The cell-based ASIC business that we know today was born in the early 1980s and was pioneered by companies like LSI Logic and VLSI Technology. Some of this history is covered in Chapter 2 of our book, “Fabless: The Transformation of the Semiconductor Industry”. The ASIC business truly changed the world. Prior to this revolution, custom chips were only available to huge, integrated device manufacturers. These behemoth organizations housed massive design teams, mask-making equipment and wafer fabs. They did it all.
Once ASICs began to flourish, all of that changed. The custom chip market became democratized. Suddenly, anyone with a vision and a reasonable budget could build a custom chip. The result was the ubiquitous deployment of semiconductor technology for custom applications of all kinds. Products became smaller, smarter and more sophisticated. We continue to see this trend today. In spite of this dramatic impact, ASIC has become something of a boutique market. In the 1980s and 1990s many analysts tracked market size, growth, and weighed competing technologies to implement custom chips. Today, hardly anyone tracks this business in spite of its revolutionary impact on our world. I did find a Gartner reference that predicts the ASIC market will be about $27B by 2020 which I think is conservative. AI and other heavy duty applications running on specialized ASICs dominate general purpose silicon so be optimistic, absolutely.
So, where does all that ASIC money go? The answer to that question is definitely changing. There are a lot of specialty suppliers for various forms of analog, mixed signal or sensor-based designs. Rather than get lost in all that detail, let’s look at the top-end of the market. Who is doing the most advanced designs? For many years LSI Logic was king of that hill. So was IBM Microelectronics. There were also several strong Japanese suppliers (e.g., NEC). ST Microelectronics in Europe participated in the ASIC business as well. In Taiwan, Global Unichip and Faraday are still in the mix. In China it is Brite Semi and Verisilicon. Back in the US, Open-Silicon and eSilicon dominate. eSilicon pivoted to be a top-end supplier a few years ago, more on them later.
Fast-forwarding to today, things are different. The Japanese suppliers have all but disappeared with the exception of Socionext which is the combination of the LSI businesses from Fujitsu and Panasonic. ST Microelectronics is still on the playing field, albeit with a somewhat unclear focus. LSI Logic became part of Avago who then bought Broadcom. So, there is still ASIC inside Broadcom but it’s dwarfed by the rest of the business and in theory they compete with their ASIC customers. The once mighty IBM Microelectronics got swallowed up by GLOBALFOUNDRIES a few years ago and recently there was a new chapter to that story.
In a surprising announcement, GF announced it would put all 7nm and below technology development on hold. The company did have active development in this area and some advanced manufacturing in Malta, N.Y. With no technology roadmap at 7nm and below, GF is likely to become a boutique foundry. Certainly, a very large boutique foundry. We’ll see how their technology roadmap unfolds over time.
But the second part of their recent announcement is also headline news. GF will spin out its ASIC business as an independent, wholly-owned subsidiary. The new company will source designs above 7nm from GF and purchase wafers on the open market at 7nm and below. It will be interesting to see how a wholly-owned subsidiary of one foundry gets advanced technology (think PDKs) from another foundry. Or maybe the GF ASIC unit will be put up for sale?
We’ll have to see how this new ASIC supplier fits in the market. One thing is for sure, all these changes could spell significant opportunity for pure-play, focused top-end ASIC suppliers like eSilicon, absolutely.