On The Verge

Anyone who has been following the IoT/IoE or whatever-you-want-to-call-it movement knows we’re on the eve of far-reaching, life-altering change. There will be billions of connected devices, all streaming information to gigascale cloud datacenters using big data analytics and deep machine learning.

Somewhere along the way, we’ll discover important, useful information from all this that will have a positive impact on humanity. It will allow us to improve quality of life, optimize all kinds of business ventures and extend life itself. These are thought-provoking, important topics to consider.  For now, we’ll just have to get by with gadgets that monitor our exercise patterns and take joy in controlling all kinds of things from our smart phones.

Since I live in the world of custom chips, ASICs if you prefer, there is another relevant and important trend that I’ve been watching and pondering for some time now. All those billions of connected devices will need semiconductor technology to achieve their connectedness.  This is for certain, and this implies a lot of some kind of chip or chips will be sold.  That’s great news for somebody, but who?

If you look at most of the IoT products on the market today, the vast majority of them use off-the-shelf parts. Standard chips are easier to access at lower cost than custom chips. If you can find one that comes close to what you need, you’re off and running. This is how most markets start. At some point, custom chips, or ASICS, become important typically for one of two reasons.

• My volumes are going up, and so are my price pressures. I’m actually only using a fraction of the standard part’s functionality that I buy. The rest “goes to waste.” I can’t afford to do this anymore. The cost of an ASIC is justified – If I can make a chip that does EXACTLY what I need and nothing else, I’ll win.
• All my competitors use the same chip as me. I can’t differentiate. If I can build an ASIC, I can put functionality in hardware that my competitors won’t have. I’ll win.

And so it goes in many markets. The need for custom silicon arises, and ASIC companies reap the benefits. With all those billions of connected devices, there has to be a tidal wave of ASIC demand coming, right? The killer aps will emerge and either point 1 or 2, above, will kick in. I believe this to be true, with a twist.

The customers from IoT who need ASICs will be different than past customers. They will be system designers, not chip designers.  They will have a strong grasp on the needs of their market, and they will need ASICs to remain competitive. But, they won’t really know how to design a chip. They will all need chips built with trailing-edge technology nodes (modest processing power with lots of sensors implies that). They will need ultra-low-power designs (for battery life). And they will need those chips FAST (competitive pressures abound).

I find this all very exciting.  It means that if ASIC suppliers can deliver moderate-complexity chips faster, easier and more reliably than today’s standard, a huge market is waiting. Solving these problems is quite doable, especially when you consider the power of internet-based solution delivery. This is the key to unlocking this market. So watch these trends over the next year or two. There will be important innovations.

If you look at semiconductor business cycles, we’ve been through a few. It started with an elite market owned by the IDMs – those rich enough to own everything.  This was followed by the high-growth, democratized era of ASICs – chips for the common person. Due to the extreme cost of designing and manufacturing an advanced SoC, we are once again in an elite market. I believe IoT ASIC demand will usher in another democratized era. The only question is when…