Issue link: https://iconnect007.uberflip.com/i/1506834
66 SMT007 MAGAZINE I SEPTEMBER 2023 e imec roadmap for solder- based FLI with pre-applied under- fill shows that most flip chip sol- der interconnects will reach the lower limit at around 7, possibly 5-micron pitch. Everything aer will be copper-copper hybrid bonding. How will advanced packaging change the life of the typical EMS company? Let's focus on automotive, where the role of the former Tier 1s (direct EMS suppliers to the automotive OEMs) is changing dramati- cally as OEMs move to cozier direct relations with the semiconductor suppliers. e nature of the supply chain is therefore changing dra- matically. From a reliability aspect, if you're talking about electronics assemblies and systems that, say, you're putting into a car, it will always be difficult to have electronics that are large and monolithic. ere are more thermal and exter- nal physical stresses than there are in a stan- dard dielectric liquid-cooled server, so it's dif- ficult to see how to reliably integrate very large heterogeneously integrated modules into a car without some form of mechanical decoupling. at means there will almost always be a need for distributed computing within the vehicle. Rather than vehicle compute and control being a completely centralized thing, with each sensor running to a central brain, as a purely "soware defined vehicle," it's much more likely that we will see a central process- ing unit receiving inputs from "edge com- pute" (or maybe more accurately vehicle "cor- ner compute") sensor clusters integrating lidar and radar and the other kinds of sensor inputs. ese clusters will probably be located on the corners of the car using machine learning algo- rithms to parse sensor inputs into a data for- mat where the central processor assembly can make decisions and drive actuation. Of course, in a car vs. a drone, it's relatively simple: You're talking about merely controlling direction and the speed that you're moving at, and there really isn't much more needed than that. If your application is more data center-oriented, you're not so concerned about environmen- tal stresses on your design. That takes you in a certain direction with more room for the use of very large advanced packages. Is that over the horizon for EMS companies—a need to choose a path and specialize? Yes, absolutely. However, not everyone needs a Class1 cleanroom and a hybrid bonder; flip chip and wire bonding aren't going away. Just as there will always be some room for wave soldering, you must be in a market with an appreciable profit margin. For example, most wire bonding is being done for the consumer, industrial, and automotive markets, simply because of its known good reliability. Custom- ers will need flexibility and technology that's well understood in terms of failure modes and reliability. With regard to the challenges in the U.S., EMS providers are way behind in their ability to supply locally because of offshoring. Only something like 2–3% of the packaging industry is based in the United States, with some com- panies planning to bring advanced packaging back (or more appropriately to develop new abilities, to compete with Asia and Europe) to the U.S. ese players are looking to develop the capabilities to manufacture cutting edge assemblies, to be able to do difficult hetero- geneous integration, and to actually manufac- ture the very high-end substrates that will be needed, in increasingly cleanroom environ- ments. To compete with the low-cost areas of the world, we'll be using more automation. By automating processes, we can reduce man- power costs and make it economically feasi- ble because, at the end of the day, we can pour Dr. Andy Mackie