Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/internet-of-things-a-semiconductor-dilemma.7681/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Internet of Things: A Semiconductor Dilemma

Anshuman

New member
Internet of Things brings up an interesting dilemma for semiconductor companies. While semiconductor companies make revenue with high volume custom chips (i.e. ASICs & SoCs), the IoT market is too fragmented and diverseto create those high volumes. Nevertheless, analysts continue to predict that IoT will bring more revenue into semiconductor companies. What are we missing?

The shortest answer comes from Gartner which predicts 50% of IoT solutions will come from startups that are less than three years old.

Traditionally, the semiconductor business model has been a sequential value chain model.e.g. Seagate bought custom controllers from Marvell for its hard drives. The model adds value at each stage from the smallest blocks of IP to design and manufacture of full chips and their subsequent sale in high volumes to one or two large customers. This small customer base makes acronyms ASIC & SoC stand more for Application Specific Individual Customer and Solely One Customer chips respectively.

On the other hand, Arduino and the Makers popularized a platform business model like Apple's App economy for hardware prototyping and small volume production. The Arduino shields enabled developers to quickly slap daughter cards of various functionalities on the board and program to specific applications. This agile software-like hardware development, along with open source ecosystems, enables online marketplaces in buying and selling hardware based gadgets. These small volume solutions cannot justify custom chips as ingredients, but they do gobble off-the-shelf standard chips. This app-economy-like model changes the rules of engagement and strategy drastically for semiconductor suppliers as it outsources innovation into the platform and ecosystem.

By many an estimate in 2013, 57% of the US adults (135 Million) identified themselves as Makers. Numerous low-volume products are created and consumed by Makers and early adopters using commodity semiconductors in high volume. This thriving market is evident with the success of component suppliers, like Adafruit and Sparkfun, sales platforms like Tindie and Etsy, and IoT funding trends on Kickstarter and Indiegogo.

If in the process a product takes off and a unicorn, as they say in the VC world, is born, the semiconductor supplier could be a default choice for a custom chip.Many semiconductor suppliers, hoping to net those evasive IoTunicorns, have launched Arduino-like rapid-prototyping boards and arecreating ecosystems around them. For most semis it's a major challenge as the value chain metrics of success, fine tuned over decades, fall short of providing the desired results.

The processes and metrics for viably birthing and nurturing the next Nestor Pebble still seem like an art to many mid to large semiconductor companies.The success of an app economy hinges totally on its developer community. The more vibrant the community the larger and more insightful it gets. Growing on “network effects,” the community provides many advantages e.g. content marketing, shorter time to market, quicker feedback from multiple stakeholders, entry into new markets, early visibility into downstream value chain entities etc.along with strong analytical insights from the community interactions.

Vision Mobile puts the 2015 active IoT developer community at 4.5 Million and growing to 10 Million by 2020. Those that leverage this community are bound to outperform their competitors over time. Startups and smaller semiconductor players are at a major advantage. Firstly, their small size makes them better suited to exploit the agile hardware paradigm. And secondly, the low barriers of entry at the new marketplace make them competitive with their individual niche-product platforms. Intel, MediaTek, STMicroelectronics, etc. all have a variety of rapid prototyping cousins to Arduino and Raspberry Pi, while it was SiliconLabs' chips that lit the audience's bracelets when Coldplay performed at Super Bowl 2016. These music synchronized lighting bracelets are made by Xylobands.

Clearly, for semiconductor companies, curating a developer community is not a choice but an imperative as community members move on to create 50% of the future IoT market. As President Obama pointed out – “Today’s D.I.Y. is tomorrow’s ’Made in America.’”
@anshufellow #Chips4Makers​
 
Is there even such a thing as a chip design/manufacturing startup anymore? Very few come to mind. Cavium? But that was 10 years ago.
 
Is there even such a thing as a chip design/manufacturing startup anymore? Very few come to mind. Cavium? But that was 10 years ago.

According to TSMC, and I agree 100%, a new generation of semiconductor companies is evolving that we call Fabless System Companies which is a system company that will do their own chips. Apple is the biggest example. Amazon just bought a semiconductor design firm etc...

The benefit is that fabless system companies have other revenue streams besides the chips themselves so margins are much higher. The Amazon Echo for example or video doorbells where you pay for cloud storage.

DIY is fine but if you are going to go into HVM and need a competitive edge you really need to do your own chips. Mobile IoT is also an issue where battery power or size is a factor. You will need to get a handful of merchant chips down to one IoT SoC.
 
Makers are just a step in the evolution of some products. Look at the auto or almost any major industry, you start with a few pioneers, then thousands of small makers, then there is consolidation with a few major manufacturers and a number of small players. In the last scenario, the components for all end up being made by a few specialists. On the high end you have TSM making the sophisticated parts and companies like Vishay and Amphenol making the small parts.
 
Last edited:
Makers are just a step in the evolution of some products. Look at the auto or almost any major industry, you start with a few pioneers, then thousands of small makers, then their is consolidation with a few major manufacturers and a number of small players. In the last scenario, the components for all end up being made by a few specialists. On the high end you have TSM making the sophisticated parts and companies like Vishay and Amphenol making the small parts.

Thats true but the industry evolution takes years to a full decade, and for Auto its probably longer. That step - which we call Makers, are today helping decide what IPs shall go into the next ASSP.

IMHO that application specific standard product will have numerous customers and and its application focus is dictated by the Makers and early adopters. Raspberry Pi alone has sold more than 8 million pieces. Surely a few startups have started by using a few of those 8 million. I am not sure when will those startups switch from Broadcom's 2835 & 2836 to ASICs. But I am sure they moved from Pis to some custom boards still using the Broadcom chips.

So till the industry as a whole consolidates IoT applications and players, IMHO there's a lot of opportunity focusing on the makers and early adopters of IoT. Atmel's blog restates Martha Stewart's post that 57% of US population identifies as Makers.

And thanks to the platform approach Broadcom has a much better understanding of its end customers.
 
Last edited:
Is there even such a thing as a chip design/manufacturing startup anymore? Very few come to mind. Cavium? But that was 10 years ago.

The semiconductor industry has its set of startups and IPOs in both process and IP, though the number has shrunk big time.
Adesto Technology a memory player had an IPO last december.
Silicon Catalyst a startup accelerator in the valley has been mentoring startups for a while now.
And then there are companies like GainSpan, Crossbar, Crocus, Everspin, to name a few in the middle.
 
One thing about the this is those makers(or preofssionals among them) are the ones writing your firmware and developer tools, and doing a much better job than companies. This saves a ton of money for the chip company, which is now able to build a chip for far cheaper and share the savings with customers, becoming the low-cost chip maker, which also wins the professional market.

You can see examples for this in the ESP32/ESP8266 from espressif which is the cost leader in IOT chips, while using the community to do a lot of the sw dev, and some hardware debugging.
 
One thing about the this is those makers(or preofssionals among them) are the ones writing your firmware and developer tools, and doing a much better job than companies. This saves a ton of money for the chip company, which is now able to build a chip for far cheaper and share the savings with customers, becoming the low-cost chip maker, which also wins the professional market.

You can see examples for this in the ESP32/ESP8266 from espressif which is the cost leader in IOT chips, while using the community to do a lot of the sw dev, and some hardware debugging.

Exactly! This concept is only enhanced by the online community around the boards that use those chips. And the chip makers reap the benefit and/or pass some on to the customers.
 
And thanks to the platform approach Broadcom has a much better understanding of its end customers.

So that's why they unloaded the IoT biz to Cypress?

I love the maker movement and the smart merchant firms are nurturing their communities. (It's the 21st Century equivalent of the university programs of the 1980s.)

The challenge is who is going to make money producing very inexpensive chips in moderate volumes supporting a fragmented space. Every vendor is looking for features to add to get their ASPs up.

I'm also concerned about the non-chip companies, and even the IoT startups that will see chips as a competitive weapon. Prototypes may very well be on merchant chips, but it's not a forgone conclusion the production chip stays with that vendor - especially true in an ARM ecosystem.
 
Don:

>> The challenge is who is going to make money producing very inexpensive chips in moderate volumes supporting a fragmented space. Every vendor is looking for features to add to get their ASPs up.

Since chips can be made very inexpensively(per chip, new nodes), if it's possible to make a single chip/platform that is both attractive to makers and volume vendors(which seems possible), the big question is how to offer support and marketing at extremely low prices. open source ecosystems seems like a way to do that.

>> it's not a forgone conclusion the production chip stays with that vendor

Well we're seeing this happen in the mobile market, with huawei making their own SOC. But that's probably easy for them . But do you see a large toy company doing the same ? they haven't done so already , and most chips will grow a lot in complexity(nodes,connectivity,power).

Altough maybe working with a specialist design company that has a lot of IP in house and can offer competitive chips is a good option, maybe someone like indie-semi could become popular. But they too benefit from selling merchant chips, low-cost ways of doing support, etc.
 
>> So that's why they unloaded the IoT biz to Cypress?

Absoluely, Broadcom understands that they cannot be as agile and nimble as Cypress in handling the IoT business. Better still I think Avago understands that better.

>> it's not a forgone conclusion the production chip stays with that vendor
>>>>
But do you see a large toy company doing the same ?

True, its way more difficult to dethrone an entrenched SoC than when starting from scratch. The semiconductor vendor has an upper hand in trying to keep the customer with themselves.

And as "making money" continues to be an issue their are high chances that agile companies will rule over non agile ones, IMHO.

I think that played a role in Xylobands deciding to go with Silicon Labs instead of a bigger player.
 
Last edited:
<style type="text/css"><!--td {border: 1px solid #ccc;}br {mso-data-placement:same-cell;}--></style>For me IoT devices is a must have for modern internet surfers. But at the same time these utilities are vulnerable to cyber-attacks posing risks to privacy and security online. Having considered this fact, I’ve always tried to be informed about all hot news concerning cybersecurity, ex like this Risk of having home smart devices hacked grows at an exponential rate TAnyway, I’ve never been hacked or had personal info stolen that I wish to you.
 
Back
Top