At my first real tech job in 1997, I had a role at Cypress Semiconductor working with engineers to draw and model designs of the products. It was during this time I developed a love of the fascinating world of semiconductors. The science behind them is remarkable. The kind of stuff where, when it’s explained to you, there is a feeling as if there is no way that should be possible, yet it is. The semiconductor industry is also very small in reality. There are only a handful of companies who have the talent to push the limits of silicon process and architecture design.
There has been a debate in the last five years about whether the silicon era is over. One of Broadcom’s founders, Henri Samueli, was fond of stating during keynotes and other speaking engagements that “We will not see any new silicon startups anymore”. Basically, his tagline was the semiconductor golden era had ended. Over the years, I have had many conversations with Henri on this topic. His logic and reasoning was no doubt sound but it seems his views discounted things like virtual/augmented reality and artificial intelligence as drivers of computing power that still does not exist today.
In context, Henri’s comments were more directed at core computing startups — components like CPUs and GPUs, or new computer architectures — and less about sensor or companion processor type startups. He also was looking at how the big silicon companies were becoming so dominant and how much capital it seemed necessary a silicon startup would need to compete. There remains truth to much of his logic today but we are seeing a fascinating renaissance in the semiconductor industry.
It turns out, the future has not yet arrived. Things like VR/AR/AI and other common buzzwords of today are only shadows of what they will be in the future. We simply do not have the computing power and underlying computing technology to bring this vision to market. While all the heavy hitters in semi — Nvidia, Intel, AMD, and Qualcomm — are working on solving these big hairy problems, we are actually seeing some new startups use combinations of licensable and customizable ARM IP, and/or lab and research institution techniques along with other kinds of technology, allowing them to create solutions that attack the gaps the big silicon companies are ignoring currently. We have companies successfully raising money to accelerate network machine learning like Wave Computing. Silicon Catalyst is a local accelerator here in the Valley that provides funding as well as other assistance helping new semiconductor companies grow and compete. I’ve even done a number of due diligence work with Silicon Valley firms who have put capital into new, mostly still stealth, semiconductor startups.
All of this is driven by the opportunities and challenges which have recently come to light from various industries I have mentioned above. Many were quickly reminded computing power is not yet commoditized once AR/VR/AI and more hit the scene. More importantly, they realize semiconductor roadmaps are guided by static scientific principles and thus, we have a pretty good idea where we will be in five years. We know that, even then, the silicon we have will not be powerful enough to get us where we want to go.
Few things visualize the upside opportunity more than Nvidia’s stock price over the last few years.
It was certainly true the past decade seemed to suggest we had enough computing power and the need for dramatically more did not seem necessary. However, innovation found a way and we now find ourselves in a situation where the future can not come fast enough. Yet, we still have to solve extremely hard problems that will be lead by semiconductor companies.
I’m a big believer in and fond of saying that there is no single greater predictor of the future than understanding semiconductor company roadmaps and what those roadmaps enable in terms of computing experiences. Given where we were a few years ago, I honestly didn’t think doing semiconductor industry analysis would be as “cool” as it was in the early 2000’s but now, the interest is there again and as big as ever. We have things like AR/VR machine learning and artificial intelligence to thank for it.
While we likely won’t see a boom like we did in the post-Fairchild Semiconductor era, it is encouraging to see investors and founders not afraid to take on the big boys and innovate.
Really well put, thank you. I do have some comments (of course).
“It was certainly true the past decade seemed to suggest we had enough computing power and the need for dramatically more did not seem necessary.”
In the consumer market, this is true and will remain true until consumers become more demanding in their applications. Right now, I’m told an SD card is too complicated. Outside the consumer market, you’re never too fast.
“While we likely won’t see a boom like we did in the post-Fairchild Semiconductor era, it is encouraging to see investors and founders not afraid to take on the big boys and innovate.”
It is encouraging, but it’s also most likely true that the real product is the company itself. Exit strategy! In other words, it exists not to compete, but to be acquired, which is fine.
very satisfying in terms of information thank you very much.
This was beautiful Admin. Thank you for your reflections.