The shape of the MCU market
IC Insights has a new report about the size of the microcontroller market.
They project over 31 billion MCUs will be sold this year, growing to almost 50 billion in 2019. That’s individual chips, not cores. These numbers dwarf PC-style microprocessor sales. They attribute most of this growth to increases in sales of smartcards, like those found in credit cards with chips, ID tags, etc. Check out this graph:
Note: ASP means average sell price.
Picture courtesy of IC Insights, The 2015 McClean Report
It’s fascinating that though shipments will grow with wild abandon, revenues won’t. ASPs are in free-fall and will tumble in half in just a couple of years.
One wonders why a company would want to be in this business! A vast increase in deliveries yet almost no gain in revenue? Digi-key lists almost 60,000 distinct part numbers for MCUs. While many of those parts are just packaging and other minor variants, the big semiconductor companies do have enormous numbers of parts that differ substantially. The Cortex-M series is supported by pretty much everyone, yet differentiation is based on memory and I/O, not on the processor. With prices in the dumper, how many of these vendors will chose to continue selling their parts?
I expect a shakeout.
The tragedy is that if a company end-of-lifes their MCUs, given how complex the peripherals are today, a lot of engineers are going to have to do major product redesigns.
I don’t know what processors are in smartcards, but given that these are pretty low-end devices you’d find a lot of 8 bitters being used. Microchip is an example of a company doing a blistering amount of business in 8 and 16 bits – they are about a $2 billion dollar company today. The company recently announced plans to buy Atmel for about $3.5 billion, which will add about a billion in sales. It’s hard to see how Microchip can integrate Atmel’s processors into their product line (and who wants 8051s anymore?), but Atmel has a rich line of other kinds of semiconductors that could open different markets for their new owner.
The first commercially successful microprocessor, Intel’s 4004, cost about $1100 in today’s dollars. That was about half a buck per transistor. Today we get a complete computer-on-a-chip with all kinds of memory and I/O for that same half buck. Astonishing.
PC sales – and corresponding PC-style processors – are at the lowest level since 2007 with only 75 million sold last quarter. Extrapolated to a full year, PC processors are now just 1% of MCU unit sales. Traditionally that number has been twice that.
The key takeaway is that the micro business really is about embedded systems. There are more dollars in selling high-end chips for PCs, but the unit shipments pale compared to MCUs. And this is good news for embedded engineers. More chips means more products that need to be designed, and falling ASPs means, as it always has in the semiconductor business, the continued explosion of demand.
Jack G. Ganssle is a lecturer and consultant on embedded development issues. He conducts seminars on embedded systems and helps companies with their embedded challenges, and works as an expert witness on embedded issues. Contact him at firstname.lastname@example.org. His website is www.ganssle.com.