Jeffrey Bornstein, SVP and CFO at General Electric Company, took time during his fireside chat at the MIT CFO Summit to make a distinction between what he called the “consumer internet” and the “industrial internet,” a term coined by GE that refers to industrial machine to machine communication.
“The architecture and the infrastructure of the industrial internet is completely different than the consumer internet,” he said. Rather than send all of the data to the cloud for analysis, the industrial internet relies on edge computing, the ability to collect and analyze data closer to the machine itself.
When a customer buys an item on Amazon, the transaction happens in seconds — not milliseconds, Bornstein said. And that’s because Amazon captures the data; sends it to the cloud; analyzes it; runs it against a profile it has of the customer; checks on the product’s digital twin, a digital representation of a physical asset; and then responds, according to Bornstein.
“When Amazon sends you back recommendations and you look at it, it takes you a couple of seconds to ingest it, and if you decide none of that is interesting, there’s no cost associated with that. There’s no cost for the false positive,” he said. “But in the industrial internet, the cost of a false positive could be catastrophic.”
A piece of industrial machinery like a jet engine, for example, needs to be able to respond to requests — such as a data anomaly — in microseconds rather than seconds, Bornstein said. Rather than send all of the data to the cloud and wait for a response, a good chunk of the data analysis needs to happen at edge — on or near the industrial machine itself.
“A very small amount of data will migrate all of the way to the cloud where you’ll do long-term data and analysis,” he said. But most of the compute happens where the machine lives.
There’s another reason for keeping compute local. “The amount of data we’re talking, just in our jet engine business, is way beyond anything Facebook or Google deals with,” he said. “So you’d never be able to pour all of that information to the cloud anyway.”