AI and Crypto Mining Surge Threatens North American Power Grid Stability

Crypto, AI growth could strain North American energy grids: NERC

Cryptocurrency mining and artificial intelligence operations drive the demand for electricity to new highs across North America as industries connect large-scale data centers and facilities to the energy grid. According to a North American Electric Reliability Corporation (NERC) report, this growth in electricity demand is expected to present challenges to forecasting and reliability. Crypto mining electricity use can vary, often scaling with market prices, which adds further complexity to energy grid management and sudden fluctuations in energy load requirements during normal operations. The NERC report highlights the strain on grid reliability and the increased risk of energy shortfalls posed by crypto mining and AI operations. It seeks to address the issues for the future to ensure a stable power supply to North America. The NERC's latest Long-Term Reliability Assessment shows that significant growth, particularly in regions like Texas, has forecast an increase of 4.6% annually to 2029 at peak summer demand -- four times more than previous projections. The report highlights that AI data centers and crypto mining facilitate unique challenges due to their energy-intensive nature and varying energy load behaviors. These energy demands can shift randomly, as seen when crypto-mining facilities adjust consumption based on electricity prices or when AI data centers ramp up energy use for processing, cooling, and storage. With crypto and AI becoming mainstream staples over time, their associated operations pose significant challenges to energy grid stability and reliability due to potential grid strain, particularly in peak periods or amid operational faults. In Texas, where crypto mining and AI hubs are concentrated, the Electric Reliability Council of Texas (ERCOT) reports increasing risks associated with contracted and non-contract energy loads. Sudden load changes in the crypto mining and AI industries potentially mimic issues seen with inverter-based resources, like disconnections during faults or price spikes, and introduce new risks for grid operators managing variable renewable energy resources. Related: MARA rolls out advanced ASIC recycling with wind power Strategies to address rising electricity consumption The NERC calls for proactive measures to address the increasing strain on the energy grid in North America, suggesting improved demand forecasting, advanced transmission planning, and expanded demand-side management (DSM) programs. ERCOT has implemented energy response and demand response programs to balance the energy grid load during critical periods. Texas has also introduced legislation, like Texas' HB 3390, which mandates improved distributed energy resources (DERs) tracking to improve reliability assessments. In tandem with the rising concerns, some mining firms are shifting toward renewable energy sources, like MARA's (formerly Marathon Digital) acquisition of a wind farm in Hansford Country, Texas.

AI Is Putting America's Electrical Grid at Risk

US and Canadian electrical grids are being strained by AI data centers, crypto mining, and electric cars. A watchdog for the utility industry says that a surge in demand for electricity could pose a problem for grid reliability across the United States and Canada as utilities have been slow to make necessary upgrades. The North American Electric Reliability Corporation said in a new report that electricity demand is growing faster than at any point in the past two decades at the same time as more than 115 gigawatts of coal-burning capacity is scheduled to shut down in the next ten years. The Financial Times first covered the report. “We are experiencing a period of profound change,†said John Moura, director of reliability assessment at NERC. The report estimates that peak summer demand will rise by 15%, or 132 gigawatts, in the next decade. That's a large increase over the 80 gigawatts that NERC estimated in last year's report, which the organization of course attributes to the rapid construction of data centers to power AI and crypto mining, as well as the rise of electric cars and heat pumps. Peak winter demand is expected to rise by 149 gigawatts in the next decade, or 18 percent, versus the previous estimate of 92 gigawatts. As anyone knows, infrastructure projects in the U.S. move at a glacial pace, and so it should be no surprise that the existing electrical grid was not designed to handle a world in which everything is battery-powered. Peak demand may have previously meant the times of day when everyone is at home, with some lights and a TV on. But now when many people come home they will be plugging in their car to charge its massive battery pack. Or they might have a heat pump, which can draw a lot of energy to heat and cool a homeâ€"a recent estimate from EnergySage found that the typical heat pump in a typical home uses 5,475 kilowatt-hours per year. They estimate that's enough electricity to run nine full-size fridges year round, or power a Tesla Model 3 for 15,000 miles. In order to accommodate all that demand pressuring the grid, energy generation needs to increase, but power lines and transformers also need to be upgraded. Although tech companies like Meta and OpenAI have been investing massively in new data centers for AI which still has a ways to go to prove its usefulness, one potential benefit of this boom is that there has been a renewed interest in bringing back nuclear. As it stands, solar and wind energy capacity is not enough to meet demands, and most grids do not have battery packs to store energy for use later. There's hope amongst those in tech that it's time for a return to nuclear as a clean, abundant energy source. They have also proposed locating data centers next to nuclear plants to avoid stressing the grid. That being said, nuclear energy isn't perfectâ€"the waste has to be stored somewhere, typically deep underground, and can take an extremely long time to decay. It also remains unclear whether the general population has forgotten past disasters and is ready to have nuclear reactors near their homes. In the past, a potential solution presented for the grid issue has been using electric cars themselves to power homes when the grid is facing peak demand, or send energy from the car back to the grid itself. However bidirectional systems have not been widely adopted by automakers or electrical utilities. Impulse Labs, a startup developing a new induction stovetop, has built a battery into the product that it believes could be used to store a buffer of energy so that households are not all trying to access the grid simultaneously. The idea would be that the battery is slowly charged by the grid, and then households access energy from that battery when needed. Impulse Labs believes that products like Tesla's Powerwall, a stationary home battery, have not been adopted more widely because it is expensive and appeals largely to doomsday preppers. By building a battery into a standard household appliance, Impulse believes it found a backdoor way to address the grid problem. NERC warns that under the current conditions, a shortfall in energy capacity could leave the supply buffer falling below required levels in nearly every jurisdiction across the U.S. during the decade. “Most of the North American [bulk power system] faces mounting resource adequacy challenges over the next 10 years as surging demand growth continues and thermal generators announce plans for retirement."

AI poses threat to North American electricity grid, watchdog warns

North America's electricity grid faces "critical reliability challenges" as power generation fails to keep pace with surging demand from artificial intelligence, the industry watchdog has warned. Soaring electricity consumption in the next decade, coupled with the closure of coal-fired plants, will place huge strain on US and Canadian grids, the North American Electric Reliability Corporation has found. The shortfall could cause blackouts during peak demand periods in both countries, and will be worsened by delays in adding solar generation capacity, batteries and hybrid resources to the grid, according to NERC. Some areas of the US could face shortfalls as soon as next year, it said. "Most of the North American [bulk power system] faces mounting resource adequacy challenges over the next 10 years as surging demand growth continues and thermal generators announce plans for retirement," NERC concluded in its 2024 Long Term Reliability Assessment report. The report is the latest warning that AI's voracious power needs threaten to overwhelm an already frail power grid as it struggles to keep pace with the energy transition. Electricity demand was growing quicker than at any point in the past two decades, NERC found, amid rapid construction of data centres to power AI and crypto mining, and as consumers buy electric vehicles and heat pumps. Peak summer demand would rise by 132 gigawatts, or 15 per cent, over the coming decade -- a sharp increase in last year's forecast for a rise of 80GW. Peak winter demand will rise by 149GW, or 18 per cent, versus 92GW before, NERC said. NERC is a not-for-profit body subject to oversight by the Federal Energy Regulatory Commission. The International Energy Agency estimates global power demand from data centres alone could top 1,000 terawatt hours by 2026 -- double 2022 levels and an increase equivalent to Germany's total electricity needs. Big Tech is scrambling to find ways to meet the staggering demand projections, announcing a host of new deals that have proved a boon to nuclear generation. But NERC warned that the demand surge would coincide with the winding- down of fossil fuel generation, with 115GW worth of capacity scheduled to be shut down in the next 10 years. The shortfall could leave the supply buffer falling below required levels in almost every jurisdiction during the decade, NERC warned. The Midcontinent System Operator, responsible for the grid in the US Midwest, faces potential shortfalls as early as next year.

Huge growth in AI usage could cause electricity blackouts in US

Electricity regulators are concerned that the huge growth in AI usage could lead to power blackouts in the US and Canada as early as next year. The prediction was published shortly after the public release of iOS 18.2 brought many more Apple Intelligence features online. Artificial intelligence services like ChatGPT and Apple's own Private Cloud Compute servers will see data center power usage double in the course of just four years, and power grids will likely struggle to keep up with this demand ... The concerns were expressed by the North American Electric Reliability Corporation (NERC), a non-profit regulator for the power-generation sector, reports the Financial Times. North America's electricity grid faces "critical reliability challenges" as power generation fails to keep pace with surging demand from artificial intelligence, the industry watchdog has warned [...] The shortfall could cause blackouts during peak demand periods in both countries as soon as next year. Demand for electricity is growing at an unprecedented rate, says the NERC in its Long Term Reliability Assessment report, with data centers powering AI services a key reason. The report aligns with forecasts by the International Energy Agency that data centers alone will exceed 1,000 terawatt hours by 2026 - double the levels in 2022. The challenge is made all the greater by the winding down of fossil fuel generation failing to be matched in pace by additional renewable energy power. The US Midwest is likely to be hit first, says NERC, but almost every area of the US and Canada may experience serious power outages within the next decade.

Can the US Power Industry Meet AI's Steep Energy Demands?

Delivering power to homes and companies is supposed to be a boring and predictable business. While it's true that the US population has been increasing and electrifying more things as it does, that growth in power use has been offset by energy savings as buildings, factories and appliances become more efficient. Indeed, electricity consumption in the US has changed little since the start of the 21st century. Until now. Suddenly, America needs more juice -- for factories and homes, electric vehicles and heating, and especially data centers and artificial intelligence. The surge in power demand is unlike anything utilities have seen in decades, perhaps not since World War II. That's going to complicate the country's already bumpy shift to clean energy, and Donald Trump's imminent return to the White House promises to shake up the transition even more. By all accounts, the industry isn't ready. "The impact of AI on the power grid came on fast," says Timothy Fox, an analyst at ClearView Energy Partners. "We're seeing load-growth projections that are unprecedented."