Tech Revolution or Nature’s Nightmare? Computing the environmental costs of AI

By Andrea Zanon

The rapid rise of artificial intelligence (AI) investment is a double-edged sword. While it promises unprecedented advancements, the environmental cost of AI’s insatiable hunger for energy is reaching a tipping point. Investment in AI has driven Big Tech share prices to fresh highs, creating the idea that AI-focused companies will grow indefinitely.

Meanwhile, investment in data centres, the lifeblood of AI operations, were estimated to attract about US$350 billion in investments in 2023 and that number is projected to balloon to $1 trillion annually by 2027. OpenAI CEO Sam Altman has even called for an immediate $5 trillion investment, representing five per cent of global GDP, to fuel further AI development. Jensen Huang, CEO of AI chipmaker NVIDIA, anticipates that corporations could spend $1 trillion over the next four years to upgrade and expand data centre infrastructure to meet the growing demand from AI across sectors.

Data centres devour energy

Large data centres, the backbone of AI operations and a core component of training and operating AI models such as Google’s Gemini or OpenAI’s GPT-4, allocate roughly 45 per cent of their energy consumption for cooling purposes. The huge energy demand is increasingly straining power grids, particularly in the United States and Canada, raising concerns about the environmental impact of the rising computing power needs. For instance, the state of Virginia in the U.S., which has the largest data centre hub in the world, has switched part of its energy infrastructure to coal-generated energy as it cannot keep up with energy demand from the growing data centre business deployed in this state. This has abysmal repercussions on the decarbonization plans set to reach Net Zero in 2040 in the U.S. Part of the switch back to dirtier energy feeds is also enhanced by the worsening of droughts, which continue to cost the U.S. and Canada dearly.

Decarbonization catastrophe coming?

Once a champion of environmental responsibility and clean technology, Google now admits a 48 per cent surge in carbon emissions over five years. Its energy-related emissions alone soared by 37 per cent in 2023, primarily driven by its investment in data centres. This contradicts its pledge to achieve net-zero emissions by 2030, raising concerns about the ability of corporate giants to sustain their green-growth efforts while pushing their corporate priorities. Similarly, Microsoft, the largest investor in ChatGPT developer OpenAI, is admitting that its 2030 net zero “moonshot” might not succeed owing to its AI strategy.

AI’s relentless growth is set to increase electricity demand by up to 20 per cent by 2030. In the face of this skyrocketing demand, and with the limited availability of renewable energy sources, there are concerns that regions, particularly in lower-income countries, will be forced to revert to carbon-intense energy sources like coal. This alarming trend is already evident in the state of Virginia, a hub for data centres, where coal is making a comeback to meet the growing energy needs.

According to the International Energy Agency (IEA), data centre energy use could double from 1,000 terawatt hours while according to SemiAnalysis, AI will result in data centres consuming at least four per cent of global energy generation by 2030. Water usage is significant too, with one study estimating that AI could correspond to about 6.6 billion cubic meters of water use by 2027, nearly two-thirds of the UK’s annual water needs.

While renewable energy offers a scalable alternative, challenges remain. To showcase the cleantech momentum, cleantech investment in 2023 was approximately $1.9 trillion (the fastest pace recorded in the past 20 years) compared to $1.1 trillion in conventional energy. Having said that, delays in renewable energy infrastructure development and the sheer scale of energy delivery required for data centres raise questions about the short-term viability of cleaner energy sources. In 2023, at COP28 in the United Arab Emirates, global governments committed to triple the world’s renewable energy resources by the end of the decade. However, when looking beyond the pledges, most countries, including Canada, will only be able to double their renewable energy production by 2030. This was confirmed by the IEA in its 2024 analysis.

Most tech giants, including Amazon, Apple, Meta, Google, and Microsoft, are acknowledging the problem, reporting an increase in emissions since 2020, largely attributed to data centre expansion and AI expansion. This admission underscores the undeniable link between unchecked AI growth and environmental degradation. Other activist corporate giants such as Apple, Amazon, and Coca-Cola joined the First Mover Coalition of the Davos World Economic Forum (WEF) and invested in green coal, green hydrogen, and green cement, among others, to send a market signal that top corporations are committed to the green growth transition. In Canada, several sectoral experts said that electricity needs of AI could catapult global tech giants to take a leading role in the energy transition and investment. Tech companies in Canada are already investing in wind, solar, nuclear, and geothermal and could become among the largest funders of clean electricity projects.

Canada’s role in AI and quantum computing

Canada is uniquely positioned to play a significant role in addressing the challenges posed by AI’s energy consumption. As one of the leading countries globally for incubating new startups in both artificial intelligence and quantum computing, Canada has the potential to drive innovation in energy-efficient AI technologies. The country’s expertise in quantum computing could lead to breakthroughs in developing more energy-efficient algorithms and hardware for AI applications. Canadian startups and research institutions are at the forefront of exploring ways to optimize AI models and reduce their environmental footprint, potentially setting new industry standards for sustainable AI development.

The Canadian government and governments, corporations, and individuals around the world must act decisively to steer AI towards a sustainable path. This includes accelerating investment in renewable energy infrastructure, developing more energy-efficient AI models, creating more efficient chip designs, and demanding transparency from tech companies about their environmental footprint and commitments. The time for complacency is over. We can either harness AI responsibly or expect to pay a huge environmental cost via increased frequency and intensity of climate-led natural hazards. This cost has been estimated to range between 14-20 per cent of global GDP by 2050 by Swiss-Re Reinsurance Group in 2022, and Nature in 2024.

As we prepare for another Climate Conference of the Parties, COP29, which this year will be hosted by Azerbaijan, a major producer of oil and natural gas ranking among the world’s top producers, we should expect more pressure and action on renewable energy, enforcing stricter energy efficiency standards, and holding corporations accountable on their Net-Zero plans.