This story originally appeared in Techonomy’s Spring 2019 magazine.

As software-based efficiency cascades across the economy, every industry is disrupted, but the impact varies tremendously. The accelerating drumbeat of artificial intelligence (AI), 5G and the Internet of Things alters the supply of goods and services. It drives down manufacturing costs and optimizes existing supply. The key concept in an age of software optimization is “capacity utilization.” The industries and companies that can best adapt as resources become used more efficiently will be those that thrive in an age of AI.

Every company has to obsess over how it can leverage digital technology to improve quality, reduce costs, or create new products. But it also matters greatly whether their industry faces growing or declining demand. What follows is a look at radically different prospects of several industries.

The Winners:

1. Airlines

This industry has always been challenging, because planes represent a huge fixed cost. The inventory–seats– becomes as useless as rotten fruit the minute a jet pulls away from the gate. In a less optimized era, airlines had to sell the last seat for 10 cents, or it went empty.

Predictive analytic software has changed everything. Airlines can now predict with astonishing accuracy how many people will buy a ticket on a specific day. When you buy one, there might be a $50 difference between two seats next to each other—and the price will probably change tomorrow. Things are so rosy at American Airlines, thanks partly to technology, that the company’s CEO, Doug Parker, not so long ago boldly declared: “We’ll never lose money again.”

Such productive efficiency gains could go further. Airlines would love to customize pricing based on a specific person’s profile, including where they live. But for all the opportunities to profit even more, they must proceed gingerly, as privacy concerns grow.

2.Industrial and heavy equipment

Technology generates vast amounts of data that can be used to predict maintenance needs for complex machinery before it malfunctions. In recent years, truck-maker Volvo introduced a system that remotely anticipates and diagnoses problems with fleets—enabling managers in remote offices to arrange efficient repairs. That saves customers time and money.

But there’s a potential negative consequence for suppliers: as technology improves the durability of its products, there may be less demand to replace older equipment. This is what happened when tire-makers introduced radial tires: improving a product too much can make it hard to capture quality improvements with higher prices. So equipment makers have two options. They can convince customers that the new predictive maintenance features are worth paying for. Or makers can transition to “equipment as a service,” making it more like software and charging an annual maintenance fee that includes equipment upgrades. With the service option, manufacturers capture more of the economic value generated by lower costs and improved quality.

The Losers:

1.Utilities

Back in the 1920s, utilities became America’s largest industry as electrification swept the country and drove the decade’s boom.  Now, the utilities sector—including electricity—is in decline. The market cap of the entire industry is less than that of Microsoft.  To survive, utilities depend on support from governments, which do things like guarantee their return of capital. But now utilities are highly vulnerable to technological disruption—especially machine-learning-based improvements in energy efficiency.

For example, Alphabet’s DeepMind Technologies uses AI to cut energy consumption at Google’s massive data centers, which require lots of energy. Power is needed for many thousands of servers, which must also be kept cool. Google feeds software years of historical data on computing loads, sensor readings and the environment. That trains neural networks to control cooling equipment, enabling Google to cut energy as much as 40%.

Even as these efficiencies broaden, massive public pressure to reduce carbon emissions is likely to contribute further to reducing demand from both business and consumers. There’s little utilities can do, except be thankful for government.

2.Autos

The industry is being transformed—in good ways, and bad—by the emergence of tech-enabled services like Uber and Lyft. Personal cars aren’t used many hours each day. But between 2014 and 2018, ride-sharing increased net demand for auto-based transportation services by about 4%—as measured by vehicle miles traveled per person. The global future of ride-sharing services remains favorable, because they’re making transportation more accessible to more people.

But this increase in car utilization isn’t necessarily good news for automakers or their suppliers. Here’s the main reason: once someone buys a car, they just pay for variables like gas, and maintenance or service. They don’t go back to an automaker regularly to buy anything else. Plus, car quality keeps improving. The average car in the U.S. stays on the road more than 11 years.

What can the auto industry do? Not much. There isn’t always a growth answer.

Alec Ellison is a longtime tech-oriented investment banker and investor. Steven Gray is Techonomy’s Head of Content.