What Heterogeneous Integration Means for EMS Providers

Estimated reading time: 4 minutes

Dr. Ravi Mahajan, an Intel Fellow and Director of Intel’s Technology and Pathfinding group, delivered a keynote at the APEX EXPO 2026 technical conference on using heterogeneous integration (HI) as a strategy and on how advanced packaging technology serves as the technical apex for implementing that strategy. Mahajan’s previous papers and industry presentations on such topics as interconnect density, signal integrity, power delivery, thermal path, and assembly yield as system-level constraints confirm him as an expert on package optimization.

Mahajan’s invitation to speak reflects the Global Electronic Association’s growing role in convening semiconductor, EMS and substrate leaders  to collaboratively develop the roadmaps, standards, and manufacturing approaches required to scale next generation HPC and AI systems. In this interview, directly after his keynote, he explains how HI is reshaping system design, why advanced packaging sits at the center of that shift, and why the industry must take a more proactive role in building the roadmap ahead.

Nolan Johnson: Ravi, in your keynote, you spoke about heterogeneous integration and advanced electronic packaging. Why is heterogeneous integration so important?

Ravi Mahajan: Computing and communications demands move fast; over the past decade or so, even faster than normal. The question for all of us is how to stitch different IPs together to respond effectively to demand. We know that not all technologies can be built on the same wafer platform: Different technologies are better optimized for different wafer platforms. Therefore, if you can't build and optimize all IPs on a single piece of silicon, the next best approach is to optimize the package. By that, I mean stitching functions together, so they work better collectively than the individual components do. The package is the ideal platform for heterogeneous integration. This idea has been around for a very long time, by the way; it was in Gordon Moore's original paper.

Johnson: I remember the term then was “multichip modules.”

Mahajan: Yes, exactly. The demand is growing steadily, and we are figuring out the best way to evolve. HI is a very good way to improve performance, as long as it's economically feasible.

Johnson: You’ve said that advanced packaging makes heterogeneous integration possible. Could you provide more details?

Mahajan: Advanced packaging is the leading edge of a technology wavefront. There are many levels of heterogeneous integration, not all of which require advanced packaging. But as we scale technology and drive advanced packaging features, the technology moves forward. Elements of advanced packaging technology trickle through to different segments. Advanced packaging is the tip of the HI spear, driving progress. HI enables people to integrate  bits and pieces, not always the latest, leading-edge technologies. We keep pushing the technology, and we always end up in a better place.

Johnson: Not to be contrary, but doesn’t the amount of computing power you put in that package drive the package? And when the modules start requiring tens of thousands of connections, that drives the package also?

Mahajan: Oh, it absolutely does. In fact, they build on each other. It's somewhat symbiotic. But that said, if technology gets ahead of the product, products can always make use of it. If product requirements outpace technology, it becomes difficult to develop technology. We’re always behind the eight ball in scaling, which is why having a technology roadmap makes sense. If we scale to what the future looks like, if we build systems to make scaling manufacturable, reliable, and performant, then we're in a better space.

Johnson: Some EMS processes already work with very large components. You mentioned, as did Matt Kelly during the keynote Q&A, that future parts will measure 200 millimeters by 100 millimeters. That’s huge, maybe the size of a typical cellphone. How do EMS companies anticipate these challenges?

Mahajan: Let's turn this around and ask: If this is a vector along which things are evolving, and we know that it is a potential future, wouldn't knowing about it in advance help solve the problem, as opposed to discovering it really late in the game? In other words, you write a technology  roadmap, and then we collectively drive the roadmap's trajectory to meet product requirements and manufacturing readiness.

Look at how ITRS (International Technology Roadmap for Semiconductors) succeeded. They wrote a trajectory and said, “This part is easy. This part is not so easy, this is difficult, and this is a showstopper.” You do the same thing here and ask, “What are my real challenges?” Then you distill them down and ask yourself, “Is it really a showstopper problem, or is it just a very difficult problem that we can collectively work on?” In doing so, we are less a victim of circumstance and more in control of our destiny.

Johnson: Would you recommend the HI roadmap as a resource document for planning? Is this a call to action for EMS companies to participate?

Mahajan: Absolutely. Not only should we think of it as a strategic document that sets the trajectory for development and research, but we should also participate in writing it by contributing manufacturing, reliability, test, and cost constraints that shape feasible HI architectures. We should participate in shaping the technology roadmap's direction. We want more and more people to get involved. I'll give you a classic, relevant example.

Devan Iyer contacted me and asked how he could help. I said, “Come and write portions of the roadmap.” He invited me into the ADEPT council to evangelize the technology roadmap. I did that; I evangelized it, and more people from industry joined in. They read the chapters. They said, “This is good. This is not so good. This is what we ought to do.” They contributed. I alluded to it in one of my slides and to the HIR chapter that the Global Electronics Association was invited to write on advanced electronic packaging at the system level. What we published this year is much better than what we published last year because it included PCB technology and PCBA and reliability insights into roadmaps in conjunction with package interconnect roadmaps, which forced people to think beyond the immediate to tomorrow, next year, and the year after that. What happens five years from now? Is it truly scalable, and how do I make it scalable?

We’re starting to open the envelope. It helps you think when you realize you really don't have an answer to a set of questions; you start to realize which research needs to be done. Once you write down the research needs, you need to set targets and fund that research.

Johnson: Ravi, thanks so much. A thought-provoking keynote, to be sure.

Mahajan: You’re welcome.