EUV: Worth the Wait

In a second post from president and chief analyst of TECHnalysis Research, LLC., Bob O’Donnell offers his view that EUV has definitely been worth the wait. Technically speaking, EUV allows more sophisticated chip designs to be made with less layers. In real-world terms, this means that the semiconductor manufacturing industry is overcoming what many perceived to be significant roadblocks in the chip building process. Read on to see what Bob has to say. You can also follow him on Twitter @bobodtech. All good things come to those who wait. We’ll sell no wine before it’s time. Patience is a virtue. We’ve all heard phrases which essentially say that some things just can’t be rushed. In the technology world, however, people have become accustomed to fast innovations and those that take time to arrive are often forgotten or written off as a dead end. For many years, such was the case with extreme ultra-violet lithography, or EUV for short. EUV is a technology that promised to enable the creation of extremely small transistors for semiconductor chips by using a light source with an extremely short wavelength to define sophisticated patterns onto silicon wafers. The basic techniques and science behind the technology were discovered in the 1980s and many companies led people to believe it would be ready for commercialization in the early 2000s. Unfortunately, mass commercialization has taken significantly longer because of the technical challenges involved in creating the extreme ultraviolet “light source” that’s at the heart of how the technology operates. Despite setbacks, however, many companies continued working on the technology, confident that it would one day be commercially available. Plus, many were driven by increasing signs that EUV was becoming critical for the ongoing evolution of the semiconductor business. In fact, because some of the impressive innovations that enabled the creation of semiconductor process nodes at 14, 10 and even 7nm started to slow down, it started to become apparent that EUV was going to be necessary to achieve 7nm and smaller manufacturing process geometries. In that context, Samsung Foundry started commercial production of chips using 7nm LPP EUV technology in October of 2018 and has now announced a roadmap that will allow the company to create 5 and then 4nm process nodes with FinFET transistors and then move down to 3nm with Gate All Around, or GAA-based transistors. The early iterations of EUV production on 7nm are used on a significant portion of multi-patterning layers, but later versions at smaller process nodes are expected to use more EUV layers. Technically speaking, EUV allows more sophisticated chip designs to be made with less layers because the precision of the EUV technology enables more details to be integrated into each layer. The bottom-line result on the forthcoming 5nm EUV process translates into a 25% increase in logic area efficiency along with either a 20% reduction in power consumption, or a 10% increase in performance (or some combination of both) versus the existing 7nm LPP process technology Samsung Foundry currently offers. In real-world terms, what this means is that the semiconductor manufacturing industry is overcoming what many perceived to be significant roadblocks in the chip building process. Instead of being stopped by the laws of physics—as some had feared—clever engineers have figured out ways to continue building smaller, faster and more power efficient chips. This, in turn, will enable the creation not only of next-generation smartphones and other artificial intelligent devices, but an entire class of completely different devices, many of which haven’t even been conceived of just yet. In particular, the world of artificial intelligence is proving to have nearly insatiable demands for computing performance and novel chip designs that can only be achieved with this next generation of silicon process technologies enabled by EUV and GAA. Right now, several leading semiconductor companies, as well as other technology companies not typically thought of as chip makers, are in the process of designing new types of accelerators optimized for machine learning, neural networks and other types of AI applications. Many of these new designs are going to need the performance capabilities and low power requirements that EUV-based semiconductor manufacturing can enable. On top of that, performance demands among traditional components, such as CPUs, are also growing. In some applications, CPUs are being used to perform calculations required for these AI-focused algorithms, while in others, huge amounts of raw data generated from sensors and other sources needs to be processed quickly in order to drive real-world decisions in manufacturing plants and other environments. Regardless of the application, it’s clear that computing demands across a variety of different chip architectures continue to increase. While there’s no question that EUV took significantly longer to achieve mass commercialization than almost anyone predicted, it’s also now true that the technology has finally arrived. Plus, it’s being proven in real-world process technology applications as we speak. Like many things in life that require time to reach their peak potential, EUV is ready for its moment to shine.