A dozen companies make CPUs, how many make ultra-precision sensors?

One hallmark of shifts in the financial markets is that old business models suddenly stop working. In 2008, that was mortgages. In 2023, once hot FinTech (and apparently most commercial banking) now looks largely untenable. Could something similar be happening in semiconductors? The “easy money” of Moore’s Law is giving way to much higher “interest rates”. This is bringing about a big shift towards custom semis, but it could also overturn many other basic assumptions of the industry.

We have recently come across an interesting potential case for this – in the sensor market of all places. We have been doing a lot of work on sensors this year, in large part because we are seeing some really powerful advances in this sector. There are a variety of reasons for this, but one of the most interesting is that new sensors companies are adopting new business models, models previously unavailable to “commodity” sensor makers. For as long as we can remember, sensors have been an afterthought in most of the electronics industry – messy, low-margin, step children of the flashier market for digital semis. We exaggerate, there have always been high margin specialty sensor segments, but overall much of this sector was overlooked by many.

This had important implications for how we built systems. Sensors, and all analog semis in many cases, got pushed to the edges. The digital components ended up with the strategic high ground. Look at the margins of the image sensors for phones, microphones, keyboards, RF systems and all the rest – for most of this century those have been brutal businesses. In most electronics categories, especially consumer electronics, the critical chips tended to be the core digital processors – CPUs in PCs, applications processors and modems in phones, etc. The companies that made those digital chips gradually subsumed other systems on their respective boards, with sensors left as after-thoughts, unable to command significant margins.

We recently heard a presentation from research firm, Yole, looking at the market for Automated Driving Assistance Systems (ADAS). This is an important market, but one which remains highly unsettled in terms of long term software and hardware architectures. The analysts argued that the automotive silicon stack would require some form of centralized automotive processor, which we agree with. Further, they argued that this chip would have to play the role of sensor fusion, combining the input of multiple sensor inputs like Lidar, Radar, Sonar, Smell, etc. Which, again, we agree with. But they then concluded the choice of that processor would in turn drive choices for all the sensors, leaving this processor maker with the strategic high ground, able to capture most of the system’s value. Ten years ago we would have agreed with this part as well, but this time we are not so certain.

For one thing, the automakers show every intention of holding onto to the driver’s seat (pun intended) when it comes to semis content in cars. If they can achieve this (by no means guaranteed) they are not going to let the processor vendor make all the choices when it comes to sensors. The sensors are also much more critical in automotive, meaning that there are likely going to be regulatory and possibly standards based considerations at play. There are also starting to be many sensor companies with significantly differentiated capabilities that they may be able to command higher margins. And finally, today’s sensor companies can see all the other changes in the world and are building software, services and solutions capabilities – which are much stickier business models. Put simply, there are some companies out there whose products are sufficiently differentiated that they can earn a stake in the ground of that strategic high ground. There are now over a dozen companies making CPUs, but how many companies have ultra-precise, real-time positional and voltage sensors? We can think of just one.

Nothing is certain. The heavy hand of history still rests pretty strongly in favor of old patterns dominating, but so much is changing in the industry, that it is possible that some important changes are coming to this market, soon.

Photo credit: Dire Straits and Hipwallpaper

Leave a Reply