The GE Evolution series locomotive is a beast of a machine.
Measuring 73 feet long and weighing in at around 436,000 pounds, the Evolution drinks diesel from a 5,300-gallon tank as it chugs around the country hauling enormous loads of iron ore, grain, or whatever else needs to be moved cheaply from point A to point B.
In the U.S., the Evolution is built in two places: an old GE industrial complex in Erie, Pennsylvania, and a new manufacturing plant just outside Fort Worth, Texas. If you visit the Texas site and enter a section within the immense factory known as the platform area, you might think you’ve stumbled upon Vulcan’s workshop. The platform area is a room about the size of a football field, where the raw chassis of locomotives are fitted together and ushered toward the assembly line. The air inside is choked with smoke. The hazy atmosphere flickers crazily from the electric arcs of welders perched on the huge steel bones of unfinished engine cars. Even with earplugs, the din from the hammering and clanging, underscored by a deep rumbling tremor from machinery, is nearly deafening.
On any given day, the biggest freight-rail haulers in the U.S.–“Class I” railroads such as Union Pacific, Norfolk Southern, and BNSF–move about 24,000 locomotives, and 365,000 freight cars, over a 140,000-mile-long network. General Electric is rolling out a suite of Industrial Internet tools for these haulers to improve efficiency. By GE’s calculation, even a 1% gain could translate into $2.8 billion in savings annually. Here’s how.
Freight locomotives carry massive loads of expensive diesel. GE’s Trip Optimizer is a type of cruise control that combs through piles of data and synthesizes them for the driver in a way that allows him to steer the locomotive to maintain the most efficient speed at all times and reduce fuel burn.
A GE locomotive is fitted with multiple sensors that transmit huge amounts of data about performance. By heeding warning signals from these sensors, analysts can recognize an impending problem, schedule a train for service, and get a part ready to install–thus averting a breakdown on the tracks.
Freight haulers try mightily to reduce something known as dwell–the time a freight car is stuck in a train yard. To quicken the breakup and sorting of trains, GE’s Yard Planner tool goes through the data on a yard’s train cars (location, destination, classification) and accelerates their dispatch.
Freight trains are slowed by congested rail networks. By using sophisticated data software and keeping track of every train’s present and future location, the Movement Planner tool intends to fit more trains on the tracks at any given time.
The upshot: an increase in average velocity of as much as 15% to 20%.
Illustrations by Justin Mezzell
In an economy that depends upon the swift, silent transmission of information, the construction of a massive locomotive can seem like a primitive anomaly. But the smoke and clamor in the big Fort Worth plant obscures some things. For instance, at roughly 220 tons, the Evolution isn’t heavy due to some sort of technological inability to make it slimmer, in the way automobiles are now “lightweighted” for better gas mileage and efficiency. The Evolution is intentionally heavy so that its traction motors can better grip the tracks. More noteworthy is the fact that while the Evolution may look old-fashioned, it is in many respects a hurtling computer. Its array of sensors and data-collecting devices complements its bulky mass with a sleek, digital agility that will grow only more impressive and more significant with time. In this fusion of old and new, this melding of heavy and light, you can see that the Evolution resembles its maker, General Electric, a company that manufactures huge things for huge customers and yet is reinventing itself–and, in the process, the very economics of heavy industry–by embracing a new kind of sophistication.
Two years ago, at a San Francisco conference billed as “Minds and Machines,” GE CEO Jeff Immelt took the stage to explain the company’s behemoths like the Evolution. The catchphrase he used that day, the “Industrial Internet,” has by now become commonplace in technology circles, even though it has been barely realized in terms of impact. At times, the Industrial Internet has been lumped alongside the so-called Internet of Things, which usually describes the effort to bestow networked connectivity on, say, your home lighting or thermostat. Yet GE’s industrial effort is more ambitious than that. Immelt’s point in his speech was that GE could no longer just build big machines like locomotives and jet engines and gas turbine power plants–“big iron,” as it’s known within the company. It now had to create a kind of intelligence within the machines, which would collect and parse their data. As he saw it, the marriage of big-data analysis and industrial engineering promised a nearly unimaginable range of improvements. A new GEnx jet engine with a multitude of sensors could spin off an awesome amount of information. GE would in turn help predict, say, when a crucial engine part required repairs. GE would use data from machines like the Evolution to optimize performance to undreamed-of levels.
In the time since the Industrial Internet’s unveiling, GE has labored to transform its vision into reality. On a frigid day last March, I spent an afternoon with Immelt at GE’s global research center in upstate New York. We toured several labs where GE is working on new energy and aviation technologies; then we chatted at length, both in private and with his executive research team, about the future of the company. In person, Immelt comes off much like he does on CNBC: informal, confident, and with a knack for simplifying what is likely the world’s most technologically complex and geographically diverse company. He’s a good talker, but a good listener, too, especially when a GE staff scientist is speaking. (Perhaps unsurprisingly, he can run a meeting like no one I’ve ever seen.) From the beginning, Immelt tells me, he knew the Industrial Internet project was something that would take at least a decade to deploy. But as someone who spends a large amount of time playing salesman to GE customers–it goes with the job when your buyers spend hundreds of millions of dollars on power plants and locomotives–he saw no choice. “I’ve been doing this for 30 years,” he says. Being able to walk into the offices of an airline or a freight CEO and tell him that data might ensure that GE jet engines or locomotives would have no unplanned downtime could change the way Immelt’s company does business. GE products could almost sell themselves–or, if his competitors had the capability to do this first, stop selling at all. In a future of intelligent devices, Immelt adds, “it’s not like you have to have a lot of fancy ideas or showmanship or salesmanship. You’re actually giving people the answer to a question they’ve asked for a long time. In the world that I live in, the industrial world, the notion of how analytics can be used, well, people find it to be beautiful, desirable, investable. They understand it inherently right away.”
Still, promising that locomotives or power plants won’t unexpectedly break down is different from making such a future actually happen. “The question is,” Immelt acknowledges, “can you do it?”
Mashing big data with big machines is “beautiful, desirable, investable,” says Immelt. It could transform GE’s business–and the economy.
That question reaches beyond the future of General Electric. In trying to build smarter machines, the company is also vying to create a new industrial age that produces broad, rippling gains for the entire global economy. What’s curious, though, is that GE’s grand endeavor isn’t about disruptive innovation, at least by Silicon Valley’s standards. Rather, it is about the vast potential of incrementalism–that is, a predicted 1% improvement in the productivity of GE’s big iron. At first glance, such a number seems terribly modest. But by merging the stolid industrial world with the fleet digital world, Immelt believes GE can achieve something revolutionary. The scale of the company’s products is so large that small gains in their productivity could, Immelt proclaims, “drive massive economic benefit.” He is talking about hundreds of billions of dollars. Immelt, who recently indicated he might step aside as GE chairman in the not-so-distant future, might also be talking about his legacy.
I am sitting in the driver’s seat of a GE locomotive, hand on the throttle, eyes on the horizon, hauling a mile-long freight train that left Kansas City, Missouri, just a few minutes ago en route to Amarillo, Texas. Hills are visible just ahead, big curves, too, and managing all this bulk tonnage is far more difficult than driving my Subaru around town. To be honest, I’m a bit out of control. Also, I’m burning through way too much diesel. To its credit, GE will not actually let me drive a big locomotive. In fact, it’s late morning and I’m in Niskayuna, New York, a few hours north of New York City. In a darkened room, I’m pushing that mile-long freight train from Kansas City via computer-aided simulation to try something known as GE’s Trip Optimizer. The optimizer is a type of hyperintelligent cruise control that can be used by locomotive engineers out in the field–the kind of engineers with actual licenses and train experience–to help them act on information they couldn’t possibly gather on their own. By taking a constant pulse of the locomotive’s geographical location, weight, speed, fuel burn, and the terrain, the Trip Optimizer can calculate what the train’s proper velocity should be at any given moment. It can remind a driver, for instance, that instead of pushing his engine hard to the crest of a big hill, he can ease up on the throttle a certain distance before the crest and let momentum carry him over. With tricks like this, the optimizer can save operators tens of millions of dollars in diesel. GE’s Industrial Internet may seem like a single, comprehensive approach to infrastructure technology. But in truth, the Industrial Internet is made from a dizzying number of components–software as well as hardware–that will be rolled out over the course of the next decade. At the moment, Trip Optimizer is an early Industrial Internet software product for locomotives that’s already being deployed by freight companies such as BNSF and Norfolk Southern. The next few years promise a range of other tools, such as those that remotely monitor equipment in real time. As Sham Chotai, the CTO of GE’s transportation division, tells me, the company is on its way to knowing “where any locomotive is at any time, what kind of problems it has, what kind of problems it will have in the future, what kind of weather it is going through, whether the operator is conserving fuel or not,” and a host of other data. In an effort to boost operational efficiency, trains will soon talk to other trains too: You slow down and we’ll pass each other on the siding. “I think in five years,” Chotai concludes, “we’ll be thinking very differently about the potential for rail in North America.”
Part of the challenge for Chotai is in trying to update both the freight trains and the aging rail systems they run on. The two are formidably complex. A GE locomotive, made up of about 200,000 parts, has always been a lot like a rolling power plant: Its engine burns fuel that (1) turns an alternator that (2) produces electric current that (3) drives an electric traction motor that (4) turns its steel wheels forward. But now that it is also a rolling electronic laboratory, a locomotive’s insides contain 6.7 miles of wiring and 250 sensors that put out 9 million data points every hour. In the coming years, the number of sensors and data points will climb precipitously. At the GE plant in Texas, the head of GE’s transportation division, Russell Stokes, offers an example of what the technology could do to improve the performance of a locomotive. “One of our big problems is if a bearing starts to fail,” Stokes says. An axle on a locomotive might then freeze, and a train will be marooned on the tracks. In remote country, this can be disastrous as well as expensive. Imagine the traffic jam stretching for miles behind the stalled train. Then imagine having to lift an immobile, 436,000-pound locomotive off the track with a crane and having to somehow transport it back to the shop. This is something that currently happens all the time.
GE has begun putting a radio-frequency sensor inside the locomotive’s gear case to transmit data on oil levels and contaminants, and by parsing that data, Stokes says they should be able to predict the conditions that lead to axle failures. Precisely what the vital signs within these machines can signify, or how long it takes for a locomotive with the equivalent of a fever to have a breakdown on the tracks, is still being worked out; and in this way, the Industrial Internet is still shiny, new, and confusing. “The goal is not just to take data I have today, but to go back and look at the data we have already and see if it shows we could have predicted a historical failure,” Stokes says. His team would look at the broken-down locomotive and comb through its data banks to try to discern a pattern. “We want to turn that into an algorithm that helps us predict the future,” Stokes explains. “We want to say: These three conditions, in this sequence, mean there’s a 90% chance this failure will happen.”
Stokes thinks that sensors that predict the degradation of parts will translate into billions of dollars of savings over time for GE’s rail customers. When data coming off the locomotive helps optimize the country’s network of trains, the gains could be astounding. It is surprising to learn that the average velocity of a freight train making long-distance runs between cities in the U.S.–Miami to New York, for instance–now ranges between 20 and 25 mph. “It’s incredible, isn’t it?” says Immelt, shaking his head. These trains don’t crawl because they’re so big; a locomotive pulling a string of boxcars can actually get up to 70 mph on open track. The average speed is so low because congestion in the train yards, breakdowns, and the frequent necessity of letting other trains pass creates slowdowns and stoppages. GE’s solution is a tool called Movement Planner that aggregates data on velocity, traffic, and location from many locomotives, and thereby increases the average speed of its customers’ trains. “Norfolk Southern will tell you a 1 mph increase for them could be worth $200 million,” says Stokes, who notes that GE’s goal is to boost speeds by up to 4 mph. When I ask Oscar Munoz, the chief operating officer of freight operator CSX, how the GE technology is helping to raise the company’s average velocity, he says, “Over 10 years, it’s gone from 18, 19 to 21 [mph]. It’s still glacial, but I think you’ll see it continue to rise.” Every uptick in velocity, he says, will indeed have huge financial implications.
Photo by Dan Cole
Machines that talk, machines that react, machines that constantly update their status–it sounds a bit like a social network… of machines.
Making locomotives smarter and more networked applies to GE’s other machines too–jet engines, gas turbines, water pumps, even oil- and gas-drilling equipment. The goal is to wring small improvements and the ensuing massive savings from all of them. A new GE wind turbine, for instance, collects data on wind speed, wind direction, and air pressure so that it can be analyzed by a nearby computer known as a controller; in response, an algorithm from a new software program might determine that a turbine blade should, say, alter its pitch to improve its power output. In some cases now, that turbine also transmits its new orientation to other turbines nearby, so the entire fleet can be synchronously improved. Machines that talk, machines that react, machines that constantly update their status–it sounds a bit like a social network… of machines.
On the television show 30 Rock, the great tragedy to befall Alec Baldwin’s character, Jack Donaghy, was that GE, the global powerhouse where he had risen up through the ranks from a microwave-oven executive, had sold NBC to Kabletown, a piddling cable company that made most of its income through pay-per-view pornography. As in all good comedy, there was a shred of truth in that–at the time, GE was selling its stake in NBC/Universal to Comcast–but within GE there were high expectations, rather than regret, about the strategy. “Over the past few years,” Beth Comstock, GE’s chief marketing officer, says, “it’s been very much about focusing the company on fewer, higher technology, more industrial businesses, and we’ve had to shed a lot of pieces of the company to get there.” Improving the company’s coherence, in other words, meant jettisoning the huge media division and radically shrinking GE’s financial-services arm, which had made GE deeply vulnerable during the recent recession. Immelt also believed that those changes would translate into improved profits and a rise in the company’s share price. GE’s value had badly trailed the S&P 500 since Immelt took over, just days before the September 11 attacks.
GE’s effort to push its machines into an era of big data has all happened during a time in which the word industrial became a touchstone for the company’s evolving identity. Indeed, even with its ready brand recognition by consumers, GE’s main customers have increasingly become corporations, utilities, governments, and hospitals, spread around 170 countries. Think of United Airlines’ jet engines, for example, or the Kingdom of Saudi Arabia’s gas-turbine power plants. In turn, the company’s main “businesses” (the term used within GE because its divisions are both immense and semiautonomous) are, in descending order of size: power and water, aviation, health care, oil and gas, and transportation. All of these are industrial to the core. GE’s home-appliances business, meanwhile, which the company attempted but failed to sell in 2008, makes only a tiny contribution to GE’s bottom line. So too does lighting, which happens to be the whole reason GE got started 122 years ago.
Immelt’s pivot has worked to a certain extent: The stock price has climbed steadily over the past five years, thanks to numbers like last year’s profit of $24.5 billion on revenue of $146 billion. Still, GE shares remain well below what they were when Immelt took charge of the company, which means his bet on the Industrial Internet is high stakes indeed.
GE, at any rate, doesn’t make its pile of money merely from selling big machines. It makes as much, or sometimes more, from servicing those machines via customer contracts, now worth some $180 billion in all, that can stretch for 20 years or more following a sale. At one GE lab I visited in Schenectady, New York, a team of about a dozen engineers sat in a large, hushed room, each manning a desk with several computer screens, monitoring the moment-to-moment performance of thousands of GE wind turbines around the country. This happens all day, every day. Similar operations exist around the world for GE locomotives, jet engines, and gas power plants. The company is in the midst of developing a closer–and far more entwined–relationship with its clients. “The first wave was: It breaks, we fix it,” Comstock says, talking about how things worked in the 1960s and 1970s. The second wave, developed in the 1980s and 1990s, were service agreements that assured customers that a GE-built jet engine or turbine would achieve a certain level of performance and would have regularly scheduled maintenance based on GE’s experience with the wear and tear of its parts.
An approaching third wave, enabled by data and analytics, does something new. It strikes an agreement between GE and a customer for a certain kind of outcome, rather than a certain kind of functionality. It’s not only about measuring whether a jet engine is working up to its specifications, or about repairing it on time, but whether it’s delivering, say, the agreed-upon amount of peak operational time. “We’re getting to the point of selling thrust, not engines,” says Brad Surak, a software manager for the company. “Or we’re selling locomotion, not locomotives. And we guarantee that back to our customers. If an engine breaks down on a Southwest Airlines jet full of passengers sitting on a tarmac and they need to take passengers out of the plane and reroute it, we pay penalties to them because of it.”
This is as good an explanation for the Industrial Internet as any. You can’t make such guarantees to customers without an approach that helps you spot patterns in the performance data that will, in turn, help avert that Southwest Airlines cancellation. To Immelt, this is why the company’s effort to collect and interpret the data from its machines is so potentially transformative. In a best-case scenario, “predictive” analytics translates into better products, better sales, happier customers, better service agreements, and better company profits. It also ensures that GE will oversee its industrial products from cradle to grave. If another company–a Silicon Valley startup, say–figures out how to do the analytics on GE’s industrial equipment first, the industrial giant might see the lifeblood of its service business threatened.
But what about the risks here? If you’re selling thrust or locomotion, you’re always on the hook for ensuring a good outcome and not just a good product. And that means swallowing an ocean of information that no company has really had to swallow before. When GE began outfitting its machines with a large number of sensors, the company realized it would be generating more data than the company or its clients knew what to do with. “A single blade in a gas turbine, if you put a lot of sensors on it, can generate 500 gigabytes of data [each day],” Sham Chotai tells me. “That’s just one blade. So that means in 30 days you’re generating as much content as the print collection of the Library of Congress.” To take another example, every pair of GEnx engines, which are installed on Boeing’s new 787 Dreamliner, generate a terabyte of information every day.
There are about 4,500 GE gas turbines around the world, each with dozens of blades. There are around 22,000 GE wind turbines–and 20,000 GE locomotives–all outfitted with multiple sensors. There will soon be thousands of GEnx jet engines taking thousands of trips a day. One thing Immelt did not explain when he unveiled the Industrial Internet was that GE was beginning to realize that the data sets coming off its industrial machines would dwarf the kind of data generated by consumers from the conventional Internet. And even with some 300,000 employees, he had already come to the conclusion that the company didn’t have the right data scientists who could make sense of it all.
If one of the defining moments for GE has been the decision to launch a quest to make intelligent machines, the other has been to shift some of the company’s crucial operations from the East Coast to the West Coast. Building the Industrial Internet has meant building a presence for GE in Silicon Valley. The man behind this effort is a mild-mannered California software executive named Bill Ruh. A few years ago, Immelt hired Ruh from Cisco and effectively gave him $1 billion to rebuild the company’s entire software and analytics approach. Ruh set up shop in San Ramon, just east of Oakland. Then he began the task of hiring 1,000 software engineers and data scientists.
Inside the fan casing of an Engine Alliance GP7200 enginePhoto by Adam Senatori
In San Ramon, you see informal interiors, collaborative work spaces, and whiteboards everywhere. For better or worse, it feels a bit Google-y.
Ruh’s mind seems as organized as a circuit board. In chatting with him, you can glimpse how analytics are not only the focus of his new job but the way he sees the world, with an answer to every question branching off, logically and sometimes ornately, into decision trees of accruing reason. When I visit GE’s new San Ramon lab, it’s a brilliant California morning, the sun reflecting off the enormous glass-and-concrete structure. After a tour of the office, Ruh says the lab is already staffed with 800 people. That means about 200 to go. The paint seems barely dry–downstairs, workmen are still nailing up drywall and running wiring past my feet. Ruh tells me, “When anyone shows up at this building from GE who works on the East Coast, or a customer who knows GE, the first thing out of their mouth is, ‘This doesn’t look like a GE building.’ ” It’s true. Ruh brought in the same designers who work on Google’s campuses to break from the tidy and conservative GE conventions. How else could the company compete on this playing field? He was not hiring people who tuck in their shirts and take multivitamins in the morning; he was hiring people who write code all night and rarely (or barely) make it to the office before 10 a.m. Thus in San Ramon you see open and informal interiors, collaborative work spaces, and whiteboards everywhere. It feels a bit messy; for better or worse, it feels a bit Google-y.
In the course of 24 months, Ruh says his team has built a new software platform, known as Predix, and late last year, GE began to deploy it. Predix serves as a new type of operating system–think of it as GE’s Android–that will create a common language for all of the company’s industrial equipment. It will make pulling predictive data from turbines and locomotives much easier, and it will improve troubleshooting by staffers in the field armed with mobile devices. “We’ve always talked about how this is a 10-year journey,” Ruh says. “And at this point, I’d say we’re really in year three.”
Industrial companies that don’t invest in data now, says GE’s Ruh, will eventually be like consumer companies that missed the Internet: “It’s going to be too late.”
Ruh can make Predix sound a bit like consumer technology–user-friendly, customizable, scalable. And that’s not a coincidence. As he sees it, the digital revolution we’ve witnessed in the consumer arena is at last ready to invade new territory. “I believe in this stuff more than you can imagine,” he says, “because I think all machines are going to get smart. They’re going to talk. The technology is there, but we have to get it to work in an industrial way.” What he means is that an Industrial Internet can pose a more stringent set of requirements than the consumer one. Ruh likes to say that if your cell phone drops a call, you get annoyed, but if your power goes down, you get angry, or fearful–or people in hospitals die. Therefore, reliability–a zero tolerance for platform failures–is essential in the software he builds. So too is cybersecurity. A jet engine, or a locomotive pulling a mile of natural-gas cars, has to be utterly impervious to data breaches, even as it is communicating its data to a secure source. A third challenge is durability. “People are buying machines that last 20 years,” Ruh says. That means his team has to create software with a life cycle that might seem absurd in the consumer sphere.
Even with these differences, Ruh thinks of this particular technological moment as akin to the early 1990s, when the consumer Internet was in its infancy and the future was difficult to discern. He is convinced that his software will enable industrial improvements that become immensely important to both GE and the larger global economy. Ruh says: “Industrial people who aren’t in the game today, who are not making the kind of investments we’re making, it’s like someone in the retail sector now saying, ‘We want to be like Amazon too.’ But you can’t be like Amazon. It’s too late. And in the industrial sector, you have to take the risk now. Because by the time it’s obvious, which is a few years from now, it’s going to be too late.”
As Ruh’s San Ramon lab has come together, it’s become clear to Immelt that by changing the technology of GE, he’s changing its culture, too. He tells me he has come to see that shift as inevitable. “Twenty-five or 30 years ago,” he says, “the thing that was differentiating companies was how well managed they are. But if you think about the companies that have just kicked butt in the past decade, they are deep-domain, technology-led, innovative companies–Google, Amazon, Apple. So I think this notion of, if the only common thread you have as an industrial company is the fact that you think you’re well managed, you can still be a pretty good company, but you’re not going to be a dominant company, a competitive company over time.” Immelt now wonders if GE was too reliant on the belief that a well-trained manager can do anything well. “That’s just not true,” he says, laughing. “They need to have technical tools.” The only way he saw GE continuing its dominance, in fact, was by leading in its technology–whether it was with state-of-the-art machines or the data that came off the state-of-the-art machines. Or ideally, both.
There are limits to how far GE can go in embracing the Valley ethos. A company that builds complex hardware can only take short-cycle software thinking–“move fast and break things,” as Mark Zuckerberg famously said–so far. When you’re making stuff that has to last 20 or 30 years, you can’t market a buggy Evolution 1.0 locomotive and follow it a few months later with the Evolution 2.0. Or as Immelt quips, “You don’t want your jet engine to be known as a minimum viable product.”
Still, in talking with Immelt and his colleagues, you often get the feeling that Silicon Valley’s success over the past decade has rattled them enough to prompt a reconsideration of not just GE’s weaknesses but also its strengths. As GE scrambled to add missing parts to its vast operations through the San Ramon lab, the executives came to reassess how they could do things with their scale and manufacturing prowess that pure digital organizations cannot. “I have a ton of respect for Marc Andreessen and the stuff he’s done,” Immelt tells me. At the launch of the Industrial Internet in 2012, he recalls that Andreessen had just written a hotly debated article for The Wall Street Journal about how software would “eat” the world. Andreessen wrote, “My own theory is that we are in the middle of a dramatic and broad technological and economic shift in which software companies are poised to take over large swathes of the economy.”
“He and I were being interviewed, and we had a GEnx jet engine next to us,” Immelt says. “And this is a very impressive thing. You could see that maybe the one thing that software wasn’t going to eat was that jet engine.” On the other hand, software will surely make that jet engine better and better. And that fact suggests that Andreessen and Immelt are both right–or, more precisely, both half-right.
Jet streams: In a single day, a plane outfitted with GE’s newest engines spins off a terabyte of data for analysis. To decode all that, GE needed to hire a new corps of data scientists.Photo by Adam Senatori
“If the only common thread you have as an industrial company is that you’re well managed, you can still be a pretty good company, but you’re not going to be dominant,” says Immelt.
Until recently, Immelt’s defining achievement at GE looked to be his efforts to move the company in a greener direction. Over the previous few years, GE had built up a huge wind turbine business and made a variety of other ecologically valuable efforts. But GE can only take green so far; this organization fundamentally exists to build or improve the infrastructure for the modern world, whether it runs on fossil fuels or not. Thus, GE has simultaneously enjoyed a booming business in equipment for oil and gas fracking, and has profited from the strong demand in diesel locomotives thanks to customers needing to haul coal around the country. To its credit, the company is developing hybrid and natural-gas locomotives that will burn cleaner. In the end, though, it will serve its customers, wherever they do business.
If things turn out the way he hopes, Immelt’s legacy might well be the Industrial Internet instead. “Often people will look at this 1% improvement and say, ‘You’re doing something small; this is not transformational,’ ” Marco Annunziata, GE’s chief economist and a close colleague of Immelt’s, tells me. “But I think that’s missing the point. What we’re trying to argue is that when you’re talking about such a huge base of machines, getting a 1% or 2% improvement is very sizable, and very important.” While Immelt is more circumspect about how far those improvements will go, Annunziata believes that the ultimate productivity gains from data and analytics will eclipse 1% or 2%, and might possibly reach double digits. “Also, you should remember these are early days,” he adds. “So the incremental improvements we are getting now are just the first steps.” Indeed, some executives I met within GE believe that the coming progression of smart technologies will lay the groundwork for a second great industrial age–a more distant future where intelligent machines can be upgraded into what the company calls “brilliant” machines. These devices wouldn’t just let you know they were going to break down. They would actually repair themselves.
We’re not there yet, Immelt warns. And he’s right. While GE has achieved some measurable successes so far–helping rail companies and airlines save on fuel and operations, for instance, and posting strong profits in most of its industrial businesses–it has years to go before it can claim victory. In the meantime, it’s worth asking how, if GE’s Industrial Internet succeeds, it will affect our day-to-day lives. In Immelt’s view, the impacts will be shared most immediately between GE and its customers such as the airlines and utilities; the changes for consumers may be pleasurably atmospheric–or in things we barely notice. More efficient jet engines would lead to fewer canceled flights.
Better mammography machines would lead to shorter wait times for results. As for freight trains, they might still be the dinosaurs of the digital era, but they would be surprisingly intelligent dinosaurs. Almost certainly, the locomotives would continue to haul the largest portion of the country’s freight (as they do now) and help tie together the invisible parts of the economy we take for granted. But presumably they would do that increasingly faster, smarter, cleaner, and cheaper. Ultimately, those benefits should translate to the consumer too.
The supermarket shelf stocked with cereal, the gasoline in your car’s tank, the orange juice in your refrigerator–all would depend in no small part on what the scientists in California can do to find patterns in the data spinning off those trains. And someday, over breakfast, it might occur to you that there’s something in your life, something you can’t quite put your finger on, that’s ever so slightly different, and that maybe, just maybe, things are a hairbreadth better.