Avnet regularly surveys engineers to learn what they’re thinking. That sort of information is quite important to Alex Iuorio, vice president of supplier development at Avnet. In this interview, Alex talks about what he’s learned from the most recent survey and its implications to the supply market in 2026 and beyond. No surprise, AI plays a remarkably large role in all the current trends. View the complete survey report here.
Nolan Johnson: Alex, would you give me a description of the survey, why you did it, and what you were trying to accomplish?
Alex Iuorio: We’ve been doing the Avnet Insights Survey for several years, but over the past two years, we've focused on artificial intelligence (AI) in a variety of contexts. We surveyed 1,200 engineers, and then interpreted or interpolated what they were telling us about AI. Like everyone else, we were trying to discern the market.
We thought about the second phase of AI adoption: the utilization or embedding of AI technologies in phones and PCs. We've all talked about whether it will accelerate the PC refresh cycle, or even the phone refresh cycle.
The third phase is arguably concurrent with the second: AI at the edge. That's what really excites us about our customer base. About 10 years ago, a leader at Intel, a friend of mine, coined the phrase, “rise of the edge.” What she rightly understood was that all those edge device manufacturers would want to be connected to the net. At the time, I thought, “That's all my customers at some point. Who will build something that they don't want connected?”
Of course, that's more or less come to pass. So, through the survey, we've tried to discern what the customer base will need from their AI tools. From the survey, we learned that it's important for us to discern how AI is being used to facilitate the design and manufacturing processes.
We wanted to determine how respondents use AI internally and externally. There was a 33% increase—from 42% to 56%—over last year’s response regarding AI’s external usage. That's a fantastic increase, but to be perfectly honest, it wouldn't surprise me if it had come back at 70 or 80%, because people need to adapt to be competitive, while using those technologies to compress manufacturing cycle time and improve their time-to-money metrics.
Johnson: What sort of functional role does your typical respondent have?
Iuorio: They are computer scientists, design engineers, manufacturing engineers, and software engineers. We try to get a good cross-section with the 1,200 responses.
Johnson: With respect to your respondents, how are they feeling about the market and the current business environment?
Iuorio: If there's one thing that intrigued us, it would be that morale is super high: 77% of the respondents believe the next year will be really tough, probably governed in the short-term by geopolitical issues. But there's this sense that something better is coming next.
I've never seen anything like it in my life. We conduct a quarterly market analysis, essentially creating a proxy for the market that we serve. So, 90 days ago, my served market was forecast to close 2025 with 5.8% growth. Today, that same market looks more likely to come in at 10%.
Johnson: That's a massive step function.
Iuorio: Here’s what it's done to CAGRs, which should smooth out the curve. When I look at the three-year CAGR for semiconductors, knowing that CAGR should, by definition, smooth out the curve, normally you see three-year CAGR moves of, say, 0.2 or 0.3. Ninety days ago, the CAGR was 4.3; now it's at 7.1. Those forecasted dollars are spread across that three-year period; they continue to accelerate.
The overall morale is high, although the short-term morale is a little guarded. Now, there could be some anomaly that drives long-term forecasts down, but assuming that doesn't happen, we're at least locked in for a couple of years.
I was at the keynote at CES, and what really struck me was AI’s presence. A year ago, we were talking about budgets rivaling the U.S. defense budget: $380 to $420 billion. The presenter showed data on that same spend projected out to 2030, and it ramps up to $2.3 trillion. It was just staggering.
Johnson: With the forecasted enthusiastic market, how are engineers implementing AI? What are your survey respondents doing?
Iuorio: We're focused on that “AI at the edge” piece. I believe 100% of engineers are scrambling to determine how to add AI functionality to their equipment, to whatever degree that might be.
Our solution is to support AI at the edge and take advantage of this curve, which we believe is accelerating, with a product called IoT Connect™. We took that platform, and we’re characterizing microcontroller devices for our top five manufacturers. We characterize them for operability and do so securely. We can offer free access to a sandbox where you can work with the devices, develop a network, and virtualize it.
For each of these vendors, we have characterized their devices. You buy the device, we give you access, and you can do something right out of the box. We are creating solutions that accelerate that development, thereby accelerating their time to market.
Johnson: This is a situation where, to exercise that device, whether it’s in hand or it’s a model in your CAD tool, is crucial. Will it meet my specifications, and is the timing right? In a traditional breadboard method, we make a prototype, get the hardware, and stick it down, only to find out it doesn't work, which extends your design cycles. We can't afford that anymore.
Iuorio: That's what we're trying to facilitate. You're in that development sandbox environment, and you're generating data as you do it. So, one of those five suppliers comes in and says, “You know what? We got this thing that houses all our algorithms and models. We'll collect data from the devices, put it in our tool, and allow the engineer to populate the devices with a validated model based on their own data.”
About AI-based tools, Avnet’s CEO once said, “This will be important when CEOs from our key supplier partners talk to me about it in our meetings.” Now, you can’t get CEOs to stop talking about this in meetings. They have other things to do, but everyone wants to maximize this. AI is a platform, but the specificity of component AI capability is what drives our companies.
Johnson: I was intrigued by the part of the survey about large language models like ChatGPT. The engineers responded that they don’t want to use a generalized model, but rather a specialized, curated dataset, trained by engineers. Can you explain that further?
Iuorio: It intrigued us, too. They’re trying to tell us that everyone wants the specificity of an engineering-developed model; that's key to what you and I need. But they know that AI usefulness, even in initial stages, is all about maintenance and updates. What users can be sure of in the public markets is that the data sets are updated. Maintaining the data is a challenge and a community makes a difference here. You don't get that benefit in a company network. They'd love to use such a model, but they don't believe it will ever be updated.
Johnson: This smells like a market opportunity for a company like Avnet. Is this a direction you're headed?
Iuorio: We're talking about it. The survey screams that there's a need, but it becomes a dollars-and-cents situation. It’s an intriguing opportunity.
Johnson: What are some of the challenges your respondents in the survey shared about incorporating AI?
Iuorio: The leader by far was data quality, which we just talked about. If you go in rank order, integration and existing tools are a huge factor—that’s the motivator for creating systems like IoT Connect™. The rest of the list includes increased performance, lack of skilled personnel, and product power requirements.
Johnson: Just about any bleeding-edge technology struggles with many of these same issues. Sustainability showed up in the survey, and I'm curious about the context relative to AI?
Iuorio: They're talking about power and the grid, and what's associated with it; it’s just so much play. For example, for a new home start, the magic number for electricity usage is one kilowatt per house. Over the past five years, new home construction starts in the U.S. have averaged 1.5 million annually. We already have spotty power in some places as the grid gets stressed. The punchline is that data centers will add the equivalent of 18 million homes on the grid over the next 24 months. How can we sustain it?
Johnson: Where do you see AI and Avnet converging in the next couple of years?
Iuorio: We probably have a dozen PLCs in place to optimize our business process from start to finish. I'm very proud of our process for identifying an opportunity, scoping the project, and specifying the kind of return we expect.
You'll see more smallish solutions addressing very specific problems, and they'll continue to build such that, at some point, there will be a lot of touchless supply chain management between Avnet and the market at large. That will be the biggest change.
I own the global migration PLC. I have a team that matches up registration activities in our systems all over the world. They scrub the data to ensure a customer’s activity is collected in one place. Using these new AI tools, the proof-of-concept for that process reduced a two-week task to one minute. Now they can spend that time on the problems. Out of the 500 line items that day, there were maybe two that needed intervention; now they can focus on those issues instead.
Obviously, the implications to efficiency are clear and that's where I see us going. You'll start to see us create efficiency in how we position, how we market, and how we transact.
Johnson: Alex, this has been very insightful. Thank you.
Iuorio: Thank you, Nolan.
