To understand why General Electric is plowing $1 billion into the idea of using software to transform industry, put yourself in the shoes of Jeff Immelt, its CEO.

As recently as 2004, GE had reigned as the most valuable company on the planet. But these days, it’s not even the largest in America. Apple, Microsoft, and Google are all bigger. Software is king of the hill. And, as Immelt came to realize, GE is not that great at software.

Internal surveys had discovered that GE sold $4 billion worth of industrial software a year—the kind used to run pumps or monitor wind turbines. That’s as much as the total revenue of Salesforce.com. But these efforts were scattered and not always state-of-the-art. And that gap was turning dangerous. GE had always believed that since it knew the materials and the physics of its jet engines and medical scanners, no one could best it in understanding those machines. But companies that specialize in analytics, like IBM, were increasingly spooking GE by figuring out when big-ticket machines like a gas turbine might fail—just by studying raw feeds from gauges or vibration monitors.

This was no small thing. GE sells $60 billion a year in industrial equipment. But its most lucrative business is servicing the machines. Now software companies were looking to take a part of that pie, to get between GE and its largest source of profits. As Immelt would later say, “We cannot afford to concede how the data gathered in our industry is used by other companies.”

In 2012, GE unveiled its answer to these threats, a campaign it calls the “industrial Internet.” It included a new research lab across the bay from Silicon Valley, where it has hired 800 people, many of them programmers and data scientists.

“People have told companies like GE for years that they can’t be in the software business,” Immelt said last year. “We’re too slow. We’re big and dopey. But you know what? We are extremely dedicated to winning in the markets we’re in. And this is a to-the-death fight to remain relevant to our customers.”

Peter Evans, then a GE executive, was given the job of shaping what he calls the “meta-narrative” around GE’s big launch. Industrial companies, which prize reliability, aren’t nearly as quick to jump for new technology as consumers. So GE’s industrial-Internet pitch was structured around the huge economic gains even a 1 percent improvement in efficiency might bring to a number of industries if they used more analytics software. That number was fairly arbitrary—something safe, “just 1 percent,” recalls Evans. But here Immelt’s marketing skills came into play. “Not ‘just 1 percent’,” he said, flipping it around. GE’s slogan would be “The Power of 1 Percent.”

In a stroke, GE had shifted the discussion about where the Internet was going next. Other companies had been talking about connecting cars and people and toasters. But manufacturing and industry account for a giant slice of global GDP. “All the appliances in your home could be wired up and monitored, but the kind of money you make in airlines or health care dwarfs that,” Immelt remarked.

There is another constituency for the campaign: engineers inside GE. To them, operational software isn’t anything new. Nor are control systems—even a steam locomotive has one. But here Immelt was betting they could reinvent these systems. “You do embedded systems? My God, how boring is that? It’s like, put a bullet in your head,” says Brian Courtney, a GE manager based in Lisle, Illinois. “Now it’s the hottest job around.” At the Lisle center, part of GE’s Intelligent Platforms division, former field engineers sit in cubicles monitoring squiggles of data coming off turbines in Pakistan and oil rigs in onetime Soviet republics. Call this version 1.0 of the industrial Internet. On the walls, staff hang pictures of fish; each represents a problem, like a cracked turbine blade, that was caught early. More and more, GE will be using data to anticipate maintenance needs, says Courtney.

A challenge for GE is that it doesn’t yet have access to most of the data its machines produce. Courtney says about five terabytes of data a day comes into GE. Facebook collects 100 times as much. According to Richard Soley, head of the Industrial Internet Consortium, a trade group GE created this year, industry has been hobbled by a “lack of Internet thinking.” A jet engine has hundreds of sensors. But measurements have been collected only at takeoff, at landing, and once midflight. GE’s aviation division only recently found ways to get all the flight data. “It sounds crazy, but people just didn’t think about it,” says Soley. “It’s like the Internet revolution has just not touched the industrial revolution.”

GE is trying to close that gap. Its software center in San Ramon created an adaptation of Hadoop, big-data software used by the likes of Facebook. GE also invested $100 million in Pivotal, a cloud computing company. On the crowdsourcing site Kaggle, it launched public competitions to optimize algorithms for routing airline flights, which can save fuel.

All this could sound familiar to anyone who works with consumer Internet technology, acknowledges Bernie Anger, general manager of GE’s Intelligent Platforms division. But he says GE is thinking about what to do next to use connectivity, and more computers, to inject “new behavior” into machines. He gives the example of a field of wind turbines that communicate and move together in response to changes in wind. “We are moving into big data, but it’s not because we want to become Google,” he says. “It’s because we are dramatically evolving manufacturing.”