The Barriers to Establishing the Industrial Internet of Things: Where are the Opportunities?

This is Part 2 in a series of posts about the Industrial Internet of Things. Part 1: The Industrial Internet of Things and Why It Matters to You and Your Business can be found here.

If you’re building a startup innovating on any of the components of the IIoT, especially those mentioned in the following articles, please DM me on Twitter at @aaroncsmall as I’d love to chat.

What’s Taking So Long?

In Part 1 we saw that we’re in “Phase 1.5” of IIoT deployment: we’re busy capitalising on the falling cost of sensors and bandwidth to connect unconnected assets and make them somewhat intelligent through the use of analytics running on cloud computing platforms. This has resulted in measurable progress in terms of sensors shipped, connections made and increased revenue in the IIoT segment, but despite this (and the many VC dollars invested!), there remain significant barriers to wide-scale adoption.

This perceived lack of progress may simply result from the immaturity of the ecosystem itself. Many of the early plays have been for transverse horizontal platforms. This is exacerbated by a lack of standards and interoperability, creating confusion and delaying deployment. But as long time IoT investor Matt Turck of Firstmark Capital eloquently points out:

“Progress may seem slow in some ways, but in fact it is happening remarkably quickly when one pauses to think about the magnitude of change a fully connected world requires”.

Despite this slow progress the IIoT represents a new era of industrial competitive advantage. If we think back to the World Economic Forum’s IIoT deployment schedule, the focus of industrial processes will shift firmly to the customer. Further innovation and capability development will be required to facilitate this shift which presents an opportunity for early stage businesses and investors. Below I highlight what I think some of these opportunities might be.

Key Barriers, Trends, and Technologies

1. Distributed Analytics
2. Security
3. Interoperability

The Need for Intelligence at the Edge

“Industrial IoT [is at an earlier stage] than most of us think because distributed infrastructure remains in its infancy.” — Michael Yamnitsky, Venture Partner, Work Bench

The cloud services that exist today (i.e. from Amazon or Microsoft, for example) excel at unbundling various capabilities into micro-services like hosting or storage. These allow organisations to scale rapidly and they have been a major driver of IIoT deployment to date. But with Cisco estimating that 50 billion endpoints will be added to the IIoT by 2020, a tidal wave of data will be created that makes the current big data stack ill-suited to the IIoT.

Most industrial data arising from these endpoints will be from various sensors and disparate software programs and hence heterogenous or too noisy. Often the endpoints will be in remote, harsh, low bandwidth environments that require event driven analytics from streaming data (rather than static data sets) that are latency-sensitive and too expensive to be carried all the way back to the cloud. For example, the need to predict impending brake failure on a train, or re-route a supply chain depending on real time customer demand.

Software is Eating the World, and AI is Nibbling at the Edge…

Distributed analytics in combination with machine learning solves this problem and is why I believe we are at inflection point in value realisation from the IIoT given the advancements in machine learning in recent years. Rather than data processed and stored in the cloud, many IIoT applications will require network intelligence to reside closer to the source: what is known as the “Edge” — almost a reversal back to the distributed PC revolution of the 1990’s and 2000’s — and away from the cloud.

These resulting systems are highly focused “Systems of Intelligence“ — deeply analytical systems that bundle capability, rather than scale, to give a process advantage in solving business challenges. For example, using predictive analytics to optimise maintenance cycles in a manufacturing plant.

The cloud giants are less well suited to this task of building Industrial Systems of Intelligence that help customers solve real world business and operational process problems. While they provide scale, they lack the domain knowledge and capability on an operational level to implement their products down to the endpoint in traditional industries (hence the partnerships that are starting to form in the IIoT space, for example GE and Microsoft Azure). Furthermore, on an AI level, the monolithic and open source machine learning algorithms they are developing are equally unsuited to the remote fabric of the IIoT edge.

Conversely, incumbent industrials (such as GE, Schneider Electric, and Siemens) need to become software companies, and in particular, AI companies. This is a major challenge for them.

The opportunity therefore exists for early stage companies and investors to build barriers to entry based on the Industrial Systems of Intelligence thesis as a result of a new architecture that will emerge, comprising:

1) AI at the edge:

• Distributed analytics at the edge, powered by machine learning, will be used to make instantaneous decisions.

2) Smart Gateways / Cloudlets:

• Software infused smart gateways or “cloudlets” will be required near the edge for local data processing that may be too resource intensive for endpoints. The smart gateways will also be used as intermediary data stores in order to capture data not used in complex event processing for the training of machine learning models.

3) Centralised Cloud

• Traditional use of unbundled microservices (hosting, storage etc.) but also deep data analysis and training of machine learning models.

In addition, as compute moves closer to the edge, competitive advantage will be borne out by those that can process data faster and more efficiently. Applied or Vertical AI software in particular is highly specialised and often requires a full stack solution that makes use of some sort of proprietary hardware to collect data or bring a processing advantage. Chipmakers have realised this and a range of custom chips have been released (for example Intel’s Myriad X and Nvidia’s Jetson) for edge applications.

Example companies: Alluvium; Camgian Microsystems; C3 IoT; Cloudleaf; Falkonry; Foghorn Systems; Maana; Mythic; Nebiolo Technologies; Senseye; Sightmachine; Uptake; Vapor.io; Vimoc.

Getting it all to work together

Once we are able to deal with the incoming data tsunami, we still need to solve the interoperability challenge. Most data captured by sensors today is used to monitor discrete machines or systems. McKinsey estimates that on average, interoperability is required for 40 percent of potential value across IoT applications and by nearly 60 percent in some settings. Interoperability would therefore significantly improve performance (and prove out RoI cases) by combining sensor data from different machines and systems to provide an integrated view of performance across an entire enterprise.

This is not a trivial task however and requires integration across multiple systems and vendors, and sometimes across industries.

While standards may emerge over time (see the Open Connectivity Foundation’s 1.3 Specification, Open Mobile Alliance’s Lightweight Machine-to-Machine (LwM2M) standard, and Open Process Communication’s Unified Architecture (OPC-UA)), Systems Management Tools and API Management Tools will also play a key role in allowing access to all layers and functionality of the IIoT, from communications to data to applications.

“The next frontier is systems management software bridging disparate IoT software systems.” — Michael Yamnitsky, Venture Partner, Work Bench

Several companies and organisations are also working on using blockchain technology to address the interoperability problem. Although at a much earlier stage, this is a promising application of the technology that can help address interoperability issues as well as the security issues outlined below, through individual endpoint authentication and sharing of data over a distributed computing platform. This will be particularly useful on a transactional level, for example containers being logged, verified and billed for as they pass through customs at a port.

Example companies: Chronicled; Chain of Things; Golgi; IOTA Foundation.

Securing the IIoT

Distributed analytics architectures with billions of endpoints unfortunately provide a multifaceted “attack surface” and introduce a new security risk to enterprises. Security is in fact cited as the number one concern of executives regarding the deployment of IIoT capabilities. Many of the traditional IT security tools available today are unsuited to the billions of devices or endpoints that will result from a change in architecture to the edge. Security tools will be required that are adaptable to a range of IIoT use cases at the edge, that can evolve with an evolving industry, and can protect assets that may be long-lived and often have simple processors and operating systems that may not support sophisticated security approaches.

Like the cloud giants, traditional IT security vendors are perhaps ill-suited to instrumenting endpoints and gateways in an industrial setting, which provides an opportunity for early stage companies and investors (and potentially those already providing distributed analytics tools) to offer security systems applicable to the IIoT edge. I predict several acquisitions in this space in coming years.

Example companies: Armis; Bastille; Claroty; Cloudflare; Filament; Mocana; Newsky Security; Prove & Run; Pwnie Express; Qadium; Rubicon.

In summary…

• The need for distributed intelligence at the edge will necessitate a new architecture based on AI at the edge and smart gateways/cloudlets.
• Cloud giants lack the operational capability down to the end point in traditional industries, while industrial incumbents will struggle to build software companies.
• The opportunity exists for early stage companies and investors to facilitate the shift to a “hybrid cloud” architecture based on distributed intelligence, that is secure and interoperable.
• By doing so they can build a new moat = Industrial Systems of Intelligence that offers a business process advantage, which is where 90% of the IIoT’s value will reside.

In the next article I will share a framework for analysing opportunities in these areas around such Industrial Systems of Intelligence, and offer some advice to IIoT entrepreneurs based on conversations I’ve had with industry insiders.

Sources:

1. Internet of Things: Are We There Yet? (The 2016 IoT Landscape), Matt Turck, Firstmark Capital.
2. Industrial Internet of Things: Unleashing the Potential of Connected Products and Services, World Economic Forum.
3. 2017 Enterprise Almanac, Michael Yamnitsky, Workbench.
4. The Internet of Things: How the Next Evolution of the Internet Is Changing Everything, Cisco.
5. The New Moats: Why Systems of Intelligence are the Next Defensible Business Model, Jerry Chen, Greylock.
6. Microsoft Azure Press Release.
7. The Next Phase Of Machine Learning, Semiconductor Engineering.
8. Unlocking the potential of the Internet of Things, McKinsey Global Institute.
9. Open Connectivity Foundation.
10. Open Mobile Alliance.
11. OPC Foundation.
12. Without API Management, the Internet of Things is Just a Big Thing, Wired.
13. Decentralizing IoT networks through blockchain, Techcrunch.
14. IoT Technology Disruptions: A Gartner Trend Insight Report, Gartner.