You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
TSMC had a licensable mcu platform for 40nm since 2012, i think there are others.
Terepac/terechip solves the requirement for minimal die area due to IO .
So it's possible for quite some time to build microcontrollers for extremely cheap - sub 10 cents(probably 5 ) for a tested die - (and i guess terepac's price is low , sine their tech is disruptive).
And with everything available - IP , various RTOS's etc, the software problem could be largely handeld, for market entry at least.
So why aren't we seeing the market flooded with 40nm mcu's at extremely low prices as competition would dictate ?
But scanning through the list , looking for "extremely cheap" mcu's , the only low-end mcu's are offer by infineon(the rest look like competing on features, not price) and just looking at part search engines , they are priced comparatively to other similar chips. Same goes for the more expensive parts from infineon.
But maybe they compete on price on large volume buyers, behind closed doors, with 40nm as a signaling mechanism. But still , the fact that only one company announced low-end 40nm's raises curiosity.
1. 40nm isn't that expensive - you could probably justify development by selling $100 million per year at a decent margin?
2. the 32-bit micro-controller market is $11 billion in 2018. say you sell $100million/year at 1/2 of current cost, that's $200million/year(in today's goods) - that's 4% of the market .surely 4% of this market will be interested in some savings.
But yes, maybe $100 million isn't good enough for VC's - and once you aim at bigger market shares you'll see tough competition, and they can use same tricks - so the small guys get scared by the big guys - and the big guys are careful not to start a race to the bottom.
So basically as in the rest of the economy - manufacturing prices aren't that relevant - what determines price is competition.
Starting a price race to the bottom only works if your costs are lower than your competitors. If everyone has the same development and support costs, this is a race that you can't win. Bigger companies usually have lower costs because they're spread over bigger total volume or more devices. So a small disruptive supplier can't win unless they're cheaper -- but their materials (silicon and packaging) come from the same sources as everyone else, so they either have to have poorer support (fewer people) or lower wages (cheaper people). Only the second one delivers what customers want.
So such a startup would have to be based somewhere with low labour costs -- for certain, not the USA or Europe. But then many of the big players you listed already have large development activities in places like India and China for exactly this reason, so it's difficult to win on this alone. And any new company would have to invest in setting up costs which existing companies have already paid off, and then somehow undercut them on price while not having significantly lower costs.
Altough i'm not sure about your analysis of chinese companies not attacking the mcu market - mediatek recently released an mcu , the esp8266 is another close example.
One explanation of this is that while bigger companies have many benefits, they are also highly interested in keeping prices high, due to many reasons , one of them is that decreasing prices would probably mean decreasing share prices.
On the other hand, new entrants don't really care - so they might have some time to operate freely without interruption - and if within that time they could build some sort of a competitive barrier it could work well for them over the long term.
The question is , what long term barriers could a new company create ?
IanD, also let's not forget security and political power - the IOT raises this questions with regards to many mcu based products, so i won't wonder if the chinese government would want chinese designed mcu's deployed widely.
ARM has started pushing 55nm ULP for IoT MCUs. The question is good: why aren't there more super cheap MCUs at mature nodes?
I think the last time I checked the floor for 8/16 bit parts was around 34 cents. IC Insights has a figure for all 20 billion MCUs annually around 89 cents right now, including higher end 32-bit devices.
Some comments hint at linear price elasticity - if prices were 1/4 current levels, volumes would jump to 4x. That sounds good but it would quickly put a supplier out of business. The higher margin 32-bit suppliers won't go there, for sure, and the lower margin 8-bit suppliers are already baked down to their cost floor. Shipments cannot jump suddenly from 20 billion to 80 billion, no matter what most IoT types (and I am one of them) say.
As I posted a few months ago, that 2 cent, trillion sensor community - and they were talking about RFID-like stuff - doesn't have a lot of support. Very few vendors are interested in the economics of that scenario, especially when any complexity enters the equation.
I think there are possibilities to fab cheaper MCUs and drive the curve, but it won't be a precipitous discontinuity - it will be more like steady 10% kind of progress.
1. Supply constraints - three things limit supply - fab capacity, packaging capacity , and the general electronics supply chain.
The fab will manage - if you want a low-end mcu ,on 40nm, and can package small sizes, i once napkin calculated that you can get around a million mcu's from a wafer or something. from that to 80 billion is possible.
Packaging - you'll have to use new packaging technologies(terepac) - with output package sizes similar to current sizes, so i guess capacity limits will be about the same , so it would take a long time to grow.
As for the rest of the supply chain managing it - this looks like the kind of a thing china would love to do , and betting against the chinese on scaling ? not wise.
2. I'm not sure i agree with rapid price elasticity. Yes there are a million possible uses for for IOT microcontrollers , but they are limited not only by price, but by energy, communication price and availability, marketing and deployment, etc etc. Maybe over a 10-20 years period we could get to 4x demand size increase - again supporting the "slow" growth scenario.
3. The 8-bit guys: they could always target the 32-bit market. but it could be a hard battle if the 32-bit guys decide to enter.
So yes, in general i do agree with your 10%/year growth don.
-----------------------------------------
Why is it this way ?
The problem: disruptive price innovation don't work in this industry. What ARM did was huge, and nobody died and no new mcu entrant really succeeded.blame moore's law probably. In other industries , you would see a big mess.
So in order to compete well - you really need to come with something new you can protect.
So i wonder: What if we start from scratch - design an mcu and software that guarantees much better final product quality ? and than we aim this mcu to industrial uses, and build a simple way for owner of the final product to verify this "seal of quality". Would there be a good market for it ? would industrial owners prepare to pay a bit more for this so our chip will pricier than a commodity ?
When you factor in productivity, Chinese or Indian dev teams are not really that cheap ...... Of course, in the end, paranoia sells better than economics or strategy. So ippisl may just be right about China gov.
The greater values added for an IoT solution will come from SW. Supporting that SW will not be just a commodity, bare-butt MCU but an MCU-powered SoC that does wonders with RTOS, Thread or no Thread. This kind of SoC will be neither all that small nor all that versatile.
