As GE preps to sell some of its digital businesses — perhaps even its Predix platform, which is designed to take manufacturing data and analyze it as part of GE’s industrial IoT efforts — I’ve been thinking about the fate of companies at the edge. It all started with a conversation at Google’s recent Cloud Next event, where I chatted with David King, CEO of FogHorn, about what plant managers want.
I asked how far Google, Amazon, Microsoft, and the IT guys would travel toward the edge. In this version of the edge, I’m thinking about gateways and devices on the factory floor; in recent months, each platform provider has built software that helps extend their cloud operations to devices that run on premise. But in King’s opinion that software is still fairly minimal, and none of these big providers really want to expend a lot of effort customizing their software for specific edge cases.
To be sure, King has a dog in this hunt. His firm’s software runs machine learning models at the edge and helps beef up the cloud providers’ bare-bones efforts. But he’s not wrong. I’ve watched for the last two years the big tech guys start with cloud-centric products and then reluctantly move closer to the edge with products and services. They had hoped to leave that work to the dozens of specialist companies out there, the likes of which range from established vendors such as OSIsoft and Kepware to the large industrial companies such as Siemens, Emerson, Honeywell, and even GE.
Underlying it all is a reluctance to move away from the scaled-out cloud platforms that the IT giants know and understand how to monetize at scale and the realization that edge computing is incredibly complicated and requires levels of reliability that even the large cloud providers can’t meet. If you’re Google, providing 99.9% uptime (three nines) is excellent, but for the companies building products that remotely operate big machinery, even 10 nines isn’t enough.
There are also a mess of machines, standards, and protocols out in the real world at the actual edge, or what some people call the “far edge” or the “remote edge.” Think of it as the controller on machines, cars, or even gateway computers responsible for sending out commands to traffic signals. There are more than a 100 different machine protocols that I’m aware of, and I can guarantee there are more I’ve never seen. The big IT guys are adept at running monolithic services and apps; they want no part of this messy world.
Which means companies that can handle these standards or the industrial giants that are expert in one particular industry have a leg up. Recognizing this, we’re seeing large industrial firms buy smaller startups focused on the industrial internet of things. So while many of these startups, which are often founded by old IT guys, may anticipate an acquisition by an IT shop, the real interest will come from old-line industrial firms trying to understand the particular needs of IT where it meets operations.
Chip Childers, CTO at the Cloud Foundry Foundation, explains that the future for these heavy industry companies is to get good at software engineering. (The Cloud Foundry Foundation works with a lot of cloud and industrial edge providers.) Childers says those edge experts will still be essential for the industrial internet because they understand these machines and their processes, and that whatever you put on these products must survive for the life of a machine. “They have to bring software skills to the edge,” he says. “They have the needed history and need to get better at being more effective. They must learn to write the software and leave room for innovation later.”
This is already happening. Just this week for example, Siemens said it would spend $730 million on a company called Mendix to make developing applications for its platform easier.
All of which brings me back to GE and its potential sale of digital assets. When it comes to businesses like Predix or ServiceMax, any would-be buyers will likely come from GE’s current crop of competitors. Unless one of the big IT guys really wants to get its hands dirty and explore an entirely new style of business.
Michael Kanellos says
You hit it on the head. The ‘edge’ which can be an entire wind farm or a sole sensor, is more complex and demanding than people realize and it doesn’t fit easily with the scale up economic model of some cloud providers. (A server might last 4 years; a transformer must live for 40). I think we’re moving to what you could call the ‘how’ era where companies figure out how to assemble the most effective architectural stack. It will be similar to how IT got built out in the mid 80s