We just concluded two very successful seminars in Tokyo and Shanghai. Samsung Memory presented their HBM2 solutions, Samsung Foundry talked about their advanced 14nm FinFET solutions, ASE Group reviewed their advanced 2.5D packaging solutions, eSilicon presented our ASIC and 2.5D design/implementation and IP solutions, Rambus detailed their high-performance SerDes solutions and Northwest Logic presented their HBM2 controller solutions. A complete ecosystem to address the requirements of advanced ASICs for markets such as high-performance computing, networking, deep learning and 5G. If you wish you could have been there, read on – there is another chance.
So, I feel like I have an unfair advantage to answer the question, “what’s next?” Based on the attendance at our events and the questions the attendees asked, I would say we’ll be seeing a lot more advanced ASICs in 2018 targeting high-performance computing, networking, deep learning and 5G infrastructure. These are all hot markets, so maybe this prediction isn’t that difficult to make. There is an important aspect of my prediction however – the rise of advanced application-specific integrated circuits (ASICs).
Many markets are initially enabled with semiconductor devices in the form of application-specific standard products (ASSPs) or field-programmable gate arrays (FPGAs). The recent rise of deep learning applications has created another ASSP-like option, the graphics processing unit (GPU). All of these technologies offer a predictable unit cost with a specific set of features. This usually works well until one of two things happens:
1) Volume and cost pressures grow to the point where buying a standard product where perhaps only 60 percent of the chip is used becomes a financial problem
2) Competition in the market becomes so strong that differentiating with the same chip everyone else uses becomes a marketing problem
When one or both of the above situations occur, folks usually turn to ASICs as a way to reduce unit cost (the chip does exactly what’s needed – no wasted silicon) or as a way to increase differentiation (you can make the chip do anything you want). What’s interesting about markets like high-performance computing, networking, deep learning and 5G infrastructure is that all are already high growth, highly competitive markets. This means ASICs become an option sooner rather than later. Add to that the fact that the design cost of an ASIC (which is a custom chip built for just one customer) is typically a very small part of the unit cost of the end system. This also makes ASICs a good option sooner rather than later.
So, a robust ASIC business is likely for these markets. Being part of an advanced ASIC company, I may be biased. I don’t think so however. I will watch the deep learning market with particular interest. Efficiency and power optimization are both very important for these applications, and ASICs are an excellent way to attack both of those problems.
By the way, if you missed our live seminars, I can make another prediction. You will get to learn all about advanced ASICs on January 17, 2018 at one of two webinars, either 8AM or 6PM Pacific time. Find out more about this great second chance to hear about a complete 2.5D/HBM2 ecosystem.