After I read Brian Bailey’s IoT semiconductor design article, IoT Myth Busting, I thought of Prince’s song 1999, in particular, the line:
“So tonight I'm gonna party like it's nineteen ninety-nine.”
Without a lot of irrational exuberance, we won’t see IoT edge and fog networks soon
Most IoT applications are prototypes and proof of concepts (PoC) designed to justify enterprise budget increases and follow-on venture investment rounds. Unless we return to and party like it is 1999 when telecoms over-invested in capacity ahead of demand, the telecom carriers are not going to build the new fog and edge networks that IoT needs to grow ahead of demand. At this stage, we would have to see a return of the irrational exuberance, a term coined by Federal Reserve Chairman Alan Greenspan, used to describe the over investment and over valuation during the dot-com bubble.
The telecoms will productize 3G and 4G services customized for IoT also categorized as machine to machine (m2m,) enabling prototypes and PoCs. They will wait for the IoT market to grow before investing in the more efficient and IoT purpose-built fog and edge networks. But if telecom providers are not good predictors of IoT, what are?
Timing is everything. The forecasts quantifying IoT remind me of something the retired star tech analyst Elliott Gold said in response to the question about when the early stage videoconference industry would reach 1 million endpoints. In response, Gold said:
“Analysts are too optimistic forecasting a new category at the beginning of a growth cycle and too pessimistic forecasting the longer cycle."
Gold’s comment has never been truer than in the case of IoT.
The definition of IoT is too broad to measure
Readwrite and Spoke, in partnership, published the IoT Revolution Handbook, citing the following U.S. IoT metrics:
- 3,000 companies
- $125 billion in funding
- $613 billion in valuation
- 342,000 employees
These metrics include a very broad definition of IoT, though:
Smart consumer/user
Facilitative reality: Augmented reality (AR) and virtual reality (VR)
Connected home
Shared economy
Healthcare
Connected Cars
Smart enterprise
Utilities
Building construction
Oil and gas/energy
Retail
Smart data
Big data
Data security
AI and machine learning
Smart cloud
Cloud life cyce
Data centers
Infrastructure as a service
Platform as a service
Cloud security
Connected and autonomous things
Wearables
Robots
Vehicles
Drones
Machines
Smart networks
VPN/network security
Ethernet/wired
Satellites
Platforms
Wi-Fi
Cellular
Many of the market segments are really just common infrastructures across many tech silos, such as data security that serves many large segments, including IoT, or is just one application of a technology, such as AR and VR, that has many other applications ranging from architecture engineering and construction to entertainment.
EDA toolchains and purpose-built IoT chips may be the best indicator of IoT growth
The Readwrite/Spoke definition is a good start, but it’s to broad to drill down into IoT development. Semiconductors and electronic design automation (EDA) are very important to watch to understand IoT development. These tool makers are at the nexus of prototypes, PoCs and products. Every platform—mobile, PCs, artificial intelligence—has an integrated EDA and semiconductor toolchain. IoT will eventually have its own toolchain. If this can be predicted, IoT growth can be predicted.
According to Bailey, there is not a consensus yet about what this toolchain will finally look like, but there is a consensus that it will follow 30 years of industry history.
An IoT chip platform will emerge. Like the Intel x86 and the toolchains became the PC platform and ARM and the surrounding toolchain became the mobile platform, a company in the IoT segment will win the position as platform leader. There will also be custom solutions. But custom will not mean small quantities. Custom solutions will take existing intellectual property (IP) and apply a non-recurring engineering (NRE) investment to optimize the IP to achieve economies of scale and turn a category of prototypes and PoCs into products and take it to market. Higher levels of device integration will let these devices hit the very inexpensive and power-efficient levels that will foster forecasted high volumes.
Security will be a core component for these devices. Custom circuits implemented specifically for security will be implemented. Eventually, there will be a consensus about how to do this, but few companies are ready to roll their own until the full security software stack is understood.
Consensus has not been reached between distributed and centralized architectures either. Architectures range from completely distributed systems at the edge with little processing in a central cloud to completely unintelligent devices at the edge with all processing taking place in the cloud.
It is implied that many chip designs will land somewhere in the middle. There will be more design variety and diversity than previous platforms to achieve the integration implied by cost and power efficiency. This diversity means more engineers will need to design custom chips, which might be best served with new EDA business models as well as new toolchains.
The semiconductor and EDA industry is also a good measure of IoT development because as fierce as the competition is, most companies understand they need to share industry data. Through organizations such as semi.org, the industry shares data that they all use to forecast and time investment in new products and new capital equipment.