Students are cheating using ChatGPT. Toys “R” Us recently produced a commercial almost entirely with generative artificial intelligence. Wall Street is incorporating AI into its investment models. Evidently, the world has fallen in love with the endless possibilities surrounding AI that could advance civilization or enslave the population. But while businesses and consumers embrace AI, energy consumption levels will come into focus – and a couple of windmill eyesores will not be enough to power the next technological revolution. Whether AI or cryptocurrency, the new tech wave will require massive data centers.
How Much Energy Does AI Use?
In January, the Paris-based International Energy Agency (IEA) reported that one Google search requires 0.3 watt-hours of electricity. By comparison, a ChatGPT query demands 2.9 watt-hours. Should ChatGPT become integrated into the roughly nine billion daily searchers, the amount of electricity consumed would equal 1.5 million European Union residents. The group estimates that data centers will account for approximately 4% of global energy use by 2026.
Other numbers are eyebrow-raising. Semiconductor research firm TechInsights noted that AI chips will represent 1.5% of global power consumption over the next five years. Hugging Face, a machine-learning company, says a generative AI system imbibes about 33 times more energy than machines running task-specific software. The Electric Power Research Institute says digital hubs will drain as much as 9.1% of domestic electricity by 2030. “At 2.9 watt-hours per ChatGPT request, AI queries are estimated to require 10x the electricity of traditional Google queries,” the non-profit organization stated. Another report suggests that AI consumes as much as 3.2 gallons of water for every kilowatt hour of energy across Microsoft’s global data centers.
Over the last decade, the tech sector has pledged to trim its emissions. However, in the wake of an AI boom, Google confirmed that emissions have soared by 48% in just five years. Microsoft’s output has climbed by roughly 30% since 2020. The chief OpenAI investor reported that it maintains more than five gigawatts of installed server capacity, exceeding the annual energy consumption of Hong Kong or Portugal.
Energy demands from AI have captured the attention of policymakers at home and abroad.
Democrats introduced the Artificial Intelligence Environmental Impacts Act of 2024, which directs the National Institute of Standards and Technology to establish standards to gauge and report the wide array of environmental effects of AI. The European Union will install a new framework next year mandating “high-risk A.I. systems” to report their energy consumption volumes and resource use.
“There is a Dickensian quality to the use of AI when it comes to our environment: It can make our planet better, and it can make our planet worse,” said Sen. Ed Markey (D-MA). “Our AI Environmental Impacts Act would set clear standards and voluntary reporting guidelines to measure AI’s impact on our environment. The development of the next generation of AI tools cannot come at the expense of the health of our planet.”
But while the world tries to fund the multi-trillion-dollar transition to green energy, the question will need to be asked: Are solar panels and windmills enough to power AI?
AI Sipping on Texas Tea
America’s electric grid will require a massive transformation to power the growth in data centers erected nationwide. Renewables have struggled to satisfy current demand, and green proponents might find it challenging to keep multi-billion-dollar data centers running with windmills and solar panels. This, experts note, could be an exceptional opportunity for crude oil and natural gas.
Phil Flynn, an energy strategist at The PRICE Futures Group, noted this past spring that oil and gas will play a critical role in AI in various ways. From producing it in cleaner ways to enhancing efficiency in locating fossil fuels, it is clear oil and natural gas will likely be necessary for years to come.
“Artificial intelligence will take a lot of the guesswork away from mapping fields, and we should be able to pinpoint specific areas to drill where we can get the maximum amount of oil for the least amount of carbon emissions,” Flynn said. “There are so many ways that AI can improve U.S. oil and gas efficiency. Already, the U.S. oil and gas industry has used extensive computer programs to map underground reservoirs, [which] artificial intelligence can improve upon. It could also improve, of course, on efficient ways to drill.”
The meat and potatoes of the discussion? Employing Texas tea and the so-called bridge fuel to power AI. In May, Goldman Sachs forecast that natural gas will supply 60% of the power demand growth from AI and data centers. Wells Fargo predicted that natural gas demand could soar by ten billion cubic feet per day by 2030. Research firm Thunder Said Energy thinks AI could add 8% to US natural gas demand by 2030.
Other commodities are proving to be critical for keeping AI up and running 24/7. AI power needs are delaying the retirement of many coal-fired power plants. Copper, which is also crucial for solar and wind installations, is necessary for the AI data centers.
Oil AI Needs Is Love
The US government has showered US and foreign companies with hundreds of billions of dollars in lavish subsidies. What do taxpayers have to show for it? Not much. Consumers are shunning electric vehicles, public charging stations are hard to find, manufacturing jobs have stalled, and wind energy generation is declining. In fact, despite the current administration’s energy policies, the oil and gas sector is booming. The industry is producing around 13 million barrels of oil per day and about 100 billion cubic feet of natural gas per day. As usual, fossil fuels will bail out the world and ensure the public can generate AI videos of Will Smith eating spaghetti.
