I came into Flex, I didn't know what it was all about until I found out more about it. It is one of the world's largest contract manufacturing companies, which means that customers give us their product and we make it for them. We could be making a very complex healthcare medical care device that some of you may be using, like continuous glucose monitors and things like that. We could be making very complex automotive equipment, or we could be making a vacuum cleaner that you use in your house. And today we make a lot of power infrastructure, compute cooling products for data centers. So we're the name behind the name, all the name brands you use. I always say you probably are using something that we have manufactured sometime during the course of your day. So for us, the complexity of we need to be able to make something that's going to go into the human body and the responsibility that comes with it, and we need to make that vacuum cleaner that you don't want it to break down and be kind of service-free for the life. So that complexity of manufacturing is significant and is always misunderstood that oh, it just shows up on your doorstep somehow. And that's what I want you to know about Flex, is that these 150,000 people across the world are showing up every day making extremely complex things that are hard to make that all of you get to use. And we have to do that with high quality and make sure that whether it's a pandemic or a supply chain crisis, our customers are still happy getting our products every single day. And I didn't appreciate that 7 years ago, David, when I first came into Flex, right? And today, 7 years later, I see the magnitude of what we're doing, particularly when this conversation of building manufacturing locally and close to the consumer and reshoring conversations are going across the world. I see the significance of what we do and what we provide every single day.

Yeah. So, you know, having worked for 30 years, starting from Shawnee, Oklahoma to Hutchinson, Kansas, I have seen everything come in and out of where manufacturing gets done, where industrial companies put their base in the last 30 years, right? And I've seen the whole variety of things that has happened. I moved to China and lived and worked in China for Eaton, running their electrical business. Saw the magnitude of what's capable there. And when I came to Flex and I looked at the consumer side of our business, it was very clear that we would never be able to compete in that industry. One, being a western company that's publicly traded, right? We have to deliver shareholder value, which means you have to keep improving your financials every year. And so competing with Asian manufacturers where they have a low cost of capital and an unfair advantage was never going to work for us. And then the product life cycles were so short that every day you were making something and then tweaking it the next day, and there wasn't enough money to be made in that value chain end to end. Right? The end customer or the end manufacturer may be making a lot of money, but we weren't. And so very quickly came to a conclusion that when you're at 2-3% operating margin, you couldn't live in that product life cycle. It wasn't rocket science to figure that out. Every job I've taken, the first thing I do is look at portfolio and mix. And I said, we just needed to get out of that. We're never going to compete and be successful at it. So really deemphasize the entire consumer side of the business. We still do a lot of consumer stuff, but we only do high-end, difficult-to-do, complex products where people are going to pay a lot of money for it and our customers get to sell it at high value, then we do it. Otherwise, we exited all of those and really left that to a ton of Asian manufacturers, knowing that we could never compete in that space, and refocused the company on high-value, difficult-to-do things, charge a premium for it, vertically integrated. And that kind of philosophy of bringing manufacturing technology to the forefront is what drove our portfolio strategy.

Yeah. So let's start with, first, manufacturing and industrial companies for the last 30 years. My hypothesis, and the data suggests this very clearly, is that outside of labor arbitrage, we truly haven't seen manufacturing productivity move. Right? Look at the last two decades. Look at manufacturing productivity. Take out labor arbitrage, it really hasn't moved the needle. And there are reasons why that is the case. One of the big reasons is that factories inherently are really built with many software systems that are put together that don't talk to each other very well. And it is a worse problem for midsize and small-size manufacturing. An average factory that is good will have 50 software systems. Most factories will have like 100 to 200 software systems. So we have been, you know, fallen for all these enterprise software systems that come our way. Put them all together. None of them talk to each other, don't work seamlessly, and we make suboptimal decisions all the time on every factory floor. And that world is going to change. And I'm super excited about that world changing because I've been pissed off about it for 20 years now. So why that world's going to change is because if you can get a data layer, forget your 50, 100, 200 software systems. If you can get a data layer that is clean, that is agile, then you can start making a lot of decisions really, really fast. And I'll give you some real examples to think about this. Average utilization in a factory, doesn't matter what people tell you. They'll tell you it's 80-90%. Those are all lies because they make it sound 80-90% by saying, 'Oh, I took out equipment downtime or I took out X, Y, and Z.' Usually 50% is a big number, a good number. Usually it's running at 30% utilization. The reason is because customers mix changes, then setup changes, or factory worker didn't show up, or machine downtime. The variables are many, and you put all those together and your equipment is sitting idle most of the time. So here comes AI. You can have planning systems that are taking all these complex variables and making smart decisions. If it moves your utilization up 10% or 20%, it's able to say, well, I'm going to predict that because Thanksgiving is coming along, I'm going to have 5% more absenteeism than I typically had, and your best use of equipment will be to run this set of products for this customer that can move your utilization up 10 points. That kind of intelligence is game-changing. Average factories don't do that. So I'm super excited about AI on the factory floor and what that's going to do for productivity and efficiency, because it's going to be a game-changer. We're not going to be falling prey to all these complex software systems anymore. And I know everyone's focused on automation and humanoids and all those cool things. All that's happening. Automation's happening, hardware automation is happening. All that is game-changing and it'll be great. But I'm super excited about AI's impact on software automation and what that does to the future of industrial companies. Very important for America, by the way, because we don't have the advantage of a low cost of capital. So we can take advantage of this game-changer for us.

Come to one of our factories, David. I'll show you how they work.

Yes, please show up.

First, I'll say I agree 100% with what Jensen is saying, and I have real-life examples for this already happening. I've spent 30 years of my career watching factories move out of America to all parts of the world, and today I'm seeing a true rebuilding of factories in America, and it is being done in a smart way and in a thoughtful way because you're bringing back manufacturing in hard-to-do, complex integration, complex assembly and test. So for example, if you have to put together a compute integration rack for what goes into data centers, you can't put all that together and ship it from across the ocean. It's very complex, and there are lots of changes that happen all the time. So it's best done close to home. So setting up factories that put together those integrations, all the way from cutting the metal to putting it together, testing it, that takes a ton of power, and then being able to pop it into a data center where it's ready to go is something that's happening today in America, and that wasn't happening a few years ago because the volume wasn't significant enough, the complexity wasn't so big. So the things you're seeing coming back home are the difficult-to-do assembly and test integration. The very kind of high-power tests that are happening—those are the kinds of jobs that you're really seeing coming back into the U.S. Just a month ago, we signed another new lease close to Austin and Georgetown for a factory that's going to be up and running in 12 months. We hope to have 50 megawatts of power there, and that factory is already sold out. And that is the kind of jobs that are coming in because they will do very complicated things.

Just what I talked about: compute integration. Things that go into a data center will be produced there. But it requires skilled technicians who know how to test this kind of equipment, who know how to build that kind of equipment. It'll require machining equipment to cut all the metal and put it together. So it'll be very high-skill jobs required to run that factory. Those are the kinds of reindustrialization we're seeing today in America, which is exciting for a person like me who's seen it go out and now come back all over again.

Well, I hope there is not a danger that we will overdo the reshoring, because we can't. And the reason we can't is, first, there's not enough people in the country to do things that are not absolutely of high value. My favorite product that I always talk about is the vision and need and want to build phones in America. If you think about a basic electronic board that goes into a lot of this equipment, like a printed circuit board, it's not something that you probably want to do in America. It has environmental issues. A lot of people can do it. Why would you try to do that unless you can bring high-quality automation to do it in a more efficient, more environmentally friendly way? Then yes, that's something we should think about doing, but that's not where it is today. So we have so many kind of good quality things to do that we should be able to pick and choose and get the best of what we want to do. So when I think about PCBs, capacitors, resistors—all of those easy-to-do things—we don't need that kind of labor to do that work. I think most industrial companies are smart. They know what kinds of things to focus on. I don't think, day-to-day politics aside, politicians may talk about bringing those kinds of jobs home, but I don't think any CEO will look at it and say, 'This is what I want to use my very high-quality labor in America to be doing.'

Yes.

No, no, I hope not. In fact, my kids thought that was the name of the company, unfortunately, like most others.

Yeah, so I told you Flex is one of the world's largest contract manufacturing companies. And when I came into Flex 7 years ago, the thing I was looking to do was build on certain technologies that I thought could drive higher value for Flex. So we exited a lot of portfolio. We spun off another company called Nextracker, which is a $17 billion valuation today. So we've done a lot of portfolio work, but within that portfolio work we started building technology that we thought would be important for the future. My history is in the energy space. I've spent a lot of my time in that. So one of the things we started looking at was building power for powering the chips, because we were already putting all the compute trays together, and we thought compute is going to get power hungry. This was well before the ChatGPT moment. And so that's what we started building—really powering that. And then at that time I thought, well, you're going to put all this power into this rack, and then how are you going to distribute all that power all the way to the data center and to the grid? So I started buying all these power companies to put together a compute cooling power portfolio. And then the ChatGPT moment happens. Everybody wakes up. They understand that data centers are a thing. Compute is getting hot, needs to be cooled and powered. And so suddenly we have this business that is taking a lot of capital, technology. It's a product company within a contract manufacturing company, and capital allocation priorities change. So it was important for us to say let's have two management teams focus on these two companies. SpinCo is going to grow at 70% this year, 80% next year, so it has different priorities, different capital allocation look. And then Flex has a lot of work to do because we want to continue to build the same thing we did in areas like healthcare and industrial products. So here you have two companies with a fantastic opportunity to go do big things. The timing made a lot of sense, and I am excited about the energy revolution that's going to happen in this country, and that's my history and heritage. So I've chosen to go run that.

Not at all. The reason I'll tell you why is the data is staggering, right? I mean, think about this: In 2030, a data center is going to take like 975 terawatt hours of power globally, and half of that is going to be in the U.S. The amount of electricity it's going to consume is just staggering. And just 5 years ago, David, we used to make these compute racks, the stuff that goes into data centers, and that used to be like 10 kilowatts. Today, it's soon going into 600 kilowatts. What does that mean to the layperson? A 600-kilowatt rack that just goes in a cabinet in your house could power a thousand homes. Thousands of those are going into a data center. When I think about what's happening with the capex investment there, I differentiate between training and inference. A lot of the deployment today is going into training, which is required. You have to build the LLM models. You have to build the kind of intelligence into it. But think about as inference takes off—those agentic AIs we're deploying in factories, in corporate America, in homes—all of those will be working at scale and in an agile way making decisions. So if you think about the compute story, it feels like it's barely started. Think about when inference takes off and how much that is going to be taking. So yes, hyperscalers are putting $700 billion of investment this year and probably have to repeat that a few years, but the data is staggering in terms of what is going to be required. And then the electricity story is a whole different one. Because that has just started. We don't know how to make the 600-kilowatt rack and deal with the thermodynamics of the heat and the cooling and all that happens. So now we have to fix the grid, fix all the power stuff, and that journey is going to take another decade plus of investment to make it real. So we have a long runway in this. I know it feels like we're investing a lot of money, but America is ahead because of our innovation, and we love our innovation because we're so good at it. Today, we're behind in infrastructure. If we can't get ahead on that infrastructure, we're going to fall behind, we're going to lose all that innovation that we just put so much effort into. So I feel this capital deployment has legs, and at least I'm going to see it in the near future, is my view.

I would say that there is some data that's promising here. What I'm hearing and seeing is a lot of Gen Zers—I think that's the generation—want to go into skilled jobs like apprentice jobs, skilled worker jobs. So if we can have public-private partnership to really train and retrain America, go into the middle of America—Shawnee, Oklahoma, Hutchinson, Kansas, Pittsburgh, Pennsylvania, the areas I lived in—there are a lot of people who want to get retrained to do these jobs. So we have to have an apprenticeship program that is deployed in a systematic way at a statewide level to really rebuild America for those jobs. Otherwise, we'll continue to have 300,000 jobs open. And then, being an immigrant sitting here in this chair, I will say that having an immigration policy that is cohesive, consistent, and long-term will also help that story.

Yes.

That's right.

I would tell young women that, you know, a lot of young women think engineering is a daunting space to be in. And, you know, I'm very public about this. I tell everybody I'm a bad engineer. The only reason I went to engineering school was I knew it would give me a paycheck and would give me a job. And look where I am today, right? So if I can do it, a lot of young women can get into engineering and STEM and those kinds of roles, because if you have strategic thinking, if you have common sense, if you have the ability to put a vision together, you can go places. And that's all I brought to the table. So I want every young woman to aspire to say that I can do this. And I did this while being a full-time mom with a husband who's got a full-time career. So I do think one can get it all. But I want women to aspire to know that she was not a great student, she was able to do this, and she got here. So there's no reason all of you can't think about that.

That's correct. Absolutely. That's me.