AI’s Environmental Reckoning

Artificial intelligence is increasing emissions, undermining renewable energy gains and threatening our environment.

Artificial intelligence presents significant opportunities and benefits. It’s already helping doctors diagnose disease, scientists model climate change and businesses become more productive. But it has also arrived at a pivotal moment in the global energy transition. Just as the world is finally building renewable energy at the scale needed to replace fossil fuels, AI is creating extraordinary new demand for electricity. The question is no longer whether we can replace coal and gas. It’s whether we can build renewable energy fast enough to keep up with AI.

Behind every AI query is a physical machine. Behind the machine is a data centre. Behind the data centre is electricity, water, concrete, steel, rare minerals, diesel backup, transmission lines, cooling systems and waste heat. And the climate and environmental impacts are now arriving.

Microsoft’s latest figures show the problem clearly. Its climate pollution jumped by whopping 25% in 2025, largely because of the rapid expansion of data centres needed to power its artificial intelligence. Microsoft promised to become carbon negative by 2030, but its emissions are moving in the opposite direction. Its electricity use is rising sharply as its AI infrastructure expands.

And let's be clear, this isn't just a Microsoft problem. Google, Amazon, OpenAI, Meta, Tesla and the rest of the AI economy all depend on enormous computing infrastructure. The difference is that Microsoft’s numbers give us a rare glimpse of what's happening behind the marketing language. The cloud is not a cloud. It’s a warehouse full of machines, running hot, day and night.

The International Energy Agency says global data centre electricity use could more than double by 2030, reaching around 945 terawatt hours. That’s more than Japan uses today.

That matters because we’re not building clean energy in a vacuum. Every new solar farm, wind farm and battery is already needed to replace coal and gas, electrify homes, clean up transport, power industry and reduce bills. If AI data centres continue adding enormous new demand, they will consume a growing share of new renewable generation. Instead of renewables rapidly displacing fossil fuels, a large chunk of the clean energy dividend will be diverted to powering AI. The challenge is thus no longer just replacing fossil fuels. It’s building renewable energy faster than AI consumes it.

Australia isn't insulated from this. Data centre demand in the National Electricity Market is forecast to triple to nearly 12 TWh by 2030, roughly 6% of total electricity. By 2049-50 it could reach about 34 TWh, around 12% of grid-supplied electricity. That isn't some minor load. Australia is adding another major energy-intensive industry just as we are trying to decarbonise the entire economy. Major new industrial demand is arriving at the same time households are being told to electrify everything, while coal plants are retiring, transmission is delayed, and governments are already struggling to build enough renewable capacity.

And yes, punters will pay. When large new electricity users plug into the grid, someone has to pay for the generation, poles, wires, firming, grid balancing and transmission upgrades. In the United States, data centre growth is already being linked to grid stress, congestion and higher system costs, especially in places like northern Virginia where data centres are concentrated. Australia can't pretend the same thing cannot happen here.

Then there are the local impacts. Data centres don’t just consume electricity. They take up space and produce heat. A recent study estimated that land surface temperatures increased by about 2°C on average after AI data centres began operating, creating what researchers called a “data heat island effect”.

That matters in Western Sydney, Melbourne’s growth corridors, western Brisbane, outer Perth and other already hot urban areas. We know heat isn't distributed fairly. Poorer communities, renters, elderly people, outdoor workers, First Nations communities and people without air conditioning suffer first and worst. A data centre might be marketed as “digital infrastructure”, but locally it can mean more heat, more pressure on water, more grid demand and more diesel backup risk.

Water is another problem. Data centres often need large volumes of water for cooling, especially in hotter climates. Sydney data centres already use billions of litres of potable water each year, and water industry reports are now warning that planning frameworks need to deal with cooling, recycling and cost recovery properly.

None of this means AI has no useful role. It can help with medical research, climate modelling, biodiversity monitoring, translation, disability access and some forms of productivity. But the idea that every chatbot, image generator, advertising tool and corporate automation product deserves unlimited energy and water is absurd.

The climate test should be simple: if AI companies want to build giant data centres, they should have to prove they’re not making the transition harder. That means no fossil fuel-powered data centres. No creative accounting with renewable certificates that let companies claim green power while still drawing from fossil-heavy grids. No approval unless new renewable generation, storage and transmission are additional, local and delivered on time. No draining drinking water supplies - especially when recycled water or dry cooling are possible. No dumping waste heat into already overheated suburbs. No socialising the costs while privatising the profits.

AI may well be part of our future. But if its growth consumes the renewable energy we need to replace fossil fuels, increases electricity prices and places new pressure on water, communities and ecosystems, then we have a serious policy problem.

Microsofts emissions rose by 25% last year.