GE Predix — The Industrial AI Platform
The $7B bet that building a platform is easier than solving a problem
What They Said
GE positioned Predix as “the operating system for the Industrial Internet” — a cloud platform that would use AI and IoT data to optimize everything from jet engines to power turbines to oil rigs. CEO Jeff Immelt declared GE would become “a top 10 software company by 2020” and invested over $7B in GE Digital, the division built to deliver this vision. Predix was the centerpiece: an AI-powered platform that would analyze sensor data from industrial equipment and predict failures before they happened.
What Actually Happened
By 2018, GE Digital was bleeding cash, Predix adoption was anemic, and the company wrote down billions in losses. Immelt was replaced. His successor, John Flannery, slashed GE Digital’s headcount by 25%. His successor, Larry Culp, effectively unwound the entire initiative — GE Digital was restructured, Predix was repositioned as a narrow tool rather than a platform, and GE abandoned its ambition to be a software company entirely.
The platform attracted fewer than 25,000 developers — a fraction of what AWS, Azure, or even niche IoT platforms commanded. Customers reported that Predix was expensive, difficult to integrate with existing industrial systems, and tried to replace tools they already had rather than complement them.
The Root Cause
GE tried to build a horizontal platform when their customers needed vertical solutions. Predix was designed to be everything — data ingestion, analytics, AI, visualization, app development — for every industrial use case simultaneously. But a power plant operator doesn’t want a platform. They want to know when turbine bearing 7B on unit 3 is going to fail so they can schedule maintenance during the planned outage in April.
The scope creep was terminal. Instead of building a focused predictive maintenance tool for jet engines (which GE understood deeply), validating it, and expanding, GE built an entire cloud platform competing with AWS and Azure while simultaneously trying to serve aviation, power, healthcare, and oil & gas. Each industry vertical had different data formats, different regulatory requirements, different equipment lifecycles, and different buyer expectations. Predix tried to abstract all of that away, and in doing so, solved none of it well.
The Pattern to Watch For
The platform temptation kills more enterprise AI initiatives than any technical failure. The pattern: a company has genuine domain expertise in one area, builds AI that works in that area, and then says “let’s build a platform so everyone can benefit.” The platform ambition multiplies scope by 10x, dilutes engineering focus, delays time-to-value for the original use case, and creates a product that’s too generic to win in any specific market.
If your AI roadmap includes the word “platform” in the first 18 months, you’re at risk. Build the tool first. Prove it works. Sell it. Only then consider whether the underlying infrastructure is worth generalizing.
What You Should Steal
GE’s aviation division — separate from Predix — actually built excellent predictive maintenance AI for its own jet engines. The difference: it was built by domain experts for a specific fleet of equipment, integrated with existing maintenance workflows, and measured against specific operational metrics (unplanned downtime reduction). That narrow, focused system delivered real value while the grand platform vision burned billions. The lesson is the oldest one in enterprise software: solve one problem completely before trying to solve all of them.