AI has a massive appetite for power.
Data centers fueling AI consumed about 1.5% of global electricity last year. The International Energy Agency predicts that could more than double by 2030.
That increase likely means burning more fossil fuels: • Coal • Natural gas • Other polluting energy sources
More fossil fuels mean more greenhouse gases. That contributes to warming temperatures, rising sea levels, and extreme weather.
But there's another side to the story.
AI's computing power can analyze energy usage and pollution. When applied correctly, it makes systems more efficient.
Current applications include: • Making buildings more efficient • Charging devices at optimal times • Reducing oil and gas production pollution • Scheduling traffic lights to cut vehicle emissions
"I am pretty optimistic that while more and more AI use is going to continue to increase, we're going to see our ability to process be much more efficient," said Alexis Abramson, dean of Columbia University's Climate School.
She believes efficiency gains will prevent energy consumption from rising as much as predictions suggest.
Buildings are responsible for one-third of U.S. greenhouse gas pollution.
AI automatically adjusts: • Lighting • Ventilation • Heating • Cooling
It bases decisions on weather data, electricity usage, and other factors.
Bob French from building automation company 75F explains the advantage: AI schedules air conditioning and heating around when workers actually arrive and leave.
That beats manual thermostat adjustments. People tend to blast the AC to quickly cool a space. That wastes energy.
Automated systems are especially useful for smaller buildings where overhauling the entire HVAC system isn't cost-effective.
AI can schedule when devices charge based on: • Grid demand • Electricity prices • Renewable energy availability
Charging electric vehicles overnight when demand is low saves money and reduces strain on the power grid.
Even though renewable energy is growing, oil and gas production continues.
AI helps make that production less polluting by: • Predicting equipment failures before they leak • Optimizing drilling patterns • Reducing methane emissions
Traffic congestion wastes fuel and increases emissions.
AI-powered traffic management systems: • Analyze real-time traffic patterns • Adjust light timing dynamically • Reduce idling time • Cut vehicle emissions
AI detects when equipment is about to fail.
Fixing problems before breakdowns: • Prevents energy waste from inefficient operations • Reduces emergency repairs that consume extra resources • Extends equipment lifespan
Here's the big question: Will AI's efficiency gains outweigh its own energy consumption?
Nobody knows for sure yet.
On one side: AI data centers are consuming exponentially more power. Training large language models requires massive computational resources.
On the other: AI is identifying inefficiencies across entire industries. Small percentage improvements across millions of buildings, vehicles, and industrial processes add up.
Abramson's optimism is based on efficiency improvements in AI processing itself. As the technology gets better at doing more with less power, the energy equation changes.
AI's environmental impact isn't black and white.
Yes, it consumes tremendous amounts of energy. That's a real problem, especially if that power comes from fossil fuels.
But AI is also enabling efficiency breakthroughs that weren't possible before. Analyzing billions of data points to optimize energy use across entire systems requires computing power humans don't have.
The key is ensuring AI's efficiency benefits outpace its own consumption.
That requires: • Making AI processing more efficient • Powering data centers with renewable energy • Deploying AI for environmental applications, not just chatbots and image generators • Measuring actual impact, not just theoretical benefits
For investors and policymakers, this means AI's environmental story is complicated. It's not just a climate villain or a climate hero. It's both.
The challenge is maximizing the benefits while minimizing the costs. Building more data centers powered by coal would be disastrous. Building data centers powered by renewables while using AI to cut emissions everywhere else could be net positive.
The technology itself is neutral. How we deploy it determines whether AI becomes part of the climate solution or makes the problem worse.
Right now, we're in the experimental phase. Scientists are finding promising applications. But scaling those solutions to match AI's growing energy footprint remains an open question.
The optimists believe efficiency wins. The pessimists worry consumption will outpace savings. The realists say it depends entirely on choices we make in the next few years about energy sources and AI deployment priorities.