Is AI Really Killing the Polar Bears?

Artificial intelligence is being blamed for a growing list of environmental problems. One of the most dramatic claims is that using AI is helping destroy the climate — and, eventually, killing polar bears.

There is a real connection.

But it is much smaller and more complicated than that claim makes it sound.

The actual chain looks like this:

AI uses data centres. Data centres use electricity. Some of that electricity is produced using fossil fuels. Burning fossil fuels releases greenhouse gases. Those gases warm the planet. A warmer planet reduces Arctic sea ice. Polar bears depend on that ice.

Every step in that chain is real.

The question is not whether AI has an environmental impact.

The important question is:

How large is that impact compared with everything else causing climate change?

That changes the picture considerably.

---

1. AI does use a lot of electricity

Artificial intelligence requires powerful computer processors.

Training large AI models can require thousands of specialized chips operating for weeks or months. Once those models exist, millions of people asking them questions, generating images and creating video also requires electricity.

Those computers operate inside data centres.

Data centres also need cooling systems, networking equipment, storage and backup power. All of this consumes energy.

And demand is growing quickly.

Data centres accounted for about 1.5% of global electricity use in 2024, according to the International Energy Agency. Estimates suggest that their share could approach 3% by 2030 as AI, cloud computing and other digital services expand.

That is significant.

But it also means something important:

About 97% of the world’s electricity in 2030 would still be used for things other than data centres.

AI is not consuming most of the world’s electricity.

It is a new and rapidly growing part of a much larger energy system.

---

2. Electricity itself is not the main climate problem

Computers do not produce carbon dioxide simply because they are computers.

The climate impact depends largely on where their electricity comes from.

A data centre powered by coal creates much more greenhouse-gas pollution than one powered by hydroelectricity, nuclear power, wind or solar power.

That distinction is sometimes lost when people talk about the environmental impact of AI.

The real chain is not:

AI → climate change

It is closer to:

AI → electricity demand → electricity source → greenhouse-gas emissions → climate change

That makes the energy system extremely important.

If the same AI computation is powered using low-carbon electricity, its climate impact can fall dramatically.

The problem therefore existed long before AI appeared.

Much of the world still produces energy by burning coal, oil and natural gas.

---

3. Fossil fuels are still the main driver

Climate change did not begin with ChatGPT.

Human greenhouse-gas emissions have been increasing for more than a century because industrial societies burn enormous quantities of fossil fuels.

Oil powers cars, trucks, ships and aircraft.

Natural gas heats buildings and generates electricity.

Coal still produces electricity and industrial heat in many countries.

Factories use fossil fuels to produce steel, cement, chemicals and countless other products.

Agriculture and land-use changes produce additional greenhouse gases.

AI entered this enormous system very recently.

That does not mean its emissions should be ignored.

It means they should be kept in proportion.

If we are trying to understand why the planet is warming, the main problem remains the massive global dependence on fossil fuels — not any single new device or technology that uses electricity.

---

4. The scale matters

Numbers can sound frightening without context.

The International Energy Agency projects global data-centre electricity consumption to rise from about 415 terawatt-hours in 2024 to around 945 terawatt-hours in 2030.

That is a very large increase.

For comparison, global renewable generation increased by 858 terawatt-hours in 2024, including 474 terawatt-hours from solar, according to Ember.

That comparison does not make data-centre growth meaningless.

It shows that the global energy system is enormous and changing rapidly.

We are simultaneously adding:

• AI data centres,
• electric vehicles,
• heat pumps,
• industrial electrification,
• air conditioning,
• new factories,
• renewable power,
• nuclear plants,
• battery storage,
• and billions of additional electrical devices.

The important question is therefore not simply:

How much electricity will AI use?

It is:

What kind of electricity will supply that demand?

If growing demand is supplied mainly by fossil fuels, emissions increase.

If it is supplied by low-carbon energy, the climate impact is much smaller.

---

5. What about all the water?

AI is also criticized for using water.

Again, there is a real issue underneath the headlines.

Some data centres use water for cooling. Electricity generation can also use large quantities of water.

But water is different from carbon dioxide.

Carbon dioxide released in one country mixes into the global atmosphere and contributes to global climate change.

Water use is mainly a local problem.

A data centre using large amounts of water in a drought-prone region can create serious problems for the surrounding community.

The same data centre located where water is plentiful, using recycled water or using a different cooling system may have a much smaller effect.

So simply stating that “AI uses water” does not tell us whether the environmental damage is large or small.

Location, technology and water availability matter.

That is a reason to improve how data centres are built and regulated.

It is not evidence that AI is one of the world’s main environmental threats.

---

6. We have seen this pattern before

New technologies often attract disproportionate attention because they are unfamiliar.

Railways were once blamed for frightening livestock and damaging the countryside.

Factories, automobiles, television, computers, the internet and smartphones all produced waves of concern when they began changing society.

Some of those concerns were justified.

Some were exaggerated.

Usually the technology eventually became part of a much larger system, and society learned how to regulate it, improve it and reduce its harmful effects.

AI is going through that stage now.

It is highly visible.

It is changing quickly.

Many people do not yet understand how it works.

That makes it an easy target.

Meanwhile, the older and much larger environmental problems can begin to seem normal because we have lived with them for decades.

A gasoline-powered car does not feel like new technology.

A natural-gas furnace does not feel unusual.

A coal-fired power station may have operated for fifty years.

But their familiarity does not make their emissions disappear.

There is a danger in becoming so focused on the small new source of energy demand that we lose sight of the much larger system supplying that energy.

---

7. So what does this have to do with polar bears?

Polar bears depend heavily on Arctic sea ice.

They use it as a platform for hunting seals and moving across large areas.

As the planet warms, Arctic sea ice declines.

That creates serious challenges for polar-bear populations in parts of the Arctic.

Human-caused greenhouse-gas emissions are driving that warming.

AI contributes to those emissions when the electricity used to operate AI systems comes from fossil fuels.

So it would be wrong to say AI has no connection at all to climate change or polar bears.

But it would be equally wrong to treat AI as though it were the main cause.

Imagine that global greenhouse-gas emissions are a huge river.

AI is a new stream flowing into it.

That stream may grow.

We should watch it.

We should prevent it from becoming unnecessarily dirty.

But stopping that stream while leaving the enormous river untouched would not solve the problem.

The environmental goal should therefore be bigger than making AI more efficient.

We need more efficient data centres.

We need better cooling systems.

We need careful decisions about where data centres are built.

We need more renewable and nuclear electricity.

And most importantly, we need to continue replacing the fossil fuels that are responsible for most of the warming already taking place.

AI should be part of that environmental discussion.

It should not become a distraction from it.

Making AI cleaner matters. Replacing fossil fuels matters much more.